fix(db): bump temporary-notes migration timestamp past share-aliases (#201)

develop merged 20260626T130000-share-aliases; rename this PR's migration to
20260626T140000 so the two no longer share a timestamp prefix.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

.env.example

@@ -187,3 +187,11 @@ MCP_DOCMOST_PASSWORD=
 # Per-request output-token ceiling for the anonymous assistant (default: 512).
 # Worst-case output per accepted call = agent steps (5) × this value.
 # SHARE_AI_MAX_OUTPUT_TOKENS=512
+#
+# Second cost backstop: a cluster-wide per-workspace rolling-DAY token budget
+# (input re-sent per step + output, summed across every accepted turn). The
+# hourly request cap above bounds how MANY calls run, not how expensive each is,
+# so this caps the owner's actual provider bill directly. Like the request cap it
+# FAILS CLOSED if Redis is unavailable (default: 1,000,000 tokens per workspace
+# per rolling day).
+# SHARE_AI_WORKSPACE_TOKEN_BUDGET_PER_DAY=1000000

.github/workflows/develop.yml

@@ -56,3 +56,160 @@ jobs:
           tags: ${{ env.IMAGE }}:develop
           cache-from: type=gha,scope=develop-amd64
           cache-to: type=gha,scope=develop-amd64,mode=max,ignore-error=true
+
+  # e2e jobs run on every develop push but DO NOT gate the build/publish above:
+  # `build` stays `needs: test` only, so the :develop image still ships even if
+  # e2e fails. A failing e2e job turns the run red and triggers GitHub's email
+  # to the pusher — that red run + email is the intended notification, not a
+  # deploy block.
+  e2e-server:
+    runs-on: ubuntu-latest
+    env:
+      DATABASE_URL: postgresql://docmost:docmost@localhost:5432/docmost
+      REDIS_URL: redis://localhost:6379
+      APP_SECRET: ci-e2e-secret-change-me-min-32-characters
+      APP_URL: http://localhost:3000
+    services:
+      postgres:
+        image: pgvector/pgvector:pg18
+        env:
+          POSTGRES_DB: docmost
+          POSTGRES_USER: docmost
+          POSTGRES_PASSWORD: docmost
+        ports:
+          - 5432:5432
+        options: >-
+          --health-cmd "pg_isready -U docmost"
+          --health-interval 5s
+          --health-timeout 5s
+          --health-retries 20
+      redis:
+        image: redis:7
+        ports:
+          - 6379:6379
+        options: >-
+          --health-cmd "redis-cli ping"
+          --health-interval 5s
+          --health-timeout 5s
+          --health-retries 20
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up pnpm
+        uses: pnpm/action-setup@v4
+
+      - name: Set up Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Build editor-ext
+        run: pnpm --filter @docmost/editor-ext build
+
+      - name: Run migrations
+        run: pnpm --filter ./apps/server migration:latest
+
+      - name: Run server e2e
+        run: pnpm --filter ./apps/server test:e2e
+
+  # Same rationale as e2e-server: this job is intentionally NOT in
+  # `build.needs`. Deploy of the :develop image must not be blocked by e2e;
+  # a red run plus GitHub's email to the pusher is the notification mechanism.
+  e2e-mcp:
+    runs-on: ubuntu-latest
+    env:
+      DATABASE_URL: postgresql://docmost:docmost@localhost:5432/docmost
+      REDIS_URL: redis://localhost:6379
+      APP_SECRET: ci-e2e-secret-change-me-min-32-characters
+      APP_URL: http://localhost:3000
+      NODE_ENV: production
+    services:
+      postgres:
+        image: pgvector/pgvector:pg18
+        env:
+          POSTGRES_DB: docmost
+          POSTGRES_USER: docmost
+          POSTGRES_PASSWORD: docmost
+        ports:
+          - 5432:5432
+        options: >-
+          --health-cmd "pg_isready -U docmost"
+          --health-interval 5s
+          --health-timeout 5s
+          --health-retries 20
+      redis:
+        image: redis:7
+        ports:
+          - 6379:6379
+        options: >-
+          --health-cmd "redis-cli ping"
+          --health-interval 5s
+          --health-timeout 5s
+          --health-retries 20
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+
+      - name: Set up pnpm
+        uses: pnpm/action-setup@v4
+
+      - name: Set up Node
+        uses: actions/setup-node@v4
+        with:
+          node-version: 22
+          cache: pnpm
+
+      - name: Install dependencies
+        run: pnpm install --frozen-lockfile
+
+      - name: Build editor-ext
+        run: pnpm --filter @docmost/editor-ext build
+
+      - name: Build server
+        run: pnpm server:build
+
+      - name: Build mcp
+        run: pnpm --filter @docmost/mcp build
+
+      - name: Run migrations
+        run: pnpm --filter ./apps/server migration:latest
+
+      - name: Start server (prod)
+        # Capture stdout/stderr so a start-up crash (bind error, stack trace,
+        # migration mismatch) is diagnosable; without this the only signal is
+        # the generic health-loop timeout below, ~120s later.
+        run: pnpm --filter ./apps/server start:prod > /tmp/server.log 2>&1 &
+
+      - name: Wait for server health
+        run: |
+          for i in $(seq 1 60); do
+            if curl -fsS http://localhost:3000/api/health > /dev/null; then
+              echo "Server is healthy"
+              exit 0
+            fi
+            sleep 2
+          done
+          echo "Server did not become healthy in time"
+          exit 1
+
+      - name: Dump server log on failure
+        if: failure()
+        run: cat /tmp/server.log || true
+
+      - name: Seed admin
+        run: |
+          curl -fsS -X POST http://localhost:3000/api/auth/setup \
+            -H "Content-Type: application/json" \
+            -d '{"name":"E2E","email":"e2e@example.com","password":"E2ePassword123","workspaceName":"E2E"}'
+
+      - name: Run mcp e2e
+        env:
+          DOCMOST_API_URL: http://localhost:3000/api
+          DOCMOST_EMAIL: e2e@example.com
+          DOCMOST_PASSWORD: E2ePassword123
+        run: pnpm --filter @docmost/mcp test:e2e

.gitignore

@@ -42,6 +42,7 @@ lerna-debug.log*
 .nx/installation
 .nx/cache
 .claude/worktrees/
+.claude/tmp/
 
 # TypeScript incremental build artifacts
 *.tsbuildinfo

AGENTS.md

@@ -283,37 +283,46 @@ Vite SPA. Code is organized by feature under `apps/client/src/features/*` (mirro
 
 ### Cutting a release
 
-The git tag is the source of truth for the displayed version (UI reads `git describe --tags`); the `package.json` bump is metadata only. Steps:
+The git tag is the source of truth for the displayed version (the client UI reads `git describe --tags` via `vite.config.ts`); the `package.json` bump is metadata that backs the server `/version` endpoint (`version.service.ts`).
 
-1. Make sure `main` is clean and pushed (`git status`, `git push`).
+**Golden rule — tag on `develop` first, merge to `main` afterwards.** Cut the version-bump commit on `develop`, put the tag on *that* commit, and push it. Merge `develop` into `main` later (it does not block the tag or the release). Because the tag is in `develop`'s ancestry from the moment it is created, `git describe` on `develop` — and the `ghcr.io/vvzvlad/gitmost:develop` image — reports the new version immediately, with **no back-merge dance**. Do **not** tag `main`'s merge commit; that is the mistake described in the pitfall below (we hit it twice).
+
+Steps:
+
+1. Make sure `develop` is up to date, clean, and pushed to **both** remotes (`git status`; `git push gitea develop && git push github develop`).
 2. Pick `vX.Y.Z` (SemVer): **minor** bump for a batch of features, **patch** for fixes only. Review what landed with `git log <last-tag>..HEAD --no-merges`.
-3. Bump `"version"` to `X.Y.Z` in the **root** `package.json`, `apps/client/package.json`, and `apps/server/package.json` (keep all three in sync). Leave `packages/mcp` alone — it is versioned independently. Commit with the bare version as the subject, e.g. `0.91.0` (matches past bump commits).
-4. Update `CHANGELOG.md` (Keep a Changelog format): add a `## [X.Y.Z] - YYYY-MM-DD` section summarising `git log vPREV..HEAD --no-merges` grouped by type (Breaking / Added / Changed / Fixed / Removed), and add the `compare/vPREV...vX.Y.Z` link at the bottom. Fold the bump + changelog into the release commit.
-5. Tag the release commit with a **lightweight** tag (existing release tags are lightweight): `git tag vX.Y.Z`.
-6. Push commit and tag: `git push origin main && git push origin vX.Y.Z`. Pushing the `v*` tag triggers `release.yml` (multi-arch GHCR images + a draft GitHub Release).
-7. **Back-merge the release into `develop`** so develop builds report the new version: `git checkout develop && git merge --no-ff main && git push origin develop` (push to Gitea as well if that is the canonical remote).
+3. Bump `"version"` to `X.Y.Z` in the **root** `package.json`, `apps/client/package.json`, and `apps/server/package.json` (keep all three in sync). Leave `packages/mcp` alone — it is versioned independently. Commit **on `develop`** with the bare version as the subject, e.g. `0.94.1` (matches past bump commits).
+4. For a real release (skip for a bare hotfix tag), update `CHANGELOG.md` (Keep a Changelog format): add a `## [X.Y.Z] - YYYY-MM-DD` section summarising `git log vPREV..HEAD --no-merges` grouped by type (Breaking / Added / Changed / Fixed / Removed), and the `compare/vPREV...vX.Y.Z` link at the bottom. Fold it into the bump commit.
+5. Tag that develop commit with a **lightweight** tag (existing release tags are lightweight): `git tag vX.Y.Z`.
+6. Push the branch **and** the tag to **both** writable remotes — `git push <branch>` does **not** push tags, and tags are per-remote:
+   ```bash
+   git push gitea develop && git push gitea vX.Y.Z
+   git push github develop && git push github vX.Y.Z
+   ```
+   Pushing the `v*` tag to `github` triggers `release.yml` (multi-arch GHCR images + a draft GitHub Release). The tag *must* exist on `github`, because the `:develop` and release images are built there by GitHub Actions and `git describe` on the runner only sees the tags present on `github` (not your local clone or `gitea`).
+7. Merge `develop` into `main` when ready (commonly later — this does not gate the release):
+   ```bash
+   git checkout main
+   git merge --ff-only develop   # or a merge commit if fast-forward is not possible
+   git push gitea main && git push github main
+   ```
+   The tag is already reachable from `main` (it lives in the `develop` history that `main` now contains), so `main` reports `vX.Y.Z` too — no extra tagging needed.
 
-#### Why develop keeps showing the *previous* version (and why step 7 matters)
+#### Pitfall: tagging `main` instead of `develop` (the mistake to avoid)
 
-The UI version is `git describe --tags --always` (see `vite.config.ts`), which walks **backwards from the current commit** and picks the **nearest tag reachable in that commit's ancestry**, then appends `-<commits-since-tag>-g<short-hash>`.
+`git describe --tags --always` (see `vite.config.ts`) walks **backwards from the current commit** and picks the **nearest tag reachable in that commit's ancestry**, then appends `-<commits-since-tag>-g<short-hash>`.
 
-The release tag (`vX.Y.Z`) is created on **`main`'s release merge commit**, and that commit is **not** in `develop`'s history. So until the release is back-merged, `git describe` on `develop` cannot see the new tag and falls back to the *previous* reachable tag. Result: every develop build — and the `ghcr.io/vvzvlad/gitmost:develop` image — keeps reporting e.g. `v0.91.0-NNN-g<hash>` even though `main` is already tagged `v0.93.0`. This is the classic git-flow pitfall: the version on `develop` does **not** advance just because a release was tagged on `main`.
+The wrong flow we fell into twice: merge `develop` into `main` *first*, then tag `main`'s **release merge commit**. That merge commit is **not** in `develop`'s history, so `git describe` on `develop` cannot see the new tag and falls back to the *previous* reachable one. Result: every develop build — and the `ghcr.io/vvzvlad/gitmost:develop` image — keeps reporting e.g. `v0.93.0-NNN-g<hash>` even though a release was "cut". Tagging on `develop` (the golden rule above) avoids this entirely: the tag is in `develop`'s ancestry from the start, and `main` still gets it once `develop` is merged in.
 
-Back-merging `main → develop` (step 7) pulls the tagged release commit into `develop`'s ancestry, after which develop builds correctly show `vX.Y.Z-NNN-g<hash>`. If `develop` already drifted (release tagged but never back-merged), just run step 7 now — no new tag is needed.
+Second gotcha — the tag must exist on the remote CI builds from. `git describe` names a tag **ref**, not just a commit. The `:develop` and release images are built by GitHub Actions (`develop.yml` / `release.yml`, `actions/checkout` with `fetch-depth: 0`), so the version they print depends on which tags exist **on the `github` remote** — not on your local clone or on `gitea`. `git push <branch>` does **not** push tags; push them explicitly to **each** remote (`gitea` and `github`). A tag that only lives on `gitea` is invisible to the GitHub build.
 
-##### The tag must also exist on the remote that CI builds from (multi-remote gotcha)
+If you already tagged `main` (or `develop` still shows the old version), recover without re-tagging:
 
-`git describe` names a tag **ref**, not just a commit — so the back-merge is *necessary but not sufficient*. The develop image is built by GitHub Actions (`develop.yml`, `actions/checkout` with `fetch-depth: 0`, then `git describe --tags --always`), so the version it prints depends on which tags exist **on the `github` remote**, not on your local clone or on `gitea`.
+1. Make the tagged commit reachable from `develop` — either back-merge `main → develop` (`git checkout develop && git merge --no-ff main`), or confirm the tagged commit is already an ancestor of `develop`.
+2. Make sure the tag exists on `github`: compare `git ls-remote --tags github` with `gitea`, and push the missing one (`git push github vX.Y.Z` / `git push gitea vX.Y.Z`). Pushing a `v*` tag to `github` also fires `release.yml` — expected, just be aware.
+3. Re-run the develop build (`gh workflow run Develop`, or push any commit to `develop`) so `git describe` re-resolves with the tag now in scope.
 
-This repo has two writable remotes — `gitea` (canonical, where commits land) and `github` (where the `:develop` and release images are built) — plus `upstream` (docmost, never push). **`git push <branch>` does NOT push tags**; tags must be pushed explicitly and *to each remote separately*. A release tag that only lives on `gitea` is invisible to the GitHub Actions build: even with the tagged commit fully in `develop`'s history (step 7 done), `git describe` on the GitHub runner falls back to the previous tag it *does* have, so the develop image keeps showing e.g. `v0.91.0-NNN` while `git describe` locally already says `v0.93.0-NN`.
-
-Fix / checklist when develop still shows the old version after a back-merge:
-
-1. Confirm the tag is missing on github: `git ls-remote --tags github` (compare with `gitea`).
-2. Push it there: `git push github vX.Y.Z` (and `git push gitea vX.Y.Z` if it is missing on gitea too). Note: pushing a `v*` tag to `github` also triggers `release.yml` (multi-arch GHCR images + draft Release) — expected, but be aware.
-3. Re-run the develop build (`gh workflow run Develop`, or push any commit to `develop`) so `git describe` re-resolves with the tag now present.
-
-(The `git push origin ...` in steps 6–7 above is shorthand — there is no `origin` remote here; substitute `gitea` **and** `github` as appropriate, and always push release tags to both.)
+(There is no `origin` remote here — push to `gitea` **and** `github` explicitly, and always push release tags to both.)
 
 ## Planning docs
 

CHANGELOG.md

@@ -10,8 +10,49 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ## [Unreleased]
 
+## [0.94.0] - 2026-06-26
+
+This release makes AI chat durable and fast: assistant turns are persisted to
+the database step by step and exported server-side, the desktop app no longer
+freezes at 100% CPU on long agent runs, and MCP writes are badged with
+unspoofable AI attribution. It also reworks footnotes (Pandoc-style reuse and
+per-reference back-links), hardens page moves and duplication against cycles
+and lost edits, and caps the anonymous public-share assistant with a
+per-workspace rolling-day token budget.
+
 ### Added
 
+- **Custom pretty-links for shared pages (`/l/:alias`).** A page editor can give
+  any publicly shared page a short, memorable, workspace-scoped vanity address
+  backed by a new `share_aliases` table. Hitting `/l/<alias>` issues a `302`
+  (never `301`, since the target is retargetable) to the canonical
+  `/share/<key>/p/<slug>` page; an unknown, dangling, or no-longer-readable alias
+  serves the plain SPA index so that the existence of a name never leaks. An
+  alias can be moved to another page (with a confirm-reassign guard) and the
+  foreign key is `ON DELETE SET NULL`, so deleting the target leaves a dangling
+  alias any workspace member can reclaim. (#205)
+
+- **Temporary notes — auto-move to Trash after a workspace lifetime.** A note can
+  be marked temporary so it auto-moves to Trash once a configurable workspace
+  lifetime elapses (default `DEFAULT_TEMPORARY_NOTE_HOURS` = 24h) unless made
+  permanent first. The deadline is frozen at creation time, so later changes to
+  the workspace setting never reschedule existing notes; an hourly background
+  sweep trashes notes past their deadline (children ride along). An open
+  temporary note shows a banner with a "Make permanent" rescue action; restoring
+  a note from Trash disarms the timer so it is not immediately re-trashed.
+  Operators configure the lifetime per workspace. (#201)
+
+- **Persistent AI-chat history as the source of truth + server-side export.**
+  An assistant turn is now persisted to the database step by step: the row is
+  inserted upfront as `streaming` and updated as each agent step finishes, then
+  finalized once to `completed`/`error`/`aborted`. A process that dies mid-turn
+  keeps every finished step, and a startup sweep flips any dangling `streaming`
+  row (untouched for 10 minutes) to `aborted`. Chat "Copy" now exports
+  server-side from these rows (`POST /ai-chat/export`) rather than from live
+  client state, so the export is identical whether a chat is freshly streaming,
+  just switched to, or reloaded — and is available from the first turn of a new
+  chat. (#183, #174)
+
 - **AI-agent attribution for MCP writes.** Comments (and pages) created through
   the MCP endpoint by a dedicated agent account are now badged as "AI", with
   unspoofable provenance derived from a per-user `is_agent` flag (not from the
@@ -67,6 +108,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 
 ### Fixed
 
+- **AI chat: the desktop app no longer freezes at 100% CPU on long agent runs.**
+  `useChat` re-rendered on every streamed token and `MessageItem`/`ReasoningBlock`
+  re-parsed the whole transcript markdown (marked + DOMPurify) on every delta, so
+  per-turn work grew quadratically and saturated the main thread. The stream is now
+  throttled (`experimental_throttle`) to ~20 Hz and each finalized message row /
+  markdown part / reasoning block is memoized, so a long turn no longer re-parses
+  already-finished content. (#182)
 - **Editor: caret/selection landed on the wrong line when clicking inside code
   blocks and footnotes.** The affected NodeViews rendered their non-editable
   chrome (language menu, footnotes heading, footnote number marker) before the
@@ -81,6 +129,37 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   no longer froze on the previous step's authoritative usage; the current step's
   estimate is combined per-component with `max`, so the count rises smoothly and
   never jumps backwards. (#163)
+- **AI chat: "New chat" during a streaming first turn now resets the whole
+  chat, not just the role badge.** Starting a new chat mid-stream cleared the
+  header but left the in-flight turn's messages behind, so the fresh chat opened
+  pre-populated with the previous conversation; it now fully resets. (#161)
+- **AI chat: a dropped tool argument now yields an actionable error.** When the
+  model omitted a required parameter (typically `pageId`) in a parallel/batch
+  tool call, the assistant forwarded zod's raw "expected string, received
+  undefined" text; tool inputs now return a message naming each missing/invalid
+  parameter (the JSON Schema contract is unchanged and nothing is backfilled).
+  (#190)
+- **Page move: cycle checks are now atomic and depth-bounded.** Moving a page
+  under one of its own descendants is rejected in the same transaction as the
+  update (closing a TOCTOU window where two concurrent A→B / B→A moves could
+  form a cycle), and the recursive tree-traversal CTEs carry a cycle/depth guard
+  so a pre-existing cycle can no longer spin a query. (#207)
+- **Page/editor robustness batch.** Duplicating a page now copies shared
+  attachments for every referencing page (not just the first); colliding block
+  ids are de-duplicated on import/normalize so MCP addressed edits can't hit the
+  wrong node; transient collab store failures are retried so autosave edits
+  aren't lost; and an out-of-order tree move no longer drops the moved subtree.
+  (#206)
+
+### Security
+
+- **Public share AI: per-workspace rolling-day token budget.** The anonymous
+  share assistant now caps a workspace's actual token spend (input + output,
+  summed across every accepted turn) over a trailing day, on top of the hourly
+  request cap — so a caller who evades the per-IP throttle still cannot run up
+  the owner's provider bill without bound. Cluster-wide via Redis and FAILS
+  CLOSED if Redis is down; default 1,000,000 tokens/day, overridable via
+  `SHARE_AI_WORKSPACE_TOKEN_BUDGET_PER_DAY`. (#159)
 
 ## [0.93.0] - 2026-06-21
 

README.md

@@ -114,7 +114,7 @@ community feature, with no enterprise license. Open it from the page header; the
 - 🔭 **Viewer comments** — let read-only viewers leave comments.
 - 🔭 **Password-protected pages** — protect individual pages / shares with a password.
 - 🔭 **Windows / Linux app** — native desktop app for Windows and Linux.
-- 🔭 **Mobile app** — mobile apps (iOS first, Android to follow), reusing the existing responsive web UI and editor via a Capacitor wrapper, with offline planned for later. See [docs/mobile-app-plan.md](docs/mobile-app-plan.md).
+- 🔭 **Mobile app** — mobile apps (iOS first, Android to follow), reusing the existing responsive web UI and editor via a Capacitor wrapper, with offline planned for later. See [issue #195](https://gitea.vvzvlad.xyz/vvzvlad/gitmost/issues/195).
 - 🔭 **Offline mode** — offline sync & PWA support.
 - 🔭 **Editor & UX improvements** — blocks inside tables (lists, to-do items), column layout, additional heading levels, highlight blocks, custom emoji in callouts, floating images, anchor links for page mentions, toggles (shared-page width, aside/sidebar, spellcheck, ligatures), sanitized space-tree export, and mentions in breadcrumbs.
 

README.ru.md

@@ -115,7 +115,7 @@ real-time-коллаборации Docmost, поэтому запись нико
 - 🔭 **Комментарии зрителей** — возможность комментировать для пользователей с доступом только на чтение.
 - 🔭 **Защищённые паролем страницы** — защита отдельных страниц / шар паролем.
 - 🔭 **Приложение для Windows / Linux** — нативное десктоп-приложение для Windows и Linux.
-- 🔭 **Мобильное приложение** — мобильные приложения (iOS обязательно, Android как пойдёт) на базе существующей адаптивной веб-версии и редактора через обёртку Capacitor; оффлайн запланирован на будущее. См. [docs/mobile-app-plan.md](docs/mobile-app-plan.md).
+- 🔭 **Мобильное приложение** — мобильные приложения (iOS обязательно, Android как пойдёт) на базе существующей адаптивной веб-версии и редактора через обёртку Capacitor; оффлайн запланирован на будущее. См. [issue #195](https://gitea.vvzvlad.xyz/vvzvlad/gitmost/issues/195).
 - 🔭 **Офлайн-режим** — офлайн-синхронизация и поддержка PWA.
 - 🔭 **Улучшения редактора и UX** — блоки внутри таблиц (списки, чек-листы), колоночная вёрстка, дополнительные уровни заголовков, highlight-блоки, кастомные эмодзи в callout-ах, плавающие изображения, anchor-ссылки на упоминания страниц, тоглы (ширина шары, aside/сайдбар, spellcheck, лигатуры), санитизация экспорта дерева спейса и mentions в хлебных крошках.
 

apps/client/package.json

@@ -1,7 +1,7 @@
 {
   "name": "client",
   "private": true,
-  "version": "0.93.0",
+  "version": "0.94.1",
   "scripts": {
     "dev": "node scripts/copy-vad-assets.mjs && vite",
     "build": "node scripts/copy-vad-assets.mjs && tsc && vite build",

apps/client/public/locales/en-US/translation.json

@@ -258,6 +258,7 @@
   "Copy to space": "Copy to space",
   "Copy chat": "Copy chat",
   "Copied": "Copied",
+  "Failed to export chat": "Failed to export chat",
   "Duplicate": "Duplicate",
   "Select a user": "Select a user",
   "Select a group": "Select a group",
@@ -597,6 +598,17 @@
   "Are you sure you want to permanently delete '{{title}}'? This action cannot be undone.": "Are you sure you want to permanently delete '{{title}}'? This action cannot be undone.",
   "Restore '{{title}}' and its sub-pages?": "Restore '{{title}}' and its sub-pages?",
   "Move to trash": "Move to trash",
+  "Make temporary": "Make temporary",
+  "Make permanent": "Make permanent",
+  "New temporary note": "New temporary note",
+  "Temporary note": "Temporary note",
+  "Temporary notes": "Temporary notes",
+  "Temporary note — moves to trash unless made permanent": "Temporary note — moves to trash unless made permanent",
+  "Note will move to trash unless made permanent": "Note will move to trash unless made permanent",
+  "Note is now permanent": "Note is now permanent",
+  "Temporary note lifetime (hours)": "Temporary note lifetime (hours)",
+  "A temporary note is automatically moved to trash after this many hours unless it is made permanent. The deadline is fixed when the note is created.": "A temporary note is automatically moved to trash after this many hours unless it is made permanent. The deadline is fixed when the note is created.",
+  "This temporary note moves to trash {{time}} (with its sub-pages) unless made permanent.": "This temporary note moves to trash {{time}} (with its sub-pages) unless made permanent.",
   "Move this page to trash?": "Move this page to trash?",
   "Restore page": "Restore page",
   "Permanently delete": "Permanently delete",
@@ -714,6 +726,8 @@
   "Test": "Test",
   "Available tools": "Available tools",
   "No tools available": "No tools available",
+  "Failed": "Failed",
+  "OK · {{n}}": "OK · {{n}}",
   "Created successfully": "Created successfully",
   "Deleted successfully": "Deleted successfully",
   "Clear": "Clear",
@@ -1166,8 +1180,9 @@
   "Pick an agent role whose persona the public assistant adopts. The safety rules always still apply.": "Pick an agent role whose persona the public assistant adopts. The safety rules always still apply.",
   "Built-in assistant persona": "Built-in assistant persona",
   "Minimize": "Minimize",
-  "Current context size": "Current context size",
-  "Tokens generated this turn": "Tokens generated this turn",
+  "Context size / model limit": "Context size / model limit",
+  "Context window (tokens)": "Context window (tokens)",
+  "Shown as used / total in the chat header. Leave empty to hide the limit.": "Shown as used / total in the chat header. Leave empty to hide the limit.",
   "AI agent": "AI agent",
   "Take a look at the current document": "Take a look at the current document",
   "AI agent is typing…": "AI agent is typing…",
@@ -1314,5 +1329,15 @@
   "Protocol": "Protocol",
   "How chat requests are sent and how reasoning is surfaced": "How chat requests are sent and how reasoning is surfaced",
   "OpenAI-compatible (surfaces reasoning)": "OpenAI-compatible (surfaces reasoning)",
-  "OpenAI (official)": "OpenAI (official)"
+  "OpenAI (official)": "OpenAI (official)",
+  "Custom address": "Custom address",
+  "A short, memorable link you can point at any shared page.": "A short, memorable link you can point at any shared page.",
+  "Use 2-60 lowercase letters, digits and hyphens": "Use 2-60 lowercase letters, digits and hyphens",
+  "This address is already in use": "This address is already in use",
+  "Move custom address?": "Move custom address?",
+  "Move here": "Move here",
+  "The address \"{{alias}}\" currently points to \"{{title}}\". Move it to this page?": "The address \"{{alias}}\" currently points to \"{{title}}\". Move it to this page?",
+  "The address \"{{alias}}\" is already in use. Move it to this page?": "The address \"{{alias}}\" is already in use. Move it to this page?",
+  "Failed to set custom address": "Failed to set custom address",
+  "Failed to remove custom address": "Failed to remove custom address"
 }

apps/client/public/locales/ru-RU/translation.json

@@ -257,6 +257,7 @@
   "Copy": "Копировать",
   "Copy to space": "Копировать в пространство",
   "Copied": "Скопировано",
+  "Failed to export chat": "Не удалось экспортировать чат",
   "Duplicate": "Дублировать",
   "Select a user": "Выберите пользователя",
   "Select a group": "Выберите группу",
@@ -606,6 +607,17 @@
   "Are you sure you want to permanently delete '{{title}}'? This action cannot be undone.": "Вы уверены, что хотите окончательно удалить '{{title}}'? Это действие невозможно отменить.",
   "Restore '{{title}}' and its sub-pages?": "Восстановить '{{title}}' и её подстраницы?",
   "Move to trash": "Переместить в корзину",
+  "Make temporary": "Сделать временной",
+  "Make permanent": "Сделать постоянной",
+  "New temporary note": "Новая временная заметка",
+  "Temporary note": "Временная заметка",
+  "Temporary notes": "Временные заметки",
+  "Temporary note — moves to trash unless made permanent": "Временная заметка — уедет в корзину, если не сделать постоянной",
+  "Note will move to trash unless made permanent": "Заметка уедет в корзину, если не сделать её постоянной",
+  "Note is now permanent": "Заметка теперь постоянная",
+  "Temporary note lifetime (hours)": "Время жизни временной заметки (часы)",
+  "A temporary note is automatically moved to trash after this many hours unless it is made permanent. The deadline is fixed when the note is created.": "Временная заметка автоматически уезжает в корзину через указанное число часов, если не сделать её постоянной. Дедлайн фиксируется при создании заметки.",
+  "This temporary note moves to trash {{time}} (with its sub-pages) unless made permanent.": "Эта временная заметка уедет в корзину {{time}} (вместе с подстраницами), если не сделать её постоянной.",
   "Move this page to trash?": "Переместить эту страницу в корзину?",
   "Restore page": "Восстановить страницу",
   "Permanently delete": "Удалить навсегда",
@@ -703,13 +715,19 @@
   "Ask the AI agent…": "Спросите AI-агента…",
   "Copy chat": "Копировать чат",
   "Created successfully": "Успешно создано",
-  "Current context size": "Текущий размер контекста",
-  "Tokens generated this turn": "Токенов сгенерировано за ход",
+  "Context size / model limit": "Размер контекста / лимит модели",
+  "Context window (tokens)": "Окно контекста (токены)",
+  "Shown as used / total in the chat header. Leave empty to hide the limit.": "Показывается в шапке чата как использовано / всего. Пусто — лимит скрыт.",
   "Delete this chat?": "Удалить этот чат?",
   "Deleted successfully": "Успешно удалено",
   "Edited by AI agent on behalf of {{name}}": "Отредактировано AI-агентом от имени {{name}}",
   "Failed to delete chat": "Не удалось удалить чат",
   "Failed to rename chat": "Не удалось переименовать чат",
+  "Failed": "Ошибка",
+  "OK · {{n}}": "OK · {{n}}",
+  "Test": "Тест",
+  "No tools available": "Инструменты недоступны",
+  "Available tools": "Доступные инструменты",
   "Minimize": "Свернуть",
   "No chats yet.": "Чатов пока нет.",
   "Send": "Отправить",
@@ -1168,5 +1186,15 @@
   "Protocol": "Протокол",
   "How chat requests are sent and how reasoning is surfaced": "Как отправляются запросы чата и как показывается reasoning",
   "OpenAI-compatible (surfaces reasoning)": "OpenAI-совместимый (показывает reasoning)",
-  "OpenAI (official)": "OpenAI (официальный)"
+  "OpenAI (official)": "OpenAI (официальный)",
+  "Custom address": "Пользовательский адрес",
+  "A short, memorable link you can point at any shared page.": "Короткая запоминающаяся ссылка, которую можно направить на любую опубликованную страницу.",
+  "Use 2-60 lowercase letters, digits and hyphens": "Используйте 2–60 строчных букв, цифр и дефисов",
+  "This address is already in use": "Этот адрес уже занят",
+  "Move custom address?": "Переместить пользовательский адрес?",
+  "Move here": "Переместить сюда",
+  "The address \"{{alias}}\" currently points to \"{{title}}\". Move it to this page?": "Адрес «{{alias}}» сейчас указывает на «{{title}}». Переместить его на эту страницу?",
+  "The address \"{{alias}}\" is already in use. Move it to this page?": "Адрес «{{alias}}» уже используется. Переместить его на эту страницу?",
+  "Failed to set custom address": "Не удалось задать пользовательский адрес",
+  "Failed to remove custom address": "Не удалось удалить пользовательский адрес"
 }

apps/client/src/features/ai-chat/components/ai-chat-window.tsx

@@ -6,7 +6,6 @@ import {
   useRef,
   useState,
 } from "react";
-import { type UIMessage } from "@ai-sdk/react";
 import { Group, Loader, Tooltip } from "@mantine/core";
 import {
   IconArrowsDiagonal,
@@ -40,12 +39,13 @@ import {
 } from "@/features/ai-chat/queries/ai-chat-query.ts";
 import ConversationList from "@/features/ai-chat/components/conversation-list.tsx";
 import ChatThread from "@/features/ai-chat/components/chat-thread.tsx";
-import { buildChatMarkdown } from "@/features/ai-chat/utils/chat-markdown.ts";
+import { exportAiChat } from "@/features/ai-chat/services/ai-chat-service.ts";
 import { useChatSession } from "@/features/ai-chat/hooks/use-chat-session.ts";
 import {
   shouldCollapseOnOutsidePointer,
   isHeaderClick,
 } from "@/features/ai-chat/utils/collapse-helpers.ts";
+import { selectContextBadge } from "@/features/ai-chat/utils/context-badge.ts";
 import { useClipboard } from "@/hooks/use-clipboard";
 import { notifications } from "@mantine/notifications";
 import classes from "@/features/ai-chat/components/ai-chat-window.module.css";
@@ -121,7 +121,7 @@ function clampGeom(g: {
  * ported from the GitmostAgent.jsx design.
  */
 export default function AiChatWindow() {
-  const { t } = useTranslation();
+  const { t, i18n } = useTranslation();
   const clipboard = useClipboard({ timeout: 500 });
   const queryClient = useQueryClient();
   const [windowOpen, setWindowOpen] = useAtom(aiChatWindowOpenAtom);
@@ -162,31 +162,6 @@ export default function AiChatWindow() {
   const { data: messageRows, isLoading: messagesLoading } =
     useAiChatMessagesQuery(activeChatId ?? undefined);
 
-  // Live snapshot of the active thread's useChat state, kept up to date by
-  // ChatThread. Lets the export include the in-progress (not-yet-persisted)
-  // streaming turn. A ref avoids re-rendering this window on every token.
-  const liveThreadRef = useRef<{
-    messages: UIMessage[];
-    isStreaming: boolean;
-    banner: string | null;
-  }>({
-    messages: [],
-    isStreaming: false,
-    banner: null,
-  });
-
-  // Live turn-token total (reasoning + output) for the in-flight turn, pushed up
-  // (THROTTLED to ~8 Hz inside ChatThread) so the header badge ticks mid-stream.
-  // `null` means no turn is in flight -> the badge falls back to the persisted
-  // context size below.
-  const [liveTurnTokens, setLiveTurnTokens] = useState<number | null>(null);
-  // Whether the on-screen thread currently holds at least one message. Reported
-  // reactively by ChatThread (the live snapshot lives in a non-reactive ref). This
-  // lets the "Copy chat" button stay available for a brand-new, not-yet-persisted
-  // chat whose first turn is in flight or was interrupted — that case has no
-  // persisted rows yet, so a persisted-rows-only gate would hide the button (#174).
-  const [hasLiveContent, setHasLiveContent] = useState(false);
-
   // The page the user is currently viewing. AiChatWindow lives in a pathless
   // parent layout route, so useParams() can't see :pageSlug. Match the full
   // pathname against the authenticated page route instead so "the current page"
@@ -213,7 +188,9 @@ export default function AiChatWindow() {
   const {
     threadKey,
     waitingForHistory,
+    startFreshThread,
     onTurnFinished,
+    onServerChatId,
     cancelPendingAdoption,
   } = useChatSession({
     activeChatId,
@@ -234,12 +211,25 @@ export default function AiChatWindow() {
   // just-failed chat after they chose a fresh one.
   const startNewChat = useCallback((): void => {
     cancelPendingAdoption();
+    // Force a fresh, empty thread UNCONDITIONALLY (#161). Pressing "New chat"
+    // while a brand-new chat's first turn is still streaming leaves activeChatId
+    // null (the real id is adopted only at turn end), so setActiveChatId(null)
+    // alone is a no-op and the reconciler never remounts — the chat/stream/history
+    // would persist and only the role badge would drop. This always remounts the
+    // thread into a clean new chat.
+    startFreshThread();
     setActiveChatId(null);
     setHistoryOpen(false);
     setDraft("");
     // Default the picker back to "Universal assistant" for the fresh chat.
     setSelectedRoleId(null);
-  }, [cancelPendingAdoption, setActiveChatId, setDraft, setSelectedRoleId]);
+  }, [
+    cancelPendingAdoption,
+    startFreshThread,
+    setActiveChatId,
+    setDraft,
+    setSelectedRoleId,
+  ]);
 
   const selectChat = useCallback(
     (chatId: string): void => {
@@ -254,20 +244,19 @@ export default function AiChatWindow() {
     [cancelPendingAdoption, setActiveChatId, setDraft, setSelectedRoleId],
   );
 
-  // The active chat object (for its title) and an export gate: only enable the
-  // export button when an existing chat with loaded persisted rows is active.
+  // The active chat object (for its title) and an export gate. The export is now
+  // SERVER-sourced (the DB is the single source of truth — #183): the assistant
+  // row is persisted upfront + per step, so even a brand-new chat whose first
+  // turn is streaming/interrupted has a server row to render. Enable the button
+  // whenever a persisted chat is active (`activeChatId` is set). For a BRAND-NEW
+  // chat that id is adopted EARLY — at the stream's `start` chunk via
+  // onServerChatId (#174) — so the Copy button is available during the first
+  // turn's stream, not only after it terminates.
   const activeChat = useMemo(
     () => chats?.items?.find((c) => c.id === activeChatId) ?? null,
     [chats, activeChatId],
   );
-  // Export is available when there is anything to export: either persisted rows
-  // for the active chat, OR a live on-screen thread with at least one message.
-  // The live arm covers a brand-new chat whose first turn is streaming or was
-  // interrupted before the server persisted any row (#174); the persisted arm is
-  // the steady-state path for an already-saved chat (#160).
-  const canExport =
-    hasLiveContent ||
-    (!!activeChatId && !!messageRows && messageRows.length > 0);
+  const canExport = !!activeChatId;
 
   // The role to display in the header and as the assistant's name. Prefer the
   // persisted role of an existing chat (chat-list JOIN); fall back to the role
@@ -284,53 +273,21 @@ export default function AiChatWindow() {
     return picked ? { name: picked.name, emoji: picked.emoji } : null;
   }, [activeChat, enabledRoles, selectedRoleId]);
 
-  // Build a Markdown export from the already-loaded persisted rows (no network
-  // call) and copy it to the clipboard. The "Copied" notification is the
-  // feedback.
-  const handleCopy = useCallback(() => {
-    // Export gate. There must be SOMETHING to export — either a live on-screen
-    // message or a persisted row. A brand-new chat whose first turn is streaming
-    // or was interrupted has live messages but no persisted rows yet; it still
-    // exports the on-screen thread WYSIWYG (#174). Only a truly empty chat (no
-    // live messages and no rows) is non-exportable (the button is hidden too —
-    // see `canExport`).
-    const live = liveThreadRef.current;
-    const hasRows = !!messageRows && messageRows.length > 0;
-    if (live.messages.length === 0 && !hasRows) return;
-    // WYSIWYG export: the live on-screen messages ARE the document (so a partial
-    // reply from an interrupted turn — which never reached the persisted rows —
-    // is exported just as it appears). The persisted rows enrich each live
-    // message (token usage / error / timestamp) by id and serve as the fallback
-    // when the live mirror is empty. The on-screen banner is appended too. See
-    // issues #160 and #174. `chatId` may be null for a not-yet-saved chat — use a
-    // placeholder so the header line still renders.
-    const markdown = buildChatMarkdown({
-      title: activeChat?.title ?? null,
-      chatId: activeChatId ?? "unsaved",
-      live: live.messages.map((m) => ({
-        id: m.id,
-        role: m.role,
-        parts: (m.parts ?? []) as { type: string; text?: string }[],
-        metadata: m.metadata as
-          | {
-              usage?: {
-                inputTokens?: number;
-                outputTokens?: number;
-                totalTokens?: number;
-                reasoningTokens?: number;
-              };
-              error?: string;
-            }
-          | undefined,
-      })),
-      rows: messageRows,
-      isStreaming: live.isStreaming,
-      banner: live.banner,
-      t,
-    });
-    clipboard.copy(markdown);
-    notifications.show({ message: t("Copied") });
-  }, [activeChatId, messageRows, activeChat, clipboard, t]);
+  // Fetch the server-rendered Markdown export and copy it to the clipboard. The
+  // server is the single source of truth (#183): it renders the transcript from
+  // the persisted rows — including an interrupted turn's in-progress row — so the
+  // export is identical whether the chat is freshly streaming, just switched to,
+  // or reloaded. The `lang` of the active i18n drives the few localized labels.
+  const handleCopy = useCallback(async () => {
+    if (!activeChatId) return;
+    try {
+      const markdown = await exportAiChat(activeChatId, i18n.language);
+      clipboard.copy(markdown);
+      notifications.show({ message: t("Copied") });
+    } catch {
+      notifications.show({ message: t("Failed to export chat"), color: "red" });
+    }
+  }, [activeChatId, clipboard, t, i18n.language]);
 
   // Current context size for the active chat: how much the conversation now
   // occupies in the model's context window — NOT the cumulative tokens spent.
@@ -339,24 +296,19 @@ export default function AiChatWindow() {
   // shipped; older rows fall back to that turn's `usage` total. NOTE: reflects
   // PERSISTED rows (updates on chat open/switch); it does not tick live
   // mid-stream — acceptable for v1.
-  const contextTokens = useMemo(() => {
-    if (!activeChatId || !messageRows) return 0;
-    for (let i = messageRows.length - 1; i >= 0; i--) {
-      const meta = messageRows[i].metadata;
-      if (!meta) continue;
-      if (typeof meta.contextTokens === "number" && meta.contextTokens > 0) {
-        return meta.contextTokens;
-      }
-      const usage = meta.usage;
-      if (usage) {
-        const fallback =
-          usage.totalTokens ??
-          (usage.inputTokens ?? 0) + (usage.outputTokens ?? 0);
-        if (fallback > 0) return fallback;
-      }
-    }
-    return 0;
-  }, [activeChatId, messageRows]);
+  //
+  // The denominator `maxContextTokens` (the model's configured max window) is
+  // derived in the SAME backward scan: it is stamped alongside `contextTokens`
+  // on a completed turn, but the numerator and denominator are taken from the
+  // most recent row carrying EACH value independently — they may land on
+  // different rows (e.g. a fresh error row can carry contextTokens but not
+  // maxContextTokens), so we keep scanning for whichever is still unset. 0 when
+  // no row has it (older rows, or no admin-configured limit) — the badge then
+  // shows just the current size with no denominator.
+  const { contextTokens, maxContextTokens } = useMemo(
+    () => selectContextBadge(activeChatId ? messageRows : undefined),
+    [activeChatId, messageRows],
+  );
 
   // On (re)open, settle the geometry before paint (useLayoutEffect → no
   // first-frame jump): compute an initial top-right placement the first time,
@@ -547,20 +499,17 @@ export default function AiChatWindow() {
         )}
 
         <div style={{ flex: 1, display: "flex", justifyContent: "center" }}>
-          {/* While a turn streams, show the LIVE turn-token count (ticks ~8 Hz);
-              once it finishes, fall back to the persisted context size. Require
-              > 0 so the very first emit (an empty tail message, count 0) does not
-              flash a "0" badge before any token streams in (#151 review). */}
-          {liveTurnTokens !== null && liveTurnTokens > 0 ? (
-            <Tooltip label={t("Tokens generated this turn")} withArrow>
-              <span className={classes.badge}>
-                {formatTokens(liveTurnTokens)}
-              </span>
-            </Tooltip>
-          ) : contextTokens > 0 ? (
-            <Tooltip label={t("Current context size")} withArrow>
+          {/* Always show the persisted "current / max" context. The denominator
+              (the admin-configured model limit) is appended only when known;
+              not clamped when current > max (shown as-is, e.g. "210k / 200k").
+              Hidden entirely until a turn has recorded a context figure. */}
+          {contextTokens > 0 ? (
+            <Tooltip label={t("Context size / model limit")} withArrow>
               <span className={classes.badge}>
                 {formatTokens(contextTokens)}
+                {maxContextTokens > 0
+                  ? ` / ${formatTokens(maxContextTokens)}`
+                  : ""}
               </span>
             </Tooltip>
           ) : null}
@@ -674,6 +623,7 @@ export default function AiChatWindow() {
           ) : (
             <ChatThread
               key={threadKey}
+              threadKey={threadKey}
               chatId={activeChatId}
               initialRows={activeChatId ? messageRows : []}
               openPage={openPage}
@@ -685,9 +635,7 @@ export default function AiChatWindow() {
               onRolePicked={(role) => setSelectedRoleId(role.id)}
               assistantName={currentRole?.name}
               onTurnFinished={onTurnFinished}
-              liveStateRef={liveThreadRef}
-              onLiveTurnTokens={setLiveTurnTokens}
-              onLiveContentChange={setHasLiveContent}
+              onServerChatId={onServerChatId}
             />
           )}
         </div>

apps/client/src/features/ai-chat/components/chat-thread.tsx

@@ -1,11 +1,4 @@
-import {
-  useCallback,
-  useEffect,
-  useMemo,
-  useRef,
-  useState,
-  type MutableRefObject,
-} from "react";
+import { useCallback, useEffect, useMemo, useRef, useState } from "react";
 import { generateId } from "ai";
 import { ActionIcon, Box, Group, Stack, Text } from "@mantine/core";
 import { IconClockHour4, IconX } from "@tabler/icons-react";
@@ -27,7 +20,6 @@ import {
 } from "@/features/ai-chat/utils/role-launch.ts";
 import { describeChatError } from "@/features/ai-chat/utils/error-message.ts";
 import { extractServerChatId } from "@/features/ai-chat/utils/adopt-chat-id.ts";
-import { liveTurnTokens } from "@/features/ai-chat/utils/count-stream-tokens.ts";
 import {
   dequeue,
   enqueueMessage,
@@ -36,6 +28,14 @@ import {
 } from "@/features/ai-chat/utils/queue-helpers.ts";
 import classes from "@/features/ai-chat/components/ai-chat.module.css";
 
+// Throttle how often the streamed `messages` state triggers a re-render. Without
+// it, useChat updates state on EVERY token, so the whole transcript's markdown
+// (marked + DOMPurify) is re-parsed per token — on a long agent run that grows
+// into a quadratic CPU storm that pins the main thread and freezes the UI.
+// ~50ms (20 Hz) keeps streaming visually smooth while decoupling re-render cost
+// from the token rate.
+const STREAM_THROTTLE_MS = 50;
+
 /** The page the user is currently viewing, sent as chat context. */
 export interface OpenPageContext {
   id: string;
@@ -45,6 +45,11 @@ export interface OpenPageContext {
 interface ChatThreadProps {
   /** The open chat id, or null for a brand-new (not-yet-created) chat. */
   chatId: string | null;
+  /** This thread's mount key (the same value the parent uses as React `key`).
+   *  Forwarded to onTurnFinished so the session can tell a turn finishing on the
+   *  CURRENT thread from one ABANDONED by New chat mid-stream — whose onFinish/
+   *  onError still fire after unmount and must not adopt the abandoned chat (#161). */
+  threadKey?: string;
   /** Persisted rows to seed initial messages (existing chats only). */
   initialRows?: IAiChatMessageRow[];
   /** The page currently open in the workspace, or null on a non-page route.
@@ -66,32 +71,16 @@ interface ChatThreadProps {
   /** Called when a turn finishes; the parent refreshes the chat list and, for a
    *  new chat, adopts the freshly created chat id. `serverChatId` is the
    *  authoritative id the server streamed on the assistant message metadata, or
-   *  undefined on a failed turn — see adopt-chat-id.ts for the full #137 design. */
-  onTurnFinished: (serverChatId?: string) => void;
-  /** Parent-owned ref that this thread keeps updated with its live useChat
-   *  snapshot (full message list + streaming flag), so the header's
-   *  "Copy chat" export can include the in-progress, not-yet-persisted
-   *  assistant message. A ref (not state) avoids re-rendering the parent on
-   *  every streamed delta. */
-  liveStateRef?: MutableRefObject<{
-    messages: UIMessage[];
-    isStreaming: boolean;
-    banner: string | null;
-  }>;
-  /** Reports the live turn-token total (reasoning + output) for the in-flight
-   *  turn so the parent can show a header badge that ticks mid-stream. THROTTLED
-   *  here (~8 Hz) so the parent re-renders a handful of times a second, not on
-   *  every streamed delta. Called with `null` when no turn is in flight (the
-   *  parent then reverts the badge to the persisted context size). */
-  onLiveTurnTokens?: (tokens: number | null) => void;
-  /** Reports whether the live thread currently holds at least one message, so the
-   *  parent can gate the "Copy chat" button on the on-screen thread rather than on
-   *  the persisted rows alone. This stays truthy for a brand-new, not-yet-saved
-   *  chat the moment its first user message appears — so an interrupted very first
-   *  turn (no persisted rows yet) is still exportable (#174). Called with `false`
-   *  on unmount so a thread torn down by `key` on chat switch can't leave the
-   *  button enabled for the next, possibly empty, chat. */
-  onLiveContentChange?: (hasContent: boolean) => void;
+   *  undefined on a failed turn — see adopt-chat-id.ts for the full #137 design.
+   *  `finishingThreadKey` (this thread's mount key) lets the session ignore a turn
+   *  finishing on a thread already abandoned by New chat mid-stream (#161). */
+  onTurnFinished: (serverChatId?: string, finishingThreadKey?: string) => void;
+  /** Called EARLY (at the stream's `start` chunk) with the authoritative server
+   *  chat id streamed on the assistant message metadata, so a brand-new chat
+   *  adopts its real id WHILE the first turn is still streaming (#174 — makes the
+   *  Copy/export button available mid-stream). Distinct from onTurnFinished,
+   *  which fires only at the terminal outcome. */
+  onServerChatId?: (serverChatId?: string) => void;
 }
 
 /**
@@ -128,6 +117,7 @@ function rowToUiMessage(row: IAiChatMessageRow): UIMessage {
  */
 export default function ChatThread({
   chatId,
+  threadKey,
   initialRows,
   openPage,
   roleId,
@@ -135,9 +125,7 @@ export default function ChatThread({
   onRolePicked,
   assistantName,
   onTurnFinished,
-  liveStateRef,
-  onLiveTurnTokens,
-  onLiveContentChange,
+  onServerChatId,
 }: ChatThreadProps) {
   const { t } = useTranslation();
 
@@ -266,6 +254,8 @@ export default function ChatThread({
     id: chatStoreId,
     messages: initialMessages,
     transport,
+    // See STREAM_THROTTLE_MS — bounds re-render/markdown-reparse frequency.
+    experimental_throttle: STREAM_THROTTLE_MS,
     // `onFinish` (ai@6 useChat) fires from a `finally` on EVERY terminal outcome
     // — success, user Stop/abort (`isAbort`), network drop (`isDisconnect`), and
     // stream error (`isError`). Keep calling `onTurnFinished()` on all of them
@@ -277,8 +267,10 @@ export default function ChatThread({
     onFinish: ({ message, isAbort, isDisconnect, isError }) => {
       // Forward the authoritative server chatId (streamed on the assistant
       // message metadata) so the parent adopts the REAL created chat id for a new
-      // chat — see adopt-chat-id.ts for the full #137 design.
-      onTurnFinished(extractServerChatId(message));
+      // chat — see adopt-chat-id.ts for the full #137 design. `threadKey` lets the
+      // session ignore this finish if it belongs to a thread abandoned by New chat
+      // mid-stream (#161).
+      onTurnFinished(extractServerChatId(message), threadKey);
       // Show a neutral "stopped" marker for an aborted turn; the red error banner
       // (via `error`) already covers isError, and a clean finish clears any marker.
       if (isError) setStopNotice(null);
@@ -299,13 +291,33 @@ export default function ChatThread({
       // Surface the raw failure in the browser console (devtools) for debugging;
       // the UI separately shows a friendly classified banner (see errorView).
       console.error("AI chat stream error:", streamError);
-      onTurnFinished();
+      onTurnFinished(undefined, threadKey);
     },
   });
 
   // Keep the flush helper pointed at the latest sendMessage instance.
   sendMessageRef.current = sendMessage;
 
+  // EARLY chat-id adoption (#174): the server streams the authoritative chat id
+  // on the assistant message metadata at the `start` chunk (message.metadata.
+  // chatId — see adopt-chat-id.ts / chatStreamMetadata). Forward it to the parent
+  // AS SOON AS it appears (mid-stream), so a brand-new chat adopts its real id
+  // WHILE the first turn is still streaming and activeChatId-gated affordances
+  // (the Copy/export button) light up immediately, instead of only at onFinish.
+  // Keyed by the last-seen id so we forward each distinct id exactly once. The
+  // parent's onServerChatId is idempotent and a no-op once the chat has an id.
+  const lastForwardedChatIdRef = useRef<string | undefined>(undefined);
+  useEffect(() => {
+    if (!onServerChatId) return;
+    const tail = messages[messages.length - 1];
+    if (tail?.role !== "assistant") return;
+    const serverChatId = extractServerChatId(tail);
+    if (!serverChatId || serverChatId === lastForwardedChatIdRef.current)
+      return;
+    lastForwardedChatIdRef.current = serverChatId;
+    onServerChatId(serverChatId);
+  }, [messages, onServerChatId]);
+
   // Live "turn was interrupted" marker for the CURRENT session. The red error
   // banner (driven by `error`) covers the error case; this covers an aborted
   // turn, distinguishing a manual Stop (`isAbort`) from a dropped connection
@@ -328,91 +340,6 @@ export default function ChatThread({
   // the SAME on-screen banner text can be mirrored into the export (issue #160).
   const errorView = error ? describeChatError(error.message ?? "", t) : null;
 
-  // The exact banner the user sees under the message list, flattened to a single
-  // string for the "Copy chat" export so the artifact records the interruption
-  // WYSIWYG. Mirrors the JSX precedence below: error first, else the stop notice.
-  const banner = errorView
-    ? errorView.detail
-      ? `${errorView.title} — ${errorView.detail}`
-      : errorView.title
-    : stopNotice === "manual"
-      ? t("Response stopped.")
-      : stopNotice === "disconnect"
-        ? t("Connection lost — the answer was interrupted.")
-        : null;
-
-  // Mirror the live useChat snapshot into the parent-owned ref so the export
-  // (handled in AiChatWindow) can include the in-progress streaming turn AND the
-  // on-screen banner. The cleanup clears the ref on unmount so a thread torn down
-  // by `key` on chat switch can't leak its (possibly still-streaming) tail into
-  // the next chat's export before the new thread's effect repopulates the ref.
-  useEffect(() => {
-    if (!liveStateRef) return;
-    liveStateRef.current = { messages, isStreaming, banner };
-    return () => {
-      liveStateRef.current = { messages: [], isStreaming: false, banner: null };
-    };
-  }, [liveStateRef, messages, isStreaming, banner]);
-
-  // Reactively report "the live thread has content" to the parent. `liveStateRef`
-  // above is a ref (deliberately non-reactive so streaming deltas don't re-render
-  // the parent), so the export button needs a SEPARATE reactive signal to flip on
-  // for a not-yet-persisted chat. Keyed on the boolean only — identical values are
-  // a no-op setState in the parent, so this does not add per-delta re-renders.
-  const hasLiveContent = messages.length > 0;
-  useEffect(() => {
-    if (!onLiveContentChange) return;
-    onLiveContentChange(hasLiveContent);
-    return () => onLiveContentChange(false);
-  }, [onLiveContentChange, hasLiveContent]);
-
-  // Report the live turn-token total to the parent header badge, THROTTLED to
-  // ~8 Hz so the parent re-renders a few times a second instead of on every
-  // streamed delta. The tail assistant message's reasoning+output (estimate while
-  // streaming, authoritative once a step reports usage) is the live figure. When
-  // the turn ends we emit a final exact value, then `null` so the parent reverts
-  // the badge to the persisted context size.
-  const lastEmitRef = useRef(0);
-  const emitTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-  useEffect(() => {
-    if (!onLiveTurnTokens) return;
-    if (!isStreaming) {
-      // Turn ended (or never started): clear any pending throttle and revert.
-      if (emitTimerRef.current) {
-        clearTimeout(emitTimerRef.current);
-        emitTimerRef.current = null;
-      }
-      lastEmitRef.current = 0;
-      onLiveTurnTokens(null);
-      return;
-    }
-    const tail = messages[messages.length - 1];
-    const live = tail?.role === "assistant" ? liveTurnTokens(tail) : null;
-    const total = live ? live.reasoning + live.output : 0;
-    const now = Date.now();
-    const MIN_INTERVAL = 120; // ms (~8 Hz)
-    const elapsed = now - lastEmitRef.current;
-    if (elapsed >= MIN_INTERVAL) {
-      lastEmitRef.current = now;
-      onLiveTurnTokens(total);
-    } else if (!emitTimerRef.current) {
-      // Schedule a trailing emit so the FINAL value of a burst is not dropped.
-      emitTimerRef.current = setTimeout(() => {
-        emitTimerRef.current = null;
-        lastEmitRef.current = Date.now();
-        onLiveTurnTokens(total);
-      }, MIN_INTERVAL - elapsed);
-    }
-  }, [messages, isStreaming, onLiveTurnTokens]);
-
-  // Clear any pending throttle timer on unmount (chat switch via `key`) so a
-  // trailing emit can't fire into a torn-down thread's parent.
-  useEffect(() => {
-    return () => {
-      if (emitTimerRef.current) clearTimeout(emitTimerRef.current);
-    };
-  }, []);
-
   // A role was picked with autoStart=false: the role is bound but NOTHING was
   // sent, so chatId stays null and the empty state would keep showing the cards.
   // This flag hides the cards and reveals the composer (with the role indicated)

apps/client/src/features/ai-chat/components/message-item-memo.test.tsx

@@ -0,0 +1,81 @@
+import { describe, expect, it, vi } from "vitest";
+import { render } from "@testing-library/react";
+import { MantineProvider } from "@mantine/core";
+import type { UIMessage } from "@ai-sdk/react";
+
+// Stub react-i18next (the component reads `useTranslation`). Mirrors the stub in
+// reasoning-block.test.tsx.
+vi.mock("react-i18next", () => ({
+  useTranslation: () => ({ t: (key: string) => key }),
+}));
+
+// Spy on `renderChatMarkdown` so we can count parse calls per text. We keep every
+// OTHER named export of markdown.ts intact via `importActual`, and override only
+// `renderChatMarkdown` with a `vi.fn()` that returns simple HTML so the component
+// still renders. This is the seam that proves the MarkdownPart memo works: a
+// finalized text part must NOT be re-parsed on a later streamed delta.
+// `vi.hoisted` so the spy exists when the hoisted `vi.mock` factory runs.
+const { renderChatMarkdownSpy } = vi.hoisted(() => ({
+  renderChatMarkdownSpy: vi.fn((text: string) => `<p>${text}</p>`),
+}));
+vi.mock("@/features/ai-chat/utils/markdown.ts", async () => {
+  const actual = await vi.importActual<
+    typeof import("@/features/ai-chat/utils/markdown.ts")
+  >("@/features/ai-chat/utils/markdown.ts");
+  return { ...actual, renderChatMarkdown: renderChatMarkdownSpy };
+});
+
+import MessageItem from "./message-item";
+
+// matchMedia (read by MantineProvider) is stubbed globally in vitest.setup.ts.
+
+const msg = (parts: UIMessage["parts"]): UIMessage =>
+  ({ id: "m1", role: "assistant", parts }) as UIMessage;
+
+const renderRow = (message: UIMessage) =>
+  render(
+    <MantineProvider>
+      <MessageItem message={message} />
+    </MantineProvider>,
+  );
+
+/** Count how many spy calls parsed exactly `text` (filtering by the first arg). */
+const callsFor = (text: string) =>
+  renderChatMarkdownSpy.mock.calls.filter((c) => c[0] === text).length;
+
+describe("MessageItem markdown memoization", () => {
+  it("does not re-parse finalized text parts when only a tail part grows", () => {
+    renderChatMarkdownSpy.mockClear();
+
+    // Two finalized text parts.
+    const first = msg([
+      { type: "text", text: "alpha" },
+      { type: "text", text: "beta" },
+    ]);
+    const { rerender } = renderRow(first);
+
+    // Both finalized parts parsed exactly once on the initial render.
+    expect(callsFor("alpha")).toBe(1);
+    expect(callsFor("beta")).toBe(1);
+
+    // A streamed delta: a NEW message object where only a third tail part grows;
+    // the first two parts' text is byte-identical.
+    const next = msg([
+      { type: "text", text: "alpha" },
+      { type: "text", text: "beta" },
+      { type: "text", text: "gamm" },
+    ]);
+    rerender(
+      <MantineProvider>
+        <MessageItem message={next} />
+      </MantineProvider>,
+    );
+
+    // The finalized parts hit the MarkdownPart memo: still parsed at most once
+    // each across BOTH renders (the resilient invariant). The only new parse is
+    // for the changed/added tail part.
+    expect(callsFor("alpha")).toBe(1);
+    expect(callsFor("beta")).toBe(1);
+    expect(callsFor("gamm")).toBe(1);
+  });
+});

apps/client/src/features/ai-chat/components/message-item.test.ts

@@ -0,0 +1,73 @@
+import { describe, expect, it, vi } from "vitest";
+import type { UIMessage } from "@ai-sdk/react";
+
+// Stub react-i18next: importing the component module pulls in `useTranslation`,
+// and we only exercise the pure `arePropsEqual` comparator (no rendering), so a
+// minimal `t` that echoes the key is enough. Mirrors the stub in
+// reasoning-block.test.tsx.
+vi.mock("react-i18next", () => ({
+  useTranslation: () => ({ t: (key: string) => key }),
+}));
+
+import { arePropsEqual } from "./message-item";
+
+/**
+ * Tests for `arePropsEqual`, the `React.memo` comparator for MessageItem. It must
+ * return false on any visible prop/content change (so the row re-renders) and
+ * true when nothing visible changed (so a finalized row is skipped). A FIXED
+ * message id is used so a content-identical clone yields an equal signature.
+ */
+const msg = (parts: UIMessage["parts"]): UIMessage =>
+  ({ id: "m1", role: "assistant", parts }) as UIMessage;
+
+const props = (
+  message: UIMessage,
+  over: Record<string, unknown> = {},
+) => ({
+  message,
+  showCitations: true,
+  neutralizeInternalLinks: false,
+  assistantName: "AI",
+  ...over,
+});
+
+describe("arePropsEqual", () => {
+  it("returns false when showCitations differs", () => {
+    const m = msg([{ type: "text", text: "answer" }]);
+    expect(
+      arePropsEqual(props(m), props(m, { showCitations: false })),
+    ).toBe(false);
+  });
+
+  it("returns false when neutralizeInternalLinks differs", () => {
+    const m = msg([{ type: "text", text: "answer" }]);
+    expect(
+      arePropsEqual(props(m), props(m, { neutralizeInternalLinks: true })),
+    ).toBe(false);
+  });
+
+  it("returns false when assistantName differs", () => {
+    const m = msg([{ type: "text", text: "answer" }]);
+    expect(
+      arePropsEqual(props(m), props(m, { assistantName: "Other" })),
+    ).toBe(false);
+  });
+
+  it("returns true on the identity fast path (same message object, equal props)", () => {
+    const m = msg([{ type: "text", text: "answer" }]);
+    expect(arePropsEqual(props(m), props(m))).toBe(true);
+  });
+
+  it("returns true for the same content in a different message object", () => {
+    const a = msg([{ type: "text", text: "answer" }]);
+    const b = msg([{ type: "text", text: "answer" }]);
+    expect(a).not.toBe(b);
+    expect(arePropsEqual(props(a), props(b))).toBe(true);
+  });
+
+  it("returns false when content changed in a different message object", () => {
+    const a = msg([{ type: "text", text: "answer" }]);
+    const b = msg([{ type: "text", text: "answer grown" }]);
+    expect(arePropsEqual(props(a), props(b))).toBe(false);
+  });
+});

apps/client/src/features/ai-chat/components/message-item.tsx

@@ -1,3 +1,4 @@
+import { memo } from "react";
 import { Box, Text } from "@mantine/core";
 import { useTranslation } from "react-i18next";
 import type { UIMessage } from "@ai-sdk/react";
@@ -10,6 +11,7 @@ import { assistantMessageHasVisibleContent } from "@/features/ai-chat/utils/mess
 import { renderChatMarkdown } from "@/features/ai-chat/utils/markdown.ts";
 import { resolveAssistantName } from "@/features/ai-chat/utils/assistant-name.ts";
 import { reasoningTokensForPart } from "@/features/ai-chat/utils/reasoning-tokens.ts";
+import { messageSignature } from "@/features/ai-chat/utils/message-signature.ts";
 import { describeChatError } from "@/features/ai-chat/utils/error-message.ts";
 import classes from "@/features/ai-chat/components/ai-chat.module.css";
 
@@ -34,6 +36,39 @@ interface MessageItemProps {
   assistantName?: string;
 }
 
+/**
+ * One assistant text part rendered as sanitized markdown. Memoized on its inputs
+ * so a finalized text part is NOT re-parsed on every streamed delta: during a
+ * turn only the actively-growing tail part changes its `text`, so every earlier
+ * part hits the memo and skips the expensive marked + DOMPurify pass. Props are
+ * primitives, so React.memo's default shallow compare is exactly right (the
+ * `text` string is compared by value).
+ */
+const MarkdownPart = memo(function MarkdownPart({
+  text,
+  neutralizeInternalLinks,
+}: {
+  text: string;
+  neutralizeInternalLinks: boolean;
+}) {
+  const html = renderChatMarkdown(text, { neutralizeInternalLinks });
+  if (html) {
+    return (
+      <div
+        className={classes.markdown}
+        // Sanitized by renderChatMarkdown (DOMPurify) before insertion.
+        dangerouslySetInnerHTML={{ __html: html }}
+      />
+    );
+  }
+  // Fallback when markdown could not render synchronously: raw text.
+  return (
+    <Text className={classes.markdown} style={{ whiteSpace: "pre-wrap" }}>
+      {text}
+    </Text>
+  );
+});
+
 /**
  * Render a single UIMessage by iterating its `parts`:
  *  - `text` parts -> sanitized markdown.
@@ -41,12 +76,13 @@ interface MessageItemProps {
  * Other part kinds (reasoning, sources, files, step-start) are ignored for v1.
  * User messages render their text as a right-aligned plain bubble.
  *
- * This component is intentionally NOT memoized: `useChat` replaces the streaming
- * assistant message with a freshly cloned object on every streamed delta, so the
- * `message` prop identity (and its `parts`) changes each tick. Re-rendering the
- * text parts on each delta is what makes the answer stream in progressively.
+ * This component is memoized (see `arePropsEqual` at the bottom) on a cheap
+ * per-message content signature: the streaming TAIL message's signature changes
+ * on each delta so it still re-renders and streams in, while finalized rows are
+ * skipped. Each text part's markdown is itself memoized via `MarkdownPart`, so a
+ * long turn no longer re-parses the whole transcript on every token.
  */
-export default function MessageItem({
+function MessageItem({
   message,
   showCitations = true,
   neutralizeInternalLinks = false,
@@ -109,24 +145,12 @@ export default function MessageItem({
           // starts with an empty text part before the first token arrives); the
           // typing indicator covers that gap until real content streams in.
           if (!part.text.trim()) return null;
-          const html = renderChatMarkdown(part.text, {
-            neutralizeInternalLinks,
-          });
-          if (html) {
-            return (
-              <div
-                key={index}
-                className={classes.markdown}
-                // Sanitized by renderChatMarkdown (DOMPurify) before insertion.
-                dangerouslySetInnerHTML={{ __html: html }}
-              />
-            );
-          }
-          // Fallback when markdown could not render synchronously: raw text.
           return (
-            <Text key={index} className={classes.markdown} style={{ whiteSpace: "pre-wrap" }}>
-              {part.text}
-            </Text>
+            <MarkdownPart
+              key={index}
+              text={part.text}
+              neutralizeInternalLinks={neutralizeInternalLinks}
+            />
           );
         }
 
@@ -177,3 +201,26 @@ export default function MessageItem({
     </Box>
   );
 }
+
+/** Skip re-rendering a message whose visible content is unchanged. The streaming
+ *  TAIL message gets a fresh object whose signature changes each delta, so it
+ *  still re-renders and streams in; every FINALIZED message is skipped, turning a
+ *  per-token whole-transcript re-render into a tail-only one. */
+export function arePropsEqual(
+  prev: MessageItemProps,
+  next: MessageItemProps,
+): boolean {
+  if (
+    prev.showCitations !== next.showCitations ||
+    prev.neutralizeInternalLinks !== next.neutralizeInternalLinks ||
+    prev.assistantName !== next.assistantName
+  ) {
+    return false;
+  }
+  // Fast path: identical message object (finalized rows keep their identity
+  // across deltas) — skip without building signatures.
+  if (prev.message === next.message) return true;
+  return messageSignature(prev.message) === messageSignature(next.message);
+}
+
+export default memo(MessageItem, arePropsEqual);

apps/client/src/features/ai-chat/components/reasoning-block.tsx

@@ -1,4 +1,4 @@
-import { useState } from "react";
+import { memo, useMemo, useState } from "react";
 import { Box, Collapse, Group, Text, UnstyledButton } from "@mantine/core";
 import { IconChevronDown } from "@tabler/icons-react";
 import { useTranslation } from "react-i18next";
@@ -27,19 +27,23 @@ interface ReasoningBlockProps {
  * Providers that don't stream reasoning TEXT still render this block from the
  * authoritative count alone (header only, empty body) so the cost is visible.
  */
-export default function ReasoningBlock({ text, tokens }: ReasoningBlockProps) {
+function ReasoningBlock({ text, tokens }: ReasoningBlockProps) {
   const { t } = useTranslation();
   const [open, setOpen] = useState(false);
 
   // Authoritative count wins; otherwise estimate live from the streamed text.
   const count = tokens && tokens > 0 ? tokens : estimateTokens(text);
   const trimmed = text.trim();
-  // Collapse the blank-line gaps the model emits between every list item /
-  // paragraph so the reasoning renders compactly (tight lists, joined
-  // paragraphs) — see collapseBlankLines. ONLY here, not in the normal answer.
-  const html = trimmed
-    ? renderChatMarkdown(collapseBlankLines(trimmed), {})
-    : "";
+  // Memoize the markdown render so toggling `open` (or a parent re-render caused
+  // by an unrelated streamed delta) does not re-parse the reasoning text; it
+  // recomputes only when the reasoning text itself changes (while it streams in).
+  // collapseBlankLines collapses the blank-line gaps the model emits between every
+  // list item / paragraph so the reasoning renders compactly (tight lists, joined
+  // paragraphs) — ONLY here, not in the normal answer.
+  const html = useMemo(
+    () => (trimmed ? renderChatMarkdown(collapseBlankLines(trimmed), {}) : ""),
+    [trimmed],
+  );
 
   return (
     <Box className={classes.reasoningBlock} mb={6}>
@@ -87,3 +91,8 @@ export default function ReasoningBlock({ text, tokens }: ReasoningBlockProps) {
     </Box>
   );
 }
+
+// Memoized: re-renders only when `text`/`tokens` change (primitive props, default
+// shallow compare), so a parent re-render during streaming of OTHER content does
+// not re-run the markdown parse for an already-finalized reasoning block.
+export default memo(ReasoningBlock);

apps/client/src/features/ai-chat/hooks/use-chat-session.test.tsx

@@ -1,5 +1,5 @@
 import { describe, it, expect, vi, beforeEach } from "vitest";
-import { renderHook } from "@testing-library/react";
+import { renderHook, act } from "@testing-library/react";
 import { useChatSession } from "./use-chat-session";
 import type { UseChatSessionOptions } from "./use-chat-session";
 
@@ -64,7 +64,10 @@ describe("useChatSession", () => {
     result.current.onTurnFinished(undefined);
     expect(setActiveChatId).not.toHaveBeenCalled();
     // The refetch lands with the new row => adopt it.
-    rerender({ activeChatId: null, chats: { items: [{ id: "x" }, { id: "new" }] } });
+    rerender({
+      activeChatId: null,
+      chats: { items: [{ id: "x" }, { id: "new" }] },
+    });
     expect(setActiveChatId).toHaveBeenCalledWith("new");
   });
 
@@ -88,7 +91,10 @@ describe("useChatSession", () => {
     });
     result.current.onTurnFinished(undefined);
     // a was deleted, new was added — same length, but membership changed.
-    rerender({ activeChatId: null, chats: { items: [{ id: "b" }, { id: "new" }] } });
+    rerender({
+      activeChatId: null,
+      chats: { items: [{ id: "b" }, { id: "new" }] },
+    });
     expect(setActiveChatId).toHaveBeenCalledWith("new");
   });
 
@@ -171,6 +177,40 @@ describe("useChatSession", () => {
     expect(setActiveChatId).not.toHaveBeenCalledWith("late");
   });
 
+  it("#174 early adopt: onServerChatId adopts the streamed id mid-stream (Copy button available during the first turn)", () => {
+    // Brand-new chat: no id yet. The server streams the real chat id "A" on the
+    // `start` chunk WHILE the first turn is still streaming (before onTurnFinished
+    // fires at the terminal outcome). The hook must adopt it immediately so the
+    // window's activeChatId-gated Copy/export button lights up during the stream.
+    const { result, setActiveChatId } = setup({
+      activeChatId: null,
+      chats: { items: [] },
+    });
+    result.current.onServerChatId("A");
+    expect(setActiveChatId).toHaveBeenCalledWith("A");
+  });
+
+  it("#174 early adopt is in-place: threadKey stays stable (live stream not torn down)", () => {
+    const chats = { items: [] };
+    const { result, rerender } = setup({ activeChatId: null, chats });
+    const keyBefore = result.current.threadKey;
+    result.current.onServerChatId("A");
+    // Parent reflects the adopted id back in; the SAME mount key is kept so the
+    // in-flight useChat store (the streaming turn) is preserved.
+    rerender({ activeChatId: "A", chats });
+    expect(result.current.threadKey).toBe(keyBefore);
+  });
+
+  it("#174 early adopt: no-op for an existing chat and for a missing id", () => {
+    const { result, setActiveChatId } = setup({
+      activeChatId: "chat-1",
+      chats: { items: [{ id: "chat-1" }] },
+    });
+    result.current.onServerChatId("chat-1"); // already has an id
+    result.current.onServerChatId(undefined); // no streamed id
+    expect(setActiveChatId).not.toHaveBeenCalled();
+  });
+
   it("in-place adopt keeps threadKey stable; an external switch remounts", () => {
     const chats = { items: [{ id: "B" }] };
     const { result, rerender } = setup({ activeChatId: null, chats });
@@ -187,6 +227,50 @@ describe("useChatSession", () => {
     expect(result.current.threadKey).toBe("C");
   });
 
+  it("#161: New chat during a streaming first turn forces a fresh thread (remount), not just a no-op", () => {
+    // Brand-new chat whose first turn is still streaming: the id is adopted only
+    // at turn end, so activeChatId AND thread.chatId are both null. Pressing "New
+    // chat" must still remount to a clean thread even though the atom is unchanged
+    // — the render-phase reconciler (null === null) would otherwise do nothing,
+    // leaving the old chat/stream/history in place (the bug: only the role badge
+    // dropped).
+    const { result } = setup({ activeChatId: null, chats: { items: [] } });
+    const keyBefore = result.current.threadKey;
+    act(() => result.current.startFreshThread());
+    expect(result.current.threadKey).not.toBe(keyBefore);
+  });
+
+  it("#161: an abandoned thread's late onTurnFinished does NOT adopt its chat (thread-aware guard)", () => {
+    // New chat mid-stream remounts to a fresh thread, but @ai-sdk/react does not
+    // abort the abandoned stream on unmount: its onFinish still fires later with
+    // the real server id, tagged with the OLD (abandoned) mount key. That must not
+    // adopt — it would yank the user back into the chat they just left.
+    const { result, setActiveChatId, onInvalidateChatList } = setup({
+      activeChatId: null,
+      chats: { items: [] },
+    });
+    const abandonedKey = result.current.threadKey;
+    act(() => result.current.startFreshThread());
+    expect(result.current.threadKey).not.toBe(abandonedKey);
+    // The abandoned turn finishes in the background, streaming its real id "A".
+    result.current.onTurnFinished("A", abandonedKey);
+    expect(setActiveChatId).not.toHaveBeenCalledWith("A");
+    // It still refreshes the chat list so the left-behind chat shows in history.
+    expect(onInvalidateChatList).toHaveBeenCalled();
+  });
+
+  it("#161: a turn finishing on the CURRENT thread still adopts (guard is key-scoped, not blanket)", () => {
+    // The happy path must keep working: onTurnFinished tagged with the mounted
+    // thread's own key adopts in place as before.
+    const { result, setActiveChatId } = setup({
+      activeChatId: null,
+      chats: { items: [] },
+    });
+    const currentKey = result.current.threadKey;
+    result.current.onTurnFinished("A", currentKey);
+    expect(setActiveChatId).toHaveBeenCalledWith("A");
+  });
+
   it("waitingForHistory gates the loader only while opening an unloaded existing chat", () => {
     // Open an existing chat whose history is still loading => loader on.
     const { result, rerender } = setup({

apps/client/src/features/ai-chat/hooks/use-chat-session.ts

@@ -31,9 +31,26 @@ export interface UseChatSessionResult {
   threadKey: string;
   /** Show the history loader instead of the live thread. */
   waitingForHistory: boolean;
+  /** Force a brand-new, empty thread (new mount key, no chat id) UNCONDITIONALLY,
+   *  even when `activeChatId` is unchanged. The window calls this from
+   *  startNewChat so "New chat" pressed WHILE a brand-new chat's first turn is
+   *  still streaming (activeChatId still null, nothing to diverge) actually
+   *  resets the chat instead of only dropping the role badge (#161). */
+  startFreshThread: () => void;
   /** Call when a turn finishes; `serverChatId` is the authoritative streamed id
-   *  (undefined on a failed turn). Handles new-chat id adoption + invalidations. */
-  onTurnFinished: (serverChatId?: string) => void;
+   *  (undefined on a failed turn). `finishingThreadKey` is the mount key of the
+   *  thread that produced the turn (omit => "current thread", back-compatible):
+   *  a turn ABANDONED by New chat mid-stream still fires this after its thread
+   *  unmounted, so adoption is gated to the still-mounted thread (#161). Handles
+   *  new-chat id adoption + invalidations. */
+  onTurnFinished: (serverChatId?: string, finishingThreadKey?: string) => void;
+  /** Call EARLY (at the stream's `start` chunk) with the authoritative streamed
+   *  chat id so a brand-new chat adopts its real id WHILE its first turn is still
+   *  streaming — making `activeChatId`-gated affordances (e.g. the Copy/export
+   *  button, #174) available immediately. In-place adoption only (same mount key,
+   *  no list/messages invalidation — that is left to onTurnFinished at the end).
+   *  Idempotent and a no-op once the chat already has an id. */
+  onServerChatId: (serverChatId?: string) => void;
   /** Disarm any pending error-path new-chat fallback. The window calls this from
    *  startNewChat/selectChat so a late refetch can't yank the user back into a
    *  just-failed chat after they explicitly moved on. */
@@ -85,15 +102,21 @@ export function useChatSession(
   // `newThread`/`switchThread` to (re)mount, `adoptThread` for in-place adoption.
   // Initial: a non-null activeChatId switches to it; a null one gets a fresh
   // session key with no chat id yet.
-  const [thread, dispatch] = useReducer(
-    threadSessionReducer,
-    undefined,
-    () =>
-      activeChatId === null
-        ? newThread(`new-${generateId()}`)
-        : switchThread(activeChatId),
+  const [thread, dispatch] = useReducer(threadSessionReducer, undefined, () =>
+    activeChatId === null
+      ? newThread(`new-${generateId()}`)
+      : switchThread(activeChatId),
   );
 
+  // Live mirror of the mounted thread's mount key, read by onTurnFinished to tell
+  // the CURRENT thread from one ABANDONED by New chat mid-stream. @ai-sdk/react
+  // does not abort a stream on unmount and proxies callbacks through a ref, so an
+  // abandoned turn's onFinish/onError still fires AFTER its ChatThread unmounted;
+  // matching its key against this ref keeps that late finish from adopting the
+  // abandoned chat and yanking the user out of the fresh chat they opened (#161).
+  const threadKeyRef = useRef(thread.key);
+  threadKeyRef.current = thread.key;
+
   // Error-path fallback for new-chat id adoption. When a brand-new chat's first
   // turn errors BEFORE the server's `start` chunk, no authoritative chatId ever
   // reaches the client, so the primary metadata adoption cannot run. We then ARM
@@ -111,7 +134,23 @@ export function useChatSession(
   // yet) we adopt the server's AUTHORITATIVE streamed id (never the newest in the
   // list, which races a second tab — #137; see adopt-chat-id.ts).
   const onTurnFinished = useCallback(
-    (serverChatId?: string) => {
+    (serverChatId?: string, finishingThreadKey?: string) => {
+      // Thread-aware guard (#161). A turn ABANDONED by "New chat" mid-stream still
+      // fires onFinish/onError after its ChatThread unmounted (@ai-sdk/react does
+      // not abort on unmount and proxies callbacks through a ref). If that late
+      // finish ran the adoption path it would set activeChatId to the abandoned
+      // chat's real id and yank the user out of the fresh chat they just opened.
+      // So adopt / arm the fallback ONLY for the still-mounted thread; an
+      // abandoned one merely refreshes the chat list (so the left-behind chat
+      // surfaces in history) and does nothing else. A missing key (undefined)
+      // means "current thread" — keeps old call sites/tests working.
+      if (
+        finishingThreadKey !== undefined &&
+        finishingThreadKey !== threadKeyRef.current
+      ) {
+        onInvalidateChatList();
+        return;
+      }
       // Read the live id from the ref, not the closure: on a failed turn this can
       // run twice in one turn (onFinish + onError) before any re-render, and the
       // primary branch below updates the ref so the second call sees the adopted id.
@@ -150,6 +189,31 @@ export function useChatSession(
     [chats, setActiveChatId, onInvalidateChatList, onInvalidateChatMessages],
   );
 
+  // EARLY adoption (#174): adopt the authoritative streamed chat id the moment
+  // the server emits it on the `start` chunk, so a brand-new chat gets its real
+  // `activeChatId` WHILE its first turn streams — not only at terminal
+  // onTurnFinished. This makes the activeChatId-gated Copy/export button
+  // available during the first turn. Pure in-place adoption (same mount key, like
+  // the primary path) with NO invalidation: the list/messages refresh stays on
+  // onTurnFinished at the end of the turn. Reads the live id from the ref so a
+  // repeat call after adoption is a no-op (resolveAdoptedChatId only fires for a
+  // still-new chat).
+  const onServerChatId = useCallback(
+    (serverChatId?: string) => {
+      const adopted = resolveAdoptedChatId(
+        activeChatIdRef.current,
+        serverChatId,
+      );
+      if (!adopted) return;
+      activeChatIdRef.current = adopted;
+      setActiveChatId(adopted);
+      dispatch({ type: "adopt", chatId: adopted });
+      // Early adoption beat the error-path fallback to it — disarm.
+      pendingNewChatRef.current = null;
+    },
+    [setActiveChatId],
+  );
+
   // FALLBACK resolver. Armed only by onTurnFinished when a brand-new chat's first
   // turn errored before the `start` chunk (no authoritative id streamed). Once
   // the per-user list refetch lands with the just-created row, adopt the SINGLE
@@ -229,10 +293,30 @@ export function useChatSession(
     pendingNewChatRef.current = null;
   }, []);
 
+  // Force a fresh, empty thread regardless of `activeChatId` (#161). The render-
+  // phase reconciler only remounts when activeChatId diverges from thread.chatId,
+  // so "New chat" pressed while a brand-new chat's first turn is still streaming
+  // (activeChatId AND thread.chatId both null — the real id is adopted only at the
+  // end of the turn) is a no-op for it and the abandoned thread/stream/history
+  // would persist. Dispatching reconcile with a fresh key and chatId:null here
+  // always produces a new mount key, so React remounts ChatThread (a clean useChat
+  // store) and the post-dispatch state (activeChatId null === thread.chatId null)
+  // keeps the reconciler from interfering. Also disarms any pending fallback.
+  const startFreshThread = useCallback(() => {
+    pendingNewChatRef.current = null;
+    dispatch({
+      type: "reconcile",
+      chatId: null,
+      newKey: `new-${generateId()}`,
+    });
+  }, []);
+
   return {
     threadKey: thread.key,
     waitingForHistory,
+    startFreshThread,
     onTurnFinished,
+    onServerChatId,
     cancelPendingAdoption,
   };
 }

apps/client/src/features/ai-chat/services/ai-chat-service.ts

@@ -50,6 +50,24 @@ export async function deleteAiChat(chatId: string): Promise<void> {
   await api.post("/ai-chat/delete", { chatId });
 }
 
+/**
+ * Export a chat to Markdown (#183). The server renders the transcript from the
+ * persisted rows (the DB is the single source of truth — including an
+ * interrupted turn's in-progress row, persisted upfront + per step), so the
+ * client just copies the returned string. `lang` localizes the few fixed
+ * role/tool labels; defaults to English server-side when omitted.
+ */
+export async function exportAiChat(
+  chatId: string,
+  lang?: string,
+): Promise<string> {
+  const req = await api.post<{ markdown: string }>("/ai-chat/export", {
+    chatId,
+    lang,
+  });
+  return req.data.markdown;
+}
+
 /**
  * Agent roles API (`/ai-chat/roles`). `list` is available to any workspace
  * member (for the chat-creation picker); create/update/delete are admin-only
@@ -76,6 +94,8 @@ export async function updateAiRole(data: IAiRoleUpdate): Promise<IAiRole> {
 
 /** Soft-delete a role (admin). */
 export async function deleteAiRole(id: string): Promise<{ success: true }> {
-  const req = await api.post<{ success: true }>("/ai-chat/roles/delete", { id });
+  const req = await api.post<{ success: true }>("/ai-chat/roles/delete", {
+    id,
+  });
   return req.data;
 }

apps/client/src/features/ai-chat/types/ai-chat.types.ts

@@ -116,6 +116,9 @@ export interface IAiChatMessageRow {
     // turn. Distinct from `usage` (legacy cumulative totalUsage). Shown in the
     // floating window's header badge.
     contextTokens?: number;
+    // The model's max context window (denominator for the header badge); set
+    // alongside contextTokens on a completed turn; absent on older rows.
+    maxContextTokens?: number;
     // Set on an assistant row whose turn ended in a provider/stream error; the
     // raw provider error text (e.g. "402: ...") for inline display in the thread.
     error?: string;

apps/client/src/features/ai-chat/utils/chat-markdown.test.ts

@@ -1,747 +0,0 @@
-import { describe, it, expect } from "vitest";
-import { buildChatMarkdown } from "@/features/ai-chat/utils/chat-markdown.ts";
-import type { IAiChatMessageRow } from "@/features/ai-chat/types/ai-chat.types.ts";
-
-/**
- * Tests for the client-only Markdown export builder. The output embeds a live
- * `new Date().toISOString()` export timestamp; we never assert that value, only
- * the deterministic structure (headings, numbering, fenced blocks, totals).
- *
- * A pass-through translator keeps role/tool labels predictable so the
- * structural assertions are stable without an i18n runtime.
- */
-const t = (key: string, values?: Record<string, unknown>): string => {
-  if (values && typeof values.name === "string") {
-    return key.replace("{{name}}", values.name);
-  }
-  return key;
-};
-
-function row(partial: Partial<IAiChatMessageRow>): IAiChatMessageRow {
-  return {
-    id: partial.id ?? "id",
-    role: partial.role ?? "user",
-    content: partial.content ?? null,
-    metadata: partial.metadata ?? null,
-    createdAt: partial.createdAt ?? "2026-06-21T00:00:00.000Z",
-  };
-}
-
-describe("buildChatMarkdown — structure", () => {
-  it("emits the title heading, chat id and message count", () => {
-    const md = buildChatMarkdown({
-      title: "My chat",
-      chatId: "chat-123",
-      rows: [],
-      t,
-    });
-    expect(md).toContain("# My chat");
-    expect(md).toContain("- Chat ID: `chat-123`");
-    expect(md).toContain("- Messages: 0");
-    expect(md).toContain("- Exported:"); // timestamp present, value not asserted
-  });
-
-  it("falls back to the translated 'Untitled chat' for empty/blank titles", () => {
-    expect(
-      buildChatMarkdown({ title: null, chatId: "c", rows: [], t }),
-    ).toContain("# Untitled chat");
-    expect(
-      buildChatMarkdown({ title: "   ", chatId: "c", rows: [], t }),
-    ).toContain("# Untitled chat");
-  });
-
-  it("numbers rows sequentially with role headings", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({ role: "user", content: "hi" }),
-        row({ role: "assistant", content: "hello" }),
-        row({ role: "user", content: "again" }),
-      ],
-      t,
-    });
-    expect(md).toContain("## 1. You");
-    expect(md).toContain("## 2. AI agent");
-    expect(md).toContain("## 3. You");
-    // Heading numbering is strictly index+1, not e.g. role-relative.
-    expect(md).not.toContain("## 0.");
-  });
-
-  it("renders the per-row text content from `content` when no metadata.parts", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ role: "user", content: "plain body" })],
-      t,
-    });
-    expect(md).toContain("plain body");
-  });
-});
-
-describe("buildChatMarkdown — text parts", () => {
-  it("skips empty / whitespace-only text parts", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "ignored-content",
-          metadata: {
-            parts: [
-              { type: "text", text: "   " },
-              { type: "text", text: "" },
-              { type: "text", text: "kept line" },
-              // eslint-disable-next-line @typescript-eslint/no-explicit-any
-            ] as any,
-          },
-        }),
-      ],
-      t,
-    });
-    expect(md).toContain("kept line");
-    // Whitespace-only part contributed no block of its own.
-    expect(md).not.toContain("   \n\n");
-    // When metadata.parts exists, the plain `content` fallback is NOT used.
-    expect(md).not.toContain("ignored-content");
-  });
-});
-
-describe("buildChatMarkdown — tool parts", () => {
-  it("renders a tool label, name, state and fenced Input/Output blocks", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "",
-          metadata: {
-            parts: [
-              {
-                type: "tool-getPage",
-                state: "output-available",
-                input: { pageId: "p1" },
-                output: { id: "p1", title: "Home" },
-                // eslint-disable-next-line @typescript-eslint/no-explicit-any
-              } as any,
-            ],
-          },
-        }),
-      ],
-      t,
-    });
-    // Known tool name maps to its label key; raw name in backticks; done state.
-    expect(md).toContain("**Tool: Read page** (`getPage`) — done");
-    expect(md).toContain("Input:");
-    expect(md).toContain("Output:");
-    // Fenced JSON blocks contain the stringified payloads.
-    expect(md).toContain('"pageId": "p1"');
-    expect(md).toContain('"title": "Home"');
-    expect(md).toContain("```json");
-  });
-
-  it("renders the generic label for an unknown tool and surfaces errorText", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "",
-          metadata: {
-            parts: [
-              {
-                type: "tool-mysteryTool",
-                state: "output-error",
-                input: { a: 1 },
-                errorText: "boom",
-                // eslint-disable-next-line @typescript-eslint/no-explicit-any
-              } as any,
-            ],
-          },
-        }),
-      ],
-      t,
-    });
-    expect(md).toContain(
-      "**Tool: Ran tool mysteryTool** (`mysteryTool`) — error",
-    );
-    expect(md).toContain("**Error:** boom");
-  });
-
-  it("does not throw on a circular tool input (falls back to String)", () => {
-    // eslint-disable-next-line @typescript-eslint/no-explicit-any
-    const circular: any = {};
-    circular.self = circular;
-    expect(() =>
-      buildChatMarkdown({
-        title: "t",
-        chatId: "c",
-        rows: [
-          row({
-            role: "assistant",
-            content: "",
-            metadata: {
-              parts: [
-                {
-                  type: "tool-getPage",
-                  state: "input-available",
-                  input: circular,
-                  // eslint-disable-next-line @typescript-eslint/no-explicit-any
-                } as any,
-              ],
-            },
-          }),
-        ],
-        t,
-      }),
-    ).not.toThrow();
-  });
-});
-
-describe("buildChatMarkdown — fence anti-breakout", () => {
-  it("lengthens the delimiter so embedded ``` cannot break out of the block", () => {
-    // Tool input whose stringified string form contains a literal ``` run.
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "",
-          metadata: {
-            parts: [
-              {
-                type: "tool-getPage",
-                state: "output-available",
-                // A bare string passes through stringify() verbatim.
-                input: "before ``` after",
-                output: "x",
-                // eslint-disable-next-line @typescript-eslint/no-explicit-any
-              } as any,
-            ],
-          },
-        }),
-      ],
-      t,
-    });
-    // The fence around the 3-backtick content must use at least 4 backticks so
-    // the embedded ``` run cannot terminate the block.
-    expect(md).toContain("````json\nbefore ``` after\n````");
-    // Robust anti-breakout check: the opening fence delimiter is strictly
-    // longer than the longest backtick run inside the wrapped content. (A naive
-    // `not.toContain("```json...")` is a false negative — a 4-backtick fence
-    // textually contains the 3-backtick substring.)
-    const open = md.match(/(`{3,})json\nbefore/);
-    expect(open).not.toBeNull();
-    expect(open![1].length).toBeGreaterThan(3); // > the 3-backtick run in content
-  });
-
-  it("uses a 5-backtick fence when the content has a 4-backtick run", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "",
-          metadata: {
-            parts: [
-              {
-                type: "tool-getPage",
-                state: "output-available",
-                input: "a ```` b",
-                // eslint-disable-next-line @typescript-eslint/no-explicit-any
-              } as any,
-            ],
-          },
-        }),
-      ],
-      t,
-    });
-    expect(md).toContain("`````json\na ```` b\n`````");
-  });
-});
-
-describe("buildChatMarkdown — token totals", () => {
-  it("prints the total-tokens line only when the summed usage is > 0", () => {
-    const withTokens = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: { usage: { inputTokens: 10, outputTokens: 5 } },
-        }),
-      ],
-      t,
-    });
-    expect(withTokens).toContain("- Total tokens: 15");
-    // Per-row usage footer too.
-    expect(withTokens).toContain("_Tokens — in: 10, out: 5, total: 15_");
-  });
-
-  it("omits the total-tokens line when the sum is 0 / usage absent", () => {
-    const noTokens = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({ role: "user", content: "hi" }),
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: { usage: { inputTokens: 0, outputTokens: 0 } },
-        }),
-      ],
-      t,
-    });
-    expect(noTokens).not.toContain("- Total tokens:");
-  });
-
-  it("uses totalTokens when present rather than summing in/out", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: {
-            usage: { inputTokens: 3, outputTokens: 4, totalTokens: 99 },
-          },
-        }),
-      ],
-      t,
-    });
-    expect(md).toContain("- Total tokens: 99");
-  });
-
-  it("appends the reasoning figure to the row footer when reasoningTokens > 0", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: {
-            usage: { inputTokens: 10, outputTokens: 8, reasoningTokens: 3 },
-          },
-        }),
-      ],
-      t,
-    });
-    expect(md).toContain("_Tokens — in: 10, out: 8, reasoning: 3, total: 18_");
-  });
-
-  it("omits the reasoning figure when reasoningTokens is 0 / absent", () => {
-    const zero = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: {
-            usage: { inputTokens: 10, outputTokens: 5, reasoningTokens: 0 },
-          },
-        }),
-      ],
-      t,
-    });
-    expect(zero).toContain("_Tokens — in: 10, out: 5, total: 15_");
-    expect(zero).not.toContain("reasoning:");
-
-    const absent = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          role: "assistant",
-          content: "x",
-          metadata: { usage: { inputTokens: 10, outputTokens: 5 } },
-        }),
-      ],
-      t,
-    });
-    expect(absent).not.toContain("reasoning:");
-  });
-});
-
-// A minimal on-screen (live) message, matching the subset buildChatMarkdown reads.
-function live(partial: {
-  id?: string;
-  role?: string;
-  parts?: { type: string; text?: string }[];
-  metadata?: { usage?: Record<string, number>; error?: string };
-}) {
-  return {
-    id: partial.id ?? "live-id",
-    role: partial.role ?? "assistant",
-    parts: partial.parts ?? [],
-    metadata: partial.metadata,
-  };
-}
-
-describe("buildChatMarkdown — live (WYSIWYG) source", () => {
-  it("uses the live messages as the document (what's on screen), numbered from 1", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      // Persisted rows hold only the user turn; the assistant reply is live-only.
-      rows: [row({ id: "u1", role: "user", content: "persisted user" })],
-      live: [
-        live({
-          id: "u1",
-          role: "user",
-          parts: [{ type: "text", text: "on-screen user" }],
-        }),
-        live({
-          id: "a1",
-          role: "assistant",
-          parts: [{ type: "text", text: "on-screen reply" }],
-        }),
-      ],
-      isStreaming: false,
-      t,
-    });
-    expect(md).toContain("## 1. You");
-    expect(md).toContain("## 2. AI agent");
-    expect(md).toContain("on-screen user");
-    expect(md).toContain("on-screen reply");
-    // Message count reflects the LIVE document, not rows + live.
-    expect(md).toContain("- Messages: 2");
-  });
-
-  it("captures a partial reply from an interrupted (non-streaming) turn — no 'generating' note", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ id: "u1", role: "user", content: "q" })],
-      live: [
-        live({ id: "u1", role: "user", parts: [{ type: "text", text: "q" }] }),
-        live({
-          id: "a-live",
-          role: "assistant",
-          parts: [{ type: "text", text: "partial plan before the drop" }],
-        }),
-      ],
-      isStreaming: false, // the stream dropped — not streaming anymore
-      banner: "Connection lost — the answer was interrupted.",
-      t,
-    });
-    // The partial assistant answer that was on screen IS in the export.
-    expect(md).toContain("partial plan before the drop");
-    // It is NOT flagged still-generating (the turn is over, just interrupted).
-    expect(md).not.toContain("still being generated");
-    // The on-screen banner is recorded at the end.
-    expect(md).toContain("Connection lost — the answer was interrupted.");
-  });
-
-  it("flags ONLY the tail assistant as still generating, and only while streaming", () => {
-    const streaming = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [],
-      live: [
-        live({
-          id: "a",
-          role: "assistant",
-          parts: [{ type: "text", text: "done earlier" }],
-        }),
-        live({
-          id: "u",
-          role: "user",
-          parts: [{ type: "text", text: "next q" }],
-        }),
-        live({
-          id: "b",
-          role: "assistant",
-          parts: [{ type: "text", text: "streaming now" }],
-        }),
-      ],
-      isStreaming: true,
-      t,
-    });
-    // Exactly one "still being generated" note (the tail assistant).
-    expect(streaming.match(/still being generated/g)?.length).toBe(1);
-
-    const idle = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [],
-      live: [
-        live({
-          id: "b",
-          role: "assistant",
-          parts: [{ type: "text", text: "final" }],
-        }),
-      ],
-      isStreaming: false,
-      t,
-    });
-    expect(idle).not.toContain("still being generated");
-  });
-
-  it("does NOT flag a completed assistant as generating when the streaming tail is a user message", () => {
-    // The `status === "submitted"` window: the user just sent, isStreaming is
-    // already true, but the new assistant turn has no message yet so the tail is
-    // the USER message. The previous assistant answer is complete on screen and
-    // must not be marked still-generating (WYSIWYG; regression for #160 review).
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [],
-      live: [
-        live({
-          id: "a",
-          role: "assistant",
-          parts: [{ type: "text", text: "completed answer" }],
-        }),
-        live({
-          id: "u",
-          role: "user",
-          parts: [{ type: "text", text: "the new question" }],
-        }),
-      ],
-      isStreaming: true,
-      t,
-    });
-    expect(md).toContain("completed answer");
-    expect(md).not.toContain("still being generated");
-  });
-
-  it("emits the heading + note for a streaming tail assistant with empty parts", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ id: "u1", role: "user", content: "q" })],
-      live: [
-        live({ id: "u1", role: "user", parts: [{ type: "text", text: "q" }] }),
-        live({ id: "a-live", role: "assistant", parts: [] }),
-      ],
-      isStreaming: true,
-      t,
-    });
-    expect(md).toContain("## 2. AI agent");
-    expect(md).toContain("still being generated");
-  });
-});
-
-describe("buildChatMarkdown — live enrichment from persisted rows", () => {
-  it("pulls usage / error / timestamp from the persisted row matched by id", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          id: "a1",
-          role: "assistant",
-          content: "x",
-          createdAt: "2026-06-22T10:00:00.000Z",
-          metadata: {
-            usage: { inputTokens: 10, outputTokens: 5 },
-            error: "rate limited",
-          },
-        }),
-      ],
-      live: [
-        // Same id as the persisted row, but no usage/error/timestamp on the live msg.
-        live({
-          id: "a1",
-          role: "assistant",
-          parts: [{ type: "text", text: "reply" }],
-        }),
-      ],
-      isStreaming: false,
-      t,
-    });
-    expect(md).toContain("reply");
-    // Token footer + total come from the enriched row.
-    expect(md).toContain("_Tokens — in: 10, out: 5, total: 15_");
-    expect(md).toContain("- Total tokens: 15");
-    expect(md).toContain("**⚠️ Error:** rate limited");
-    // The persisted timestamp is carried into the export.
-    expect(md).toContain("<!-- 2026-06-22T10:00:00.000Z -->");
-  });
-
-  it("prefers authoritative usage already on the live message over the row's", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({
-          id: "a1",
-          role: "assistant",
-          content: "x",
-          metadata: {
-            usage: { inputTokens: 1, outputTokens: 1, totalTokens: 2 },
-          },
-        }),
-      ],
-      live: [
-        live({
-          id: "a1",
-          role: "assistant",
-          parts: [{ type: "text", text: "reply" }],
-          metadata: {
-            usage: { inputTokens: 100, outputTokens: 50, totalTokens: 150 },
-          },
-        }),
-      ],
-      isStreaming: false,
-      t,
-    });
-    // The live (authoritative, freshest) usage wins, not the stale row usage.
-    expect(md).toContain("- Total tokens: 150");
-    expect(md).not.toContain("- Total tokens: 2");
-  });
-
-  it("a current-turn live message with no matching row renders without a footer", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ id: "u1", role: "user", content: "q" })],
-      live: [
-        live({ id: "u1", role: "user", parts: [{ type: "text", text: "q" }] }),
-        live({
-          id: "a-live",
-          role: "assistant",
-          parts: [{ type: "text", text: "fresh reply" }],
-        }),
-      ],
-      isStreaming: false,
-      t,
-    });
-    expect(md).toContain("fresh reply");
-    // No persisted row for the live assistant -> no token footer, no timestamp.
-    expect(md).not.toContain("_Tokens —");
-    expect(md).not.toContain("<!-- undefined -->");
-  });
-});
-
-describe("buildChatMarkdown — fallback + banner", () => {
-  it("falls back to the persisted rows when there are no live messages", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [
-        row({ role: "user", content: "from rows" }),
-        row({
-          role: "assistant",
-          content: "answer",
-          metadata: { usage: { inputTokens: 4, outputTokens: 6 } },
-        }),
-      ],
-      live: [], // empty live mirror -> fallback path
-      isStreaming: false,
-      t,
-    });
-    expect(md).toContain("## 1. You");
-    expect(md).toContain("## 2. AI agent");
-    expect(md).toContain("from rows");
-    expect(md).toContain("- Messages: 2");
-    expect(md).toContain("- Total tokens: 10");
-  });
-
-  it("appends the on-screen banner once, after the messages", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ role: "user", content: "q" })],
-      live: [
-        live({ id: "u", role: "user", parts: [{ type: "text", text: "q" }] }),
-      ],
-      isStreaming: false,
-      banner: "Rate limit reached — try again shortly.",
-      t,
-    });
-    expect(md).toContain("_⚠️ Rate limit reached — try again shortly._");
-    // Banner comes after the (only) message block.
-    expect(md.indexOf("Rate limit reached")).toBeGreaterThan(
-      md.indexOf("## 1."),
-    );
-  });
-
-  it("omits the banner block when there is no banner", () => {
-    const md = buildChatMarkdown({
-      title: "t",
-      chatId: "c",
-      rows: [row({ role: "user", content: "q" })],
-      live: [
-        live({ id: "u", role: "user", parts: [{ type: "text", text: "q" }] }),
-      ],
-      isStreaming: false,
-      banner: null,
-      t,
-    });
-    expect(md).not.toContain("_⚠️");
-  });
-});
-
-// #174: a brand-new, not-yet-persisted chat whose first turn is streaming (or was
-// interrupted) has live messages but NO persisted rows yet, and its chat id is not
-// known (the caller passes a placeholder). The export must still capture the
-// on-screen thread WYSIWYG from the live messages alone.
-describe("buildChatMarkdown — first-turn export with no persisted base (#174)", () => {
-  it("builds the document from live messages alone when rows are empty", () => {
-    const md = buildChatMarkdown({
-      title: null,
-      chatId: "unsaved",
-      rows: [],
-      live: [
-        live({
-          id: "u1",
-          role: "user",
-          parts: [{ type: "text", text: "hello" }],
-        }),
-        live({
-          id: "a1",
-          role: "assistant",
-          parts: [{ type: "text", text: "partial reply" }],
-        }),
-      ],
-      isStreaming: true,
-      t,
-    });
-    // Both on-screen messages are serialized, numbered from 1.
-    expect(md).toContain("## 1. You");
-    expect(md).toContain("hello");
-    expect(md).toContain("## 2. AI agent");
-    expect(md).toContain("partial reply");
-    // The streaming tail assistant is flagged as in-progress.
-    expect(md).toContain("still being generated");
-    // The placeholder chat id and the live message count are recorded.
-    expect(md).toContain("- Chat ID: `unsaved`");
-    expect(md).toContain("- Messages: 2");
-    // No persisted timestamp exists for a current-turn live message.
-    expect(md).not.toContain("<!--");
-  });
-
-  it("captures an interrupted first turn (no rows, not streaming) without a generating note", () => {
-    const md = buildChatMarkdown({
-      title: null,
-      chatId: "unsaved",
-      rows: [],
-      live: [
-        live({ id: "u1", role: "user", parts: [{ type: "text", text: "q" }] }),
-        live({
-          id: "a1",
-          role: "assistant",
-          parts: [{ type: "text", text: "half an answer" }],
-        }),
-      ],
-      isStreaming: false,
-      banner: "Connection dropped — the response was cut off.",
-      t,
-    });
-    expect(md).toContain("half an answer");
-    // An interrupted (non-streaming) partial is exported as-is, no generating note.
-    expect(md).not.toContain("still being generated");
-    // The on-screen banner records the interruption.
-    expect(md).toContain("_⚠️ Connection dropped — the response was cut off._");
-  });
-});

apps/client/src/features/ai-chat/utils/chat-markdown.ts

@@ -1,308 +0,0 @@
-/**
- * Client-only Markdown builder for an AI agent chat. Serializes the already
- * persisted message rows (loaded via `useAiChatMessagesQuery`) into a single
- * Markdown string suitable for copying to the clipboard. NO network call is
- * made and NO server/DB code is touched — this reuses the rich "request
- * internals" (tool calls with input/output, per-message token usage,
- * finish/error info) that the chat already holds client-side.
- *
- * Only role labels and tool action labels are localized via the passed-in `t`
- * translator; the structural document words (Input/Output/Error/Tokens/...) are
- * plain English constants because the output is a technical artifact.
- */
-
-import type { IAiChatMessageRow } from "@/features/ai-chat/types/ai-chat.types.ts";
-import {
-  ToolUiPart,
-  getToolName,
-  toolRunState,
-  toolLabelKey,
-} from "@/features/ai-chat/utils/tool-parts.tsx";
-
-// Minimal translator signature compatible with react-i18next's `t`.
-type Translate = (key: string, values?: Record<string, unknown>) => string;
-
-interface BuildChatMarkdownArgs {
-  title: string | null;
-  chatId: string;
-  /** The live, on-screen messages — the WYSIWYG source of the export. When
-   *  present and non-empty these DRIVE the document (so it mirrors exactly what
-   *  the user sees, including a partial reply from an interrupted turn). Each is
-   *  matched to a persisted row by `id` to enrich it with token usage / error /
-   *  timestamp. When absent or empty the builder falls back to `rows`. */
-  live?: LiveMessage[];
-  /** Persisted message rows. Enrichment source (matched to `live` by id) AND the
-   *  fallback document source when `live` is empty. */
-  rows: IAiChatMessageRow[];
-  /** Whether the live thread is still streaming. Only then is the tail assistant
-   *  message flagged "still generating"; an interrupted (non-streaming) partial
-   *  reply is exported as-is and the `banner` explains the interruption. */
-  isStreaming?: boolean;
-  /** The on-screen banner text (error / dropped connection / manual stop),
-   *  appended at the end of the export so the artifact records the interruption
-   *  the user saw. */
-  banner?: string | null;
-  t: Translate;
-}
-
-/** A single AI SDK UIMessage part (text part or other). */
-interface TextLikePart {
-  type: string;
-  text?: string;
-}
-
-/** Authoritative per-turn usage the server attaches to a message / row. */
-interface UsageLike {
-  inputTokens?: number;
-  outputTokens?: number;
-  totalTokens?: number;
-  reasoningTokens?: number;
-}
-
-/** A live, on-screen message (subset of the AI SDK UIMessage we consume). */
-interface LiveMessage {
-  id: string;
-  role: "user" | "assistant" | string;
-  parts: TextLikePart[];
-  metadata?: { usage?: UsageLike; error?: string };
-}
-
-/** One message normalized for rendering, regardless of live/persisted origin. */
-interface ExportItem {
-  role: string;
-  parts: TextLikePart[];
-  usage?: UsageLike;
-  error?: string;
-  /** ISO timestamp from the persisted row, when one is known. */
-  createdAt?: string;
-  /** True only for the tail assistant message while the thread is streaming. */
-  generating: boolean;
-}
-
-/**
- * Stringify an arbitrary tool input/output value for a fenced block. Strings
- * pass through as-is; everything else is pretty-printed JSON, falling back to
- * `String(value)` if serialization throws (e.g. a circular structure).
- */
-function stringify(value: unknown): string {
-  if (typeof value === "string") return value;
-  try {
-    return JSON.stringify(value, null, 2);
-  } catch {
-    return String(value);
-  }
-}
-
-/**
- * Wrap `code` in a fenced code block whose backtick delimiter is LONGER than
- * the longest backtick run inside the content, so embedded backticks (or even
- * a literal ``` fence) never break out of the block. Minimum 3 backticks.
- */
-function fence(code: string, lang = ""): string {
-  const runs: string[] = code.match(/`+/g) ?? [];
-  const longest = runs.reduce((m, s) => Math.max(m, s.length), 0);
-  const delim = "`".repeat(Math.max(3, longest + 1));
-  return `${delim}${lang}\n${code}\n${delim}`;
-}
-
-/** Per-row token count, mirroring the header sum in ai-chat-window.tsx. */
-function rowTokens(usage: {
-  inputTokens?: number;
-  outputTokens?: number;
-  totalTokens?: number;
-  reasoningTokens?: number;
-}): number {
-  return (
-    usage.totalTokens ?? (usage.inputTokens ?? 0) + (usage.outputTokens ?? 0)
-  );
-}
-
-/** Render one message's UIMessage parts into an array of Markdown blocks
- *  (text blocks + tool blocks). Mirrors MessageItem's part handling. */
-function renderMessageParts(parts: TextLikePart[], t: Translate): string[] {
-  const out: string[] = [];
-
-  for (const part of parts) {
-    if (part.type === "text") {
-      const text = (part.text ?? "").trim();
-      // Skip empty/whitespace-only text parts (matches MessageItem).
-      if (text.length > 0) out.push(text);
-      continue;
-    }
-
-    const isToolPart =
-      part.type.startsWith("tool-") || part.type === "dynamic-tool";
-    if (!isToolPart) continue;
-
-    const tp = part as unknown as ToolUiPart;
-    const name = getToolName(tp);
-    const { key, values } = toolLabelKey(name);
-    const label = t(key, values);
-    const state = toolRunState(tp.state);
-
-    const toolLines: string[] = [
-      `**Tool: ${label}** (\`${name}\`) — ${state}`,
-    ];
-    if (tp.input !== undefined) {
-      toolLines.push("Input:");
-      toolLines.push(fence(stringify(tp.input), "json"));
-    }
-    if (tp.output !== undefined) {
-      toolLines.push("Output:");
-      toolLines.push(fence(stringify(tp.output), "json"));
-    }
-    if (tp.errorText) {
-      toolLines.push(`**Error:** ${tp.errorText}`);
-    }
-    out.push(toolLines.join("\n\n"));
-  }
-
-  return out;
-}
-
-/** Resolve a persisted row's parts: prefer the rich persisted parts, else a
- *  single text part built from the plain-text content (mirrors `rowToUiMessage`). */
-function rowParts(row: IAiChatMessageRow): TextLikePart[] {
-  return Array.isArray(row.metadata?.parts) && row.metadata.parts.length > 0
-    ? (row.metadata.parts as TextLikePart[])
-    : [{ type: "text", text: row.content ?? "" }];
-}
-
-/**
- * Normalize the export to one ordered list of {@link ExportItem}, WYSIWYG-first:
- *
- * - When `live` messages are present, THEY are the document (what the user sees,
- *   incl. an interrupted turn's partial reply). Each is matched to a persisted
- *   row by `id` to pull token usage / error / timestamp — a live message of the
- *   CURRENT turn has no matching row yet, so it simply renders without a footer.
- *   Authoritative `usage`/`error` already on the live message metadata win over
- *   the row (the server attaches usage to the streamed message at a step
- *   boundary before the row is refetched). Only the tail assistant message is
- *   flagged `generating`, and only while `isStreaming`.
- * - When `live` is empty (e.g. the export runs before the live mirror is
- *   populated), fall back to the persisted `rows` so the format never regresses.
- */
-function resolveItems(
-  live: LiveMessage[] | undefined,
-  rows: IAiChatMessageRow[],
-  isStreaming: boolean,
-): ExportItem[] {
-  if (live && live.length > 0) {
-    const rowsById = new Map(rows.map((r) => [r.id, r]));
-    // The "still generating" note may apply ONLY to an assistant message that is
-    // the actual TAIL of the list — that is where the on-screen typing indicator
-    // sits. While `status === "submitted"` (isStreaming true) right after the
-    // user hit send, the tail is the USER message and the new assistant turn has
-    // no message yet; the previous assistant answer is shown complete on screen,
-    // so it must NOT be flagged (the indicator renders as a separate bottom
-    // block, not on that answer).
-    const lastIndex = live.length - 1;
-    const tailIsStreamingAssistant =
-      isStreaming && live[lastIndex]?.role === "assistant";
-    return live.map((m, i) => {
-      const row = rowsById.get(m.id);
-      return {
-        role: m.role,
-        parts: m.parts ?? [],
-        // Authoritative usage/error already on the live message (the server
-        // attaches usage to the streamed message at a step boundary) wins over
-        // the persisted row; a current-turn live message has no matching row yet
-        // and simply renders without a token footer (the accepted WYSIWYG
-        // tradeoff — an interrupted turn loses only its token footer, not text).
-        usage: m.metadata?.usage ?? row?.metadata?.usage,
-        error: m.metadata?.error ?? row?.metadata?.error ?? undefined,
-        createdAt: row?.createdAt,
-        generating: tailIsStreamingAssistant && i === lastIndex,
-      };
-    });
-  }
-
-  return rows.map((row) => ({
-    role: row.role,
-    parts: rowParts(row),
-    usage: row.metadata?.usage,
-    error: row.metadata?.error ?? undefined,
-    createdAt: row.createdAt,
-    generating: false,
-  }));
-}
-
-/**
- * Serialize a chat to a Markdown string. Pure (apart from `new Date()` for the
- * export timestamp), so it is straightforward to unit-test.
- */
-export function buildChatMarkdown(args: BuildChatMarkdownArgs): string {
-  const { title, chatId, live, rows, isStreaming, banner, t } = args;
-  const blocks: string[] = [];
-
-  const items = resolveItems(live, rows, isStreaming === true);
-
-  const heading = (title ?? "").trim() || t("Untitled chat");
-  blocks.push(`# ${heading}`);
-
-  // Metadata bullet list. Total tokens is only shown when there is a sum.
-  const totalTokens = items.reduce(
-    (sum, item) => (item.usage ? sum + rowTokens(item.usage) : sum),
-    0,
-  );
-  const meta = [
-    `- Chat ID: \`${chatId}\``,
-    `- Exported: ${new Date().toISOString()}`,
-    `- Messages: ${items.length}`,
-  ];
-  if (totalTokens > 0) meta.push(`- Total tokens: ${totalTokens}`);
-  blocks.push(meta.join("\n"));
-
-  items.forEach((item, index) => {
-    blocks.push("---");
-
-    const roleLabel = item.role === "assistant" ? t("AI agent") : t("You");
-    blocks.push(`## ${index + 1}. ${roleLabel}`);
-
-    // Created-at kept in source as an HTML comment (out of the rendered prose).
-    // A live message of the current turn has no persisted row yet — omit it.
-    if (item.createdAt) blocks.push(`<!-- ${item.createdAt} -->`);
-
-    blocks.push(...renderMessageParts(item.parts, t));
-
-    // A generating assistant may have empty/no parts yet — the heading (above)
-    // and this note still record the in-progress turn.
-    if (item.generating) {
-      blocks.push(
-        "_⏳ This message is still being generated — the export captured a partial, in-progress response._",
-      );
-    }
-
-    // A persisted per-message error (the raw provider text) may coexist with the
-    // trailing `banner` (the classified on-screen alert) when the failed turn's
-    // row has already been refetched by export time. They describe the same
-    // failure at different fidelity; showing both is an accepted, minor redundancy.
-    if (item.error) {
-      blocks.push(`**⚠️ Error:** ${item.error}`);
-    }
-
-    const usage = item.usage;
-    if (usage) {
-      const total = usage.totalTokens ?? rowTokens(usage);
-      // Reasoning (thinking) tokens are shown only when the provider reported a
-      // positive count; old rows / non-reasoning providers omit it.
-      const reasoning =
-        usage.reasoningTokens && usage.reasoningTokens > 0
-          ? `, reasoning: ${usage.reasoningTokens}`
-          : "";
-      blocks.push(
-        `_Tokens — in: ${usage.inputTokens ?? "?"}, out: ${usage.outputTokens ?? "?"}${reasoning}, total: ${total}_`,
-      );
-    }
-  });
-
-  // Record the on-screen banner (error / dropped connection / manual stop) so
-  // the export reflects exactly what the user saw, including an interruption.
-  if (banner && banner.trim().length > 0) {
-    blocks.push("---");
-    blocks.push(`_⚠️ ${banner.trim()}_`);
-  }
-
-  // Blank line between blocks so the Markdown renders cleanly.
-  return blocks.join("\n\n");
-}

apps/client/src/features/ai-chat/utils/context-badge.test.ts

@@ -0,0 +1,90 @@
+import { describe, expect, it } from "vitest";
+import type { IAiChatMessageRow } from "@/features/ai-chat/types/ai-chat.types.ts";
+import { selectContextBadge } from "@/features/ai-chat/utils/context-badge.ts";
+
+/**
+ * Pure-helper tests for the header context badge selection. Covers the two
+ * non-obvious rules: numerator and denominator are each taken from the most
+ * recent row carrying THAT value (they may live on different rows), and a fresh
+ * row with a zero/absent value must NOT shadow an older positive one.
+ */
+const row = (metadata: IAiChatMessageRow["metadata"]): IAiChatMessageRow => ({
+  id: Math.random().toString(),
+  role: "assistant",
+  content: null,
+  metadata,
+  createdAt: "2026-01-01T00:00:00.000Z",
+});
+
+describe("selectContextBadge", () => {
+  it("returns zeros for empty / nullish input", () => {
+    expect(selectContextBadge(undefined)).toEqual({
+      contextTokens: 0,
+      maxContextTokens: 0,
+    });
+    expect(selectContextBadge(null)).toEqual({
+      contextTokens: 0,
+      maxContextTokens: 0,
+    });
+    expect(selectContextBadge([])).toEqual({
+      contextTokens: 0,
+      maxContextTokens: 0,
+    });
+  });
+
+  it("reads both figures from the most recent row that carries them", () => {
+    expect(
+      selectContextBadge([
+        row({ contextTokens: 100, maxContextTokens: 200000 }),
+        row({ contextTokens: 1500, maxContextTokens: 200000 }),
+      ]),
+    ).toEqual({ contextTokens: 1500, maxContextTokens: 200000 });
+  });
+
+  it("falls back to legacy usage total for older rows without contextTokens", () => {
+    expect(
+      selectContextBadge([
+        row({ usage: { inputTokens: 30, outputTokens: 70 } }),
+      ]),
+    ).toEqual({ contextTokens: 100, maxContextTokens: 0 });
+
+    expect(
+      selectContextBadge([row({ usage: { totalTokens: 250 } })]),
+    ).toEqual({ contextTokens: 250, maxContextTokens: 0 });
+  });
+
+  it("takes numerator and denominator from different rows", () => {
+    // Freshest row (an error turn) carries contextTokens but no max; the older
+    // completed turn carries the max. Each is picked from its own latest row.
+    expect(
+      selectContextBadge([
+        row({ contextTokens: 800, maxContextTokens: 200000 }),
+        row({ contextTokens: 1200, error: "402: nope" }),
+      ]),
+    ).toEqual({ contextTokens: 1200, maxContextTokens: 200000 });
+  });
+
+  it("does not let a fresh zero/absent max shadow an older positive max", () => {
+    expect(
+      selectContextBadge([
+        row({ contextTokens: 100, maxContextTokens: 200000 }),
+        row({ contextTokens: 1200, maxContextTokens: 0 }),
+      ]),
+    ).toEqual({ contextTokens: 1200, maxContextTokens: 200000 });
+  });
+
+  it("skips rows with null metadata", () => {
+    expect(
+      selectContextBadge([
+        row({ contextTokens: 500, maxContextTokens: 200000 }),
+        row(null),
+      ]),
+    ).toEqual({ contextTokens: 500, maxContextTokens: 200000 });
+  });
+
+  it("reports current > max as-is (no clamp)", () => {
+    expect(
+      selectContextBadge([row({ contextTokens: 250000, maxContextTokens: 200000 })]),
+    ).toEqual({ contextTokens: 250000, maxContextTokens: 200000 });
+  });
+});

apps/client/src/features/ai-chat/utils/context-badge.ts

@@ -0,0 +1,49 @@
+import type { IAiChatMessageRow } from "@/features/ai-chat/types/ai-chat.types.ts";
+
+/**
+ * Derive the header context badge figures from the persisted message rows.
+ *
+ * - `contextTokens` (numerator): how much the conversation now occupies in the
+ *   model's context window. Read from the most recent row carrying a context
+ *   figure — `contextTokens` (final-step input+output) on rows recorded after
+ *   this shipped, else that turn's legacy `usage` total for older rows.
+ * - `maxContextTokens` (denominator): the model's configured max window, stamped
+ *   alongside `contextTokens` on a completed turn.
+ *
+ * Each value is taken from the most recent row carrying THAT value
+ * independently — they may land on different rows (e.g. a fresh error row can
+ * carry `contextTokens` but not `maxContextTokens`), so the scan continues for
+ * whichever is still unset. `0` means "no row has it" (older rows, or no
+ * admin-configured limit); the badge then omits the value.
+ */
+export function selectContextBadge(
+  messageRows: readonly IAiChatMessageRow[] | undefined | null,
+): { contextTokens: number; maxContextTokens: number } {
+  let contextTokens = 0;
+  let maxContextTokens = 0;
+  if (!messageRows) return { contextTokens, maxContextTokens };
+  for (let i = messageRows.length - 1; i >= 0; i--) {
+    const meta = messageRows[i].metadata;
+    if (!meta) continue;
+    if (contextTokens === 0) {
+      if (typeof meta.contextTokens === "number" && meta.contextTokens > 0) {
+        contextTokens = meta.contextTokens;
+      } else if (meta.usage) {
+        const usage = meta.usage;
+        const fallback =
+          usage.totalTokens ??
+          (usage.inputTokens ?? 0) + (usage.outputTokens ?? 0);
+        if (fallback > 0) contextTokens = fallback;
+      }
+    }
+    if (
+      maxContextTokens === 0 &&
+      typeof meta.maxContextTokens === "number" &&
+      meta.maxContextTokens > 0
+    ) {
+      maxContextTokens = meta.maxContextTokens;
+    }
+    if (contextTokens !== 0 && maxContextTokens !== 0) break;
+  }
+  return { contextTokens, maxContextTokens };
+}

apps/client/src/features/ai-chat/utils/count-stream-tokens.test.ts

@@ -1,17 +1,5 @@
 import { describe, expect, it } from "vitest";
-import type { UIMessage } from "@ai-sdk/react";
-import {
-  estimateTokens,
-  liveTurnTokens,
-} from "@/features/ai-chat/utils/count-stream-tokens.ts";
-
-const msg = (parts: unknown[], metadata?: unknown): UIMessage =>
-  ({
-    id: Math.random().toString(),
-    role: "assistant",
-    parts,
-    metadata,
-  }) as UIMessage;
+import { estimateTokens } from "@/features/ai-chat/utils/count-stream-tokens.ts";
 
 describe("estimateTokens", () => {
   it("returns 0 for the empty string", () => {
@@ -25,147 +13,3 @@ describe("estimateTokens", () => {
     expect(estimateTokens("12345678")).toBe(2);
   });
 });
-
-describe("liveTurnTokens — estimate path", () => {
-  it("is all zeros for an undefined message", () => {
-    expect(liveTurnTokens(undefined)).toEqual({
-      reasoning: 0,
-      output: 0,
-      authoritative: false,
-    });
-  });
-
-  it("is all zeros for a parts-less message", () => {
-    expect(liveTurnTokens({ id: "x", role: "assistant" } as UIMessage)).toEqual({
-      reasoning: 0,
-      output: 0,
-      authoritative: false,
-    });
-  });
-
-  it("estimates output from text parts", () => {
-    // 8 chars -> 2 tokens.
-    const r = liveTurnTokens(msg([{ type: "text", text: "12345678" }]));
-    expect(r).toEqual({ reasoning: 0, output: 2, authoritative: false });
-  });
-
-  it("estimates reasoning from reasoning parts (kept separate from output)", () => {
-    const r = liveTurnTokens(
-      msg([
-        { type: "reasoning", text: "12345678" },
-        { type: "text", text: "abcd" },
-      ]),
-    );
-    expect(r).toEqual({ reasoning: 2, output: 1, authoritative: false });
-  });
-
-  it("accumulates across multiple text + reasoning parts (multi-step)", () => {
-    const r = liveTurnTokens(
-      msg([
-        { type: "reasoning", text: "abcd" }, // 1
-        { type: "text", text: "abcd" }, // 1
-        { type: "tool-getPage", state: "output-available" }, // ignored
-        { type: "reasoning", text: "abcd" }, // 1
-        { type: "text", text: "abcdefgh" }, // 2
-      ]),
-    );
-    expect(r).toEqual({ reasoning: 2, output: 3, authoritative: false });
-  });
-
-  it("ignores non text/reasoning parts (tools, step-start)", () => {
-    const r = liveTurnTokens(
-      msg([
-        { type: "step-start" },
-        { type: "tool-getPage", state: "input-available" },
-      ]),
-    );
-    expect(r).toEqual({ reasoning: 0, output: 0, authoritative: false });
-  });
-});
-
-describe("liveTurnTokens — authoritative path", () => {
-  it("returns authoritative usage verbatim, splitting reasoning out of output", () => {
-    // outputTokens INCLUDES reasoning in the AI SDK shape -> answer = 100 - 30.
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: "estimate would be tiny" }], {
-        usage: { inputTokens: 500, outputTokens: 100, reasoningTokens: 30 },
-      }),
-    );
-    expect(r).toEqual({ reasoning: 30, output: 70, authoritative: true });
-  });
-
-  it("treats missing reasoningTokens as 0 and keeps full output", () => {
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: "x" }], {
-        usage: { inputTokens: 10, outputTokens: 42 },
-      }),
-    );
-    expect(r).toEqual({ reasoning: 0, output: 42, authoritative: true });
-  });
-
-  it("never returns a negative output when reasoning exceeds reported output", () => {
-    const r = liveTurnTokens(
-      msg([], { usage: { outputTokens: 10, reasoningTokens: 40 } }),
-    );
-    expect(r).toEqual({ reasoning: 40, output: 0, authoritative: true });
-  });
-
-  it("falls back to the estimate when metadata has no usage object", () => {
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: "abcd" }], { chatId: "c1" }),
-    );
-    expect(r).toEqual({ reasoning: 0, output: 1, authoritative: false });
-  });
-});
-
-describe("liveTurnTokens — combined authoritative + estimate (#163)", () => {
-  it("ticks the in-flight step above the completed-steps authoritative base", () => {
-    // The authoritative usage is the sum over COMPLETED steps (step 1). The
-    // CURRENT step is streaming and its text is NOT in `usage` yet, but it IS in
-    // the parts -> the running estimate must push the live figure above the base
-    // so the badge keeps growing between step boundaries.
-    const longText = "x".repeat(800); // 800 chars -> 200 est output tokens
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: longText }], {
-        usage: { inputTokens: 500, outputTokens: 40 }, // step-1 base: 40 output
-      }),
-    );
-    // max(authOutput=40, estOutput=200) = 200 -> the counter ticks, not frozen.
-    expect(r.output).toBe(200);
-    expect(r.authoritative).toBe(true);
-  });
-
-  it("ticks reasoning of the in-flight step above the authoritative reasoning base", () => {
-    const longReasoning = "r".repeat(400); // 400 chars -> 100 est reasoning
-    const r = liveTurnTokens(
-      msg([{ type: "reasoning", text: longReasoning }], {
-        usage: { inputTokens: 100, outputTokens: 20, reasoningTokens: 20 },
-      }),
-    );
-    // reasoning: max(20, 100) = 100 ; output: max(max(0,20-20)=0, 0) = 0.
-    expect(r.reasoning).toBe(100);
-    expect(r.output).toBe(0);
-    expect(r.authoritative).toBe(true);
-  });
-
-  it("snaps to the authoritative figure once it exceeds the rough estimate", () => {
-    // Short on-screen text (estimate tiny) but a large authoritative output:
-    // the exact figure wins at the boundary (the counter never under-reports).
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: "abcd" }], {
-        usage: { inputTokens: 10, outputTokens: 5000 },
-      }),
-    );
-    expect(r.output).toBe(5000);
-  });
-
-  it("is monotonic: max never drops below the authoritative base when the estimate is smaller", () => {
-    // Mirrors the legacy 'verbatim' tests: estimate < authoritative -> unchanged.
-    const r = liveTurnTokens(
-      msg([{ type: "text", text: "tiny" }], {
-        usage: { inputTokens: 500, outputTokens: 100, reasoningTokens: 30 },
-      }),
-    );
-    expect(r).toEqual({ reasoning: 30, output: 70, authoritative: true });
-  });
-});

apps/client/src/features/ai-chat/utils/count-stream-tokens.ts

@@ -1,18 +1,11 @@
-import type { UIMessage } from "@ai-sdk/react";
-
 /**
- * Live token counting for a streaming AI-chat turn — split into REASONING
- * (thinking) and OUTPUT (answer) tokens, mirroring how Claude Code shows
- * `Thinking… · 60 tokens` next to its thinking indicator.
+ * Rough client-side token estimation for AI-chat UI affordances.
  *
- * No provider streams exact per-token usage mid-stream, so the live number is a
- * CLIENT ESTIMATE (chars/≈4 heuristic) that is reconciled to AUTHORITATIVE usage
- * once the server attaches it on a step/turn boundary (see the server's
- * `chatStreamMetadata` + the client's read of `message.metadata.usage`). When
- * authoritative usage is present we return it verbatim (the number "jumps to
- * exact"); otherwise we return the running estimate. Pure + unit-testable: it
- * never runs a real BPE tokenizer (that would be O(n²) on the hot path, bloat the
- * bundle, and be wrong for Gemini/Ollama anyway).
+ * No provider streams exact per-token usage mid-stream, so any in-flight figure
+ * is a CLIENT ESTIMATE (chars/≈4 heuristic). Pure + unit-testable: it never runs
+ * a real BPE tokenizer (that would be O(n²) on the hot path, bloat the bundle,
+ * and be wrong for Gemini/Ollama anyway). Used by the in-body reasoning counter
+ * ("Thinking · N tokens").
  */
 
 /**
@@ -24,90 +17,3 @@ export function estimateTokens(text: string): number {
   if (!text) return 0;
   return Math.ceil(text.length / 4);
 }
-
-/** Authoritative per-step/turn usage the server attaches to message metadata. */
-export interface AuthoritativeUsage {
-  inputTokens?: number;
-  outputTokens?: number;
-  totalTokens?: number;
-  reasoningTokens?: number;
-}
-
-/** Live token split for a turn's tail (streaming) assistant message. */
-export interface LiveTurnTokens {
-  /** Thinking/reasoning tokens (estimate, or authoritative when available). */
-  reasoning: number;
-  /** Answer/output tokens (estimate, or authoritative when available). */
-  output: number;
-  /** True when the numbers come from authoritative server usage, not estimate. */
-  authoritative: boolean;
-}
-
-/** Read the authoritative usage off a UIMessage's metadata, if the server set it. */
-function metadataUsage(message: UIMessage): AuthoritativeUsage | undefined {
-  const meta = message?.metadata as
-    | { usage?: AuthoritativeUsage }
-    | undefined;
-  const usage = meta?.usage;
-  if (!usage || typeof usage !== "object") return undefined;
-  return usage;
-}
-
-/**
- * Token split for the given (streaming) assistant message.
- *
- * COMBINES the authoritative server usage with the running text estimate so the
- * counter ticks in real time AND lands exact. The server only attaches
- * `metadata.usage` at a step/turn boundary (`finish-step`/`finish`) and it is
- * CUMULATIVE over COMPLETED steps — it does NOT yet include the in-flight step.
- * So a multi-step turn that returned the authoritative figure verbatim would
- * FREEZE between boundaries and jump in steps (issue #163).
- *
- * Instead we always compute the running ESTIMATE (chars/≈4 over the message's
- * `reasoning`/`text` parts, which grows on every streamed delta) and take the
- * per-component MAX of the authoritative base and the estimate:
- *   - between boundaries the estimate of the in-flight step ticks the number up;
- *   - at a boundary the authoritative figure snaps it to exact;
- *   - because the server's usage is cumulative and we only ever take the max, the
- *     number is MONOTONIC — it never drops.
- *
- * Providers that don't stream reasoning text still surface a reasoning count once
- * the authoritative usage arrives (`max(reasoningTokens, 0)`); on the pure
- * estimate path (no usage yet) such a turn shows `reasoning: 0` until then.
- */
-export function liveTurnTokens(message: UIMessage | undefined): LiveTurnTokens {
-  if (!message) return { reasoning: 0, output: 0, authoritative: false };
-
-  // Running ESTIMATE over every reasoning/text part — grows on each delta. This
-  // includes the IN-FLIGHT step, which the authoritative usage does not cover yet.
-  let estReasoning = 0;
-  let estOutput = 0;
-  for (const part of message.parts ?? []) {
-    if (part.type === "reasoning") {
-      estReasoning += estimateTokens((part as { text?: string }).text ?? "");
-    } else if (part.type === "text") {
-      estOutput += estimateTokens((part as { text?: string }).text ?? "");
-    }
-  }
-
-  const usage = metadataUsage(message);
-  if (!usage) {
-    // No authoritative usage streamed yet: the estimate IS the live figure.
-    return { reasoning: estReasoning, output: estOutput, authoritative: false };
-  }
-
-  // Authoritative sum over COMPLETED steps. `outputTokens` already INCLUDES
-  // reasoning in the AI SDK usage shape, so subtract it out for the "answer"
-  // figure (never go negative if a provider reports them inconsistently).
-  const authReasoning = usage.reasoningTokens ?? 0;
-  const authOutput = Math.max(0, (usage.outputTokens ?? 0) - authReasoning);
-
-  // Per-component max: the in-flight step's estimate ticks above the completed-
-  // steps base between boundaries, and the authoritative figure wins once it
-  // exceeds the (rough) estimate at the next boundary. Monotonic by construction.
-  return {
-    reasoning: Math.max(authReasoning, estReasoning),
-    output: Math.max(authOutput, estOutput),
-    authoritative: true,
-  };
-}

apps/client/src/features/ai-chat/utils/message-signature.test.ts

@@ -0,0 +1,241 @@
+import { describe, expect, it } from "vitest";
+import type { UIMessage } from "@ai-sdk/react";
+import { messageSignature } from "@/features/ai-chat/utils/message-signature.ts";
+
+/**
+ * Pure-helper tests for `messageSignature`, the cheap per-message content
+ * signature that drives MessageItem's memo (a streaming row's signature must
+ * change on every delta so it re-renders, while a finalized row's stays stable
+ * so it is skipped). Each test exercises ONE change signal and asserts it flips
+ * the signature; a content-identical clone must keep an EQUAL signature.
+ *
+ * The signature embeds `message.id` and `message.role`, so the `msg` factory
+ * uses a FIXED id/role here (not `Math.random()`): otherwise two messages with
+ * identical content would get different signatures and the negative case would
+ * be impossible to express.
+ */
+const msg = (
+  parts: UIMessage["parts"],
+  metadata?: unknown,
+): UIMessage =>
+  ({
+    id: "m1",
+    role: "assistant",
+    parts,
+    metadata,
+  }) as UIMessage;
+
+describe("messageSignature", () => {
+  it("changes when a text part grows", () => {
+    const before = msg([{ type: "text", text: "alpha" }]);
+    const after = msg([{ type: "text", text: "alpha beta" }]);
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when a new part is appended", () => {
+    const before = msg([{ type: "text", text: "alpha" }]);
+    const after = msg([
+      { type: "text", text: "alpha" },
+      { type: "text", text: "beta" },
+    ]);
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when a part's state flips", () => {
+    const before = msg([
+      { type: "tool-getPage", state: "input-streaming" } as never,
+    ]);
+    const after = msg([
+      { type: "tool-getPage", state: "output-available" } as never,
+    ]);
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when a tool part gains an output", () => {
+    const before = msg([
+      { type: "tool-getPage", state: "output-available" } as never,
+    ]);
+    const after = msg([
+      {
+        type: "tool-getPage",
+        state: "output-available",
+        output: { ok: true },
+      } as never,
+    ]);
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when a part gains an errorText", () => {
+    const before = msg([
+      { type: "tool-getPage", state: "output-error" } as never,
+    ]);
+    const after = msg([
+      {
+        type: "tool-getPage",
+        state: "output-error",
+        errorText: "boom",
+      } as never,
+    ]);
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when usage.reasoningTokens arrives on finish-step (text/state already frozen)", () => {
+    // The specifically-commented edge case: the authoritative turn total lands on
+    // the final finish-step AFTER the reasoning text length and state are frozen.
+    // Only the token count appears between these two snapshots, so the signature
+    // MUST still flip — otherwise the "Thinking · N tokens" header would never
+    // snap from the live estimate to the exact figure.
+    const before = msg([
+      { type: "reasoning", text: "thinking", state: "done" } as never,
+    ]);
+    const after = msg(
+      [{ type: "reasoning", text: "thinking", state: "done" } as never],
+      { usage: { reasoningTokens: 42 } },
+    );
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when metadata.error appears", () => {
+    const before = msg([{ type: "text", text: "answer" }]);
+    const after = msg([{ type: "text", text: "answer" }], { error: "boom" });
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("changes when metadata.finishReason changes (e.g. to 'aborted')", () => {
+    const before = msg([{ type: "text", text: "answer" }], {
+      finishReason: "stop",
+    });
+    const after = msg([{ type: "text", text: "answer" }], {
+      finishReason: "aborted",
+    });
+    expect(messageSignature(before)).not.toBe(messageSignature(after));
+  });
+
+  it("is UNCHANGED for a content-identical clone (different object, same values)", () => {
+    // A finalized row that is re-created as a fresh object (different parts array
+    // by reference, same parts by value) must keep an EQUAL signature, so the
+    // memo skips re-rendering it.
+    const a = msg([
+      { type: "text", text: "alpha" },
+      { type: "tool-getPage", state: "output-available", output: { ok: true } } as never,
+    ]);
+    const b = msg([
+      { type: "text", text: "alpha" },
+      { type: "tool-getPage", state: "output-available", output: { ok: true } } as never,
+    ]);
+    expect(a).not.toBe(b);
+    expect(messageSignature(a)).toBe(messageSignature(b));
+  });
+});
+
+/**
+ * Per-part-kind coupling guard for the load-bearing invariant documented at the
+ * top of message-signature.ts: the signature MUST sample every VISIBLE field the
+ * MessageItem render body draws, or the memo freezes a stale row. This is an
+ * executable lock for the part kinds rendered TODAY — read alongside
+ * `MessageItem` (message-item.tsx) and the `assistantMessageHasVisibleContent`
+ * helper (message-content.ts), which "mirrors MessageItem's render decisions
+ * EXACTLY". For each kind, mutating a field the render body DRAWS must flip the
+ * signature. If a new visible field is rendered without being added here AND to
+ * the signature, the corresponding assertion below should fail — that is the
+ * guard. (This intentionally stops short of the render-descriptor refactor:
+ * adding a part kind or a visible field still requires a human to extend both
+ * the signature and this block.)
+ */
+describe("messageSignature ↔ render coupling (per visible part kind)", () => {
+  describe("text part — render draws part.text (MarkdownPart text={part.text})", () => {
+    it("flips when the visible text changes", () => {
+      // Streaming is append-only, so the visible text only grows; the signature
+      // samples its length, so the growth is the change signal.
+      const before = msg([{ type: "text", text: "answer" }]);
+      const after = msg([{ type: "text", text: "answer extended" }]);
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+  });
+
+  describe("reasoning part — render draws text + tokens (ReasoningBlock)", () => {
+    it("flips when the visible reasoning text changes", () => {
+      const before = msg([
+        { type: "reasoning", text: "think", state: "streaming" } as never,
+      ]);
+      const after = msg([
+        { type: "reasoning", text: "think harder", state: "streaming" } as never,
+      ]);
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+
+    it("flips when the visible token count (metadata.usage.reasoningTokens) lands", () => {
+      // The header's "Thinking · N tokens" reads reasoningTokensForPart, fed by
+      // metadata.usage.reasoningTokens — a VISIBLE field that arrives on the final
+      // finish-step after text length and state are frozen.
+      const before = msg([
+        { type: "reasoning", text: "think", state: "done" } as never,
+      ]);
+      const after = msg(
+        [{ type: "reasoning", text: "think", state: "done" } as never],
+        { usage: { reasoningTokens: 99 } },
+      );
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+  });
+
+  describe("tool-* part — render draws state/errorText/citations (ToolCallCard)", () => {
+    it("flips when the run state changes (running ↔ done icon + label)", () => {
+      // toolRunState(part.state) selects the spinner/check/error icon.
+      const before = msg([
+        { type: "tool-getPage", state: "input-available" } as never,
+      ]);
+      const after = msg([
+        { type: "tool-getPage", state: "output-available" } as never,
+      ]);
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+
+    it("flips when output arrives (drives the rendered citation links)", () => {
+      // toolCitations reads part.output to render the "/p/{id}" anchors.
+      const before = msg([
+        { type: "tool-getPage", state: "output-available" } as never,
+      ]);
+      const after = msg([
+        {
+          type: "tool-getPage",
+          state: "output-available",
+          output: { id: "page-1", title: "Doc" },
+        } as never,
+      ]);
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+
+    it("flips when errorText appears (the visible red error detail line)", () => {
+      const before = msg([
+        { type: "tool-getPage", state: "output-error" } as never,
+      ]);
+      const after = msg([
+        {
+          type: "tool-getPage",
+          state: "output-error",
+          errorText: "permission denied",
+        } as never,
+      ]);
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+  });
+
+  describe("metadata banners — render draws error / aborted notices", () => {
+    it("flips when metadata.error appears (ChatErrorAlert banner)", () => {
+      const before = msg([{ type: "text", text: "answer" }]);
+      const after = msg([{ type: "text", text: "answer" }], { error: "boom" });
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+
+    it("flips when metadata.finishReason becomes 'aborted' (ChatStoppedNotice)", () => {
+      const before = msg([{ type: "text", text: "answer" }], {
+        finishReason: "stop",
+      });
+      const after = msg([{ type: "text", text: "answer" }], {
+        finishReason: "aborted",
+      });
+      expect(messageSignature(before)).not.toBe(messageSignature(after));
+    });
+  });
+});

apps/client/src/features/ai-chat/utils/message-signature.ts

@@ -0,0 +1,44 @@
+import type { UIMessage } from "@ai-sdk/react";
+
+/** Cheap content signature for one message: changes iff something VISIBLE in the
+ *  row changed. Streaming is APPEND-ONLY (text parts only grow, parts are only
+ *  appended, a tool/text part flips state once), so a per-part [type, text
+ *  length, state, error/output presence] tuple + the persisted metadata
+ *  (error/finishReason) is a sufficient change signal without comparing full
+ *  strings on every delta. WARNING — load-bearing for the MessageItem memo:
+ *  if a future part kind's VISIBLE content can change WITHOUT changing [type,
+ *  text length, state, error/output presence] (e.g. a tool that streams
+ *  `preliminary` output, or a client-side regenerate that edits a finalized
+ *  row in place), extend this signature or the memo will freeze a stale row. */
+export function messageSignature(message: UIMessage): string {
+  const parts = message.parts
+    .map((p) => {
+      const any = p as {
+        type: string;
+        text?: string;
+        state?: string;
+        errorText?: string;
+        output?: unknown;
+      };
+      return [
+        any.type,
+        any.text?.length ?? 0,
+        any.state ?? "",
+        any.errorText ? 1 : 0,
+        any.output !== undefined ? 1 : 0,
+      ].join(":");
+    })
+    .join("|");
+  const meta = message.metadata as
+    | { error?: string; finishReason?: string; usage?: { reasoningTokens?: number } }
+    | undefined;
+  // `usage.reasoningTokens` is neither append-only nor part-bound: the authoritative
+  // turn total arrives on the final `finish-step` AFTER the reasoning text length and
+  // state are already frozen. Without it in the signature the row's signature would be
+  // unchanged at that point and the re-render skipped, so the "Thinking · N tokens"
+  // header (reasoningTokensForPart) would keep the live estimate instead of snapping
+  // to the exact figure.
+  return `${message.id}#${message.role}#${parts}#${meta?.error ?? ""}#${
+    meta?.finishReason ?? ""
+  }#${meta?.usage?.reasoningTokens ?? ""}`;
+}

apps/client/src/features/editor/components/footnote/footnote-definition-view.tsx

@@ -7,7 +7,7 @@ import classes from "./footnote.module.css";
  * A 0-based backlink index -> its lowercase letter label (0 -> "a", 25 -> "z",
  * 26 -> "aa", ...), matching the Pandoc/Wikipedia "↩ a b c" convention.
  */
-function backlinkLabel(index: number): string {
+export function backlinkLabel(index: number): string {
   let out = "";
   let x = index;
   while (x >= 0) {

apps/client/src/features/editor/components/footnote/footnote-views.structure.test.tsx

@@ -75,7 +75,9 @@ vi.mock("@/features/editor/components/code-block/mermaid-view.tsx", () => ({
 }));
 
 import FootnotesListView from "./footnotes-list-view";
-import FootnoteDefinitionView from "./footnote-definition-view";
+import FootnoteDefinitionView, {
+  backlinkLabel,
+} from "./footnote-definition-view";
 import CodeBlockView from "../code-block/code-block-view";
 
 // Minimal NodeViewProps stub: definition view only touches node.attrs.id and
@@ -214,3 +216,16 @@ describe("#168 footnote definition multi-backlinks", () => {
     );
   });
 });
+
+// #185 re-review pt 7: backlinkLabel is base-26 (a..z, then aa…). The component
+// tests only cover a,b,c (index 0-2); pin the >= 26 carry boundary.
+describe("backlinkLabel base-26 boundary (#168)", () => {
+  it("maps 0->a, 25->z, 26->aa, 27->ab, 51->az, 52->ba", () => {
+    expect(backlinkLabel(0)).toBe("a");
+    expect(backlinkLabel(25)).toBe("z");
+    expect(backlinkLabel(26)).toBe("aa");
+    expect(backlinkLabel(27)).toBe("ab");
+    expect(backlinkLabel(51)).toBe("az");
+    expect(backlinkLabel(52)).toBe("ba");
+  });
+});

apps/client/src/features/editor/components/footnote/footnote.module.css

@@ -104,6 +104,19 @@
   min-width: 0;
 }
 
+/* The inner editable paragraph inherits `.ProseMirror p { margin: 0.5em 0 }`,
+   which pushes the first text line ~0.5em below the "N." marker (aligned to
+   flex-start), making the number float above the text. Drop the outer margins
+   so the marker and the first line share the same top edge — same approach
+   used for callouts in core.css. */
+.definitionContent > :first-child {
+  margin-top: 0;
+}
+
+.definitionContent > :last-child {
+  margin-bottom: 0;
+}
+
 .backLink {
   flex: 0 0 auto;
   cursor: pointer;

apps/client/src/features/editor/full-editor.tsx

@@ -26,6 +26,7 @@ import { FixedToolbar } from "@/features/editor/components/fixed-toolbar/fixed-t
 import { PageEditMode } from "@/features/user/types/user.types.ts";
 import { useAsideTriggerProps } from "@/hooks/use-toggle-aside.tsx";
 import { DeletedPageBanner } from "@/features/page/trash/components/deleted-page-banner.tsx";
+import { TemporaryNoteBanner } from "@/features/page/components/temporary-note-banner.tsx";
 import clsx from "clsx";
 import {
   currentPageEditModeAtom,
@@ -37,6 +38,7 @@ const MemoizedTitleEditor = React.memo(TitleEditor);
 const MemoizedPageEditor = React.memo(PageEditor);
 const MemoizedFixedToolbar = React.memo(FixedToolbar);
 const MemoizedDeletedPageBanner = React.memo(DeletedPageBanner);
+const MemoizedTemporaryNoteBanner = React.memo(TemporaryNoteBanner);
 
 type PageUser = {
   id: string;
@@ -103,6 +105,7 @@ export function FullEditor({
         <MemoizedFixedToolbar />
       )}
       <MemoizedDeletedPageBanner slugId={slugId} />
+      <MemoizedTemporaryNoteBanner slugId={slugId} />
       <MemoizedTitleEditor
         pageId={pageId}
         slugId={slugId}

apps/client/src/features/editor/styles/task-list.css

@@ -10,9 +10,15 @@ ul[data-type="taskList"] {
         display: flex;
 
         > label {
-            padding-top: 0.2rem;
+            /* Box exactly one text-line tall and center the checkbox in it, so the
+               checkbox lines up with the first line of the item's text. This tracks
+               the editor line-height (--mantine-line-height-xl) instead of a magic
+               padding-top that drifts from the real line box. */
             flex: 0 0 auto;
             margin-right: 0.5rem;
+            height: calc(var(--mantine-line-height-xl, 1.65) * 1em);
+            display: inline-flex;
+            align-items: center;
             user-select: none;
         }
 

apps/client/src/features/page-embed/queries/page-embed-query.ts

@@ -1,7 +1,40 @@
 import { useMutation } from "@tanstack/react-query";
 import { notifications } from "@mantine/notifications";
-import { toggleTemplate } from "@/features/page-embed/services/page-embed-api";
-import type { ToggleTemplateResponse } from "@/features/page-embed/types/page-embed.types";
+import {
+  toggleTemplate,
+  toggleTemporary,
+} from "@/features/page-embed/services/page-embed-api";
+import type {
+  ToggleTemplateResponse,
+  ToggleTemporaryResponse,
+} from "@/features/page-embed/types/page-embed.types";
+import { queryClient } from "@/main.tsx";
+
+/**
+ * After toggling a note's temporary state, mirror the new deadline into the
+ * shared page cache (keyed by both slugId and id) and refresh the sidebar so the
+ * menu label, the in-page banner, and the tree icon all reflect the change.
+ * Centralised here so the header menu and the banner can't drift apart on the
+ * cache-key plumbing.
+ */
+export function syncTemporaryExpiresInCache(
+  page: { id: string; slugId: string },
+  temporaryExpiresAt: string | null,
+) {
+  for (const key of [page.slugId, page.id]) {
+    const cached = queryClient.getQueryData<any>(["pages", key]);
+    if (cached) {
+      queryClient.setQueryData(["pages", key], {
+        ...cached,
+        temporaryExpiresAt,
+      });
+    }
+  }
+  queryClient.invalidateQueries({
+    predicate: (item) =>
+      ["sidebar-pages"].includes(item.queryKey[0] as string),
+  });
+}
 
 export function useToggleTemplateMutation() {
   return useMutation<
@@ -18,3 +51,20 @@ export function useToggleTemplateMutation() {
     },
   });
 }
+
+export function useToggleTemporaryMutation() {
+  return useMutation<
+    ToggleTemporaryResponse,
+    Error,
+    { pageId: string; temporary?: boolean }
+  >({
+    mutationFn: (data) => toggleTemporary(data),
+    onError: (err: any) => {
+      notifications.show({
+        message:
+          err?.response?.data?.message || "Failed to update temporary note",
+        color: "red",
+      });
+    },
+  });
+}

apps/client/src/features/page-embed/services/page-embed-api.ts

@@ -2,6 +2,7 @@ import api from "@/lib/api-client";
 import type {
   PageTemplateLookup,
   ToggleTemplateResponse,
+  ToggleTemporaryResponse,
 } from "../types/page-embed.types";
 
 export async function lookupTemplate(params: {
@@ -18,3 +19,11 @@ export async function toggleTemplate(params: {
   const r = await api.post("/pages/toggle-template", params);
   return r.data;
 }
+
+export async function toggleTemporary(params: {
+  pageId: string;
+  temporary?: boolean;
+}): Promise<ToggleTemporaryResponse> {
+  const r = await api.post("/pages/toggle-temporary", params);
+  return r.data;
+}

apps/client/src/features/page-embed/types/page-embed.types.ts

@@ -14,3 +14,9 @@ export type ToggleTemplateResponse = {
   pageId: string;
   isTemplate: boolean;
 };
+
+export type ToggleTemporaryResponse = {
+  pageId: string;
+  // null => the note was made permanent; ISO string => armed deadline.
+  temporaryExpiresAt: string | null;
+};

apps/client/src/features/page/components/header/page-header-menu.tsx

@@ -2,6 +2,7 @@ import { ActionIcon, Button, Group, Menu, Text, ThemeIcon, Tooltip } from "@mant
 import {
   IconArrowRight,
   IconArrowsHorizontal,
+  IconClockHour4,
   IconDots,
   IconEye,
   IconEyeOff,
@@ -24,6 +25,10 @@ import { useDisclosure, useHotkeys } from "@mantine/hooks";
 import { useClipboard } from "@/hooks/use-clipboard";
 import { useParams } from "react-router-dom";
 import { usePageQuery } from "@/features/page/queries/page-query.ts";
+import {
+  useToggleTemporaryMutation,
+  syncTemporaryExpiresInCache,
+} from "@/features/page-embed/queries/page-embed-query.ts";
 import { buildPageUrl } from "@/features/page/page.utils.ts";
 import { notifications } from "@mantine/notifications";
 import { getAppUrl } from "@/lib/config.ts";
@@ -160,6 +165,29 @@ function PageActionMenu({ readOnly }: PageActionMenuProps) {
   const { data: watchStatus } = useWatchStatusQuery(page?.id);
   const watchPage = useWatchPageMutation();
   const unwatchPage = useUnwatchPageMutation();
+  const toggleTemporary = useToggleTemporaryMutation();
+  const isTemporary = !!page?.temporaryExpiresAt;
+
+  const handleToggleTemporary = async () => {
+    if (!page?.id) return;
+    const next = !isTemporary;
+    try {
+      const res = await toggleTemporary.mutateAsync({
+        pageId: page.id,
+        temporary: next,
+      });
+      // Reflect the new deadline in the page cache so the menu label flips and
+      // any banner updates. The sidebar icon refreshes via its own query.
+      syncTemporaryExpiresInCache(page, res.temporaryExpiresAt);
+      notifications.show({
+        message: next
+          ? t("Note will move to trash unless made permanent")
+          : t("Note is now permanent"),
+      });
+    } catch {
+      // mutation surfaces the error via notifications
+    }
+  };
 
   const handleCopyLink = () => {
     const pageUrl =
@@ -309,6 +337,12 @@ function PageActionMenu({ readOnly }: PageActionMenuProps) {
           {!readOnly && (
             <>
               <Menu.Divider />
+              <Menu.Item
+                leftSection={<IconClockHour4 size={16} />}
+                onClick={handleToggleTemporary}
+              >
+                {isTemporary ? t("Make permanent") : t("Make temporary")}
+              </Menu.Item>
               <Menu.Item
                 color={"red"}
                 leftSection={<IconTrash size={16} />}

apps/client/src/features/page/components/temporary-note-banner.tsx

@@ -0,0 +1,87 @@
+import { Button, Group, Paper, Text } from "@mantine/core";
+import { IconClockHour4 } from "@tabler/icons-react";
+import { Trans, useTranslation } from "react-i18next";
+import { useTimeAgo } from "@/hooks/use-time-ago.tsx";
+import { usePageQuery } from "@/features/page/queries/page-query.ts";
+import {
+  useToggleTemporaryMutation,
+  syncTemporaryExpiresInCache,
+} from "@/features/page-embed/queries/page-embed-query.ts";
+import { useGetSpaceBySlugQuery } from "@/features/space/queries/space-query.ts";
+import { useSpaceAbility } from "@/features/space/permissions/use-space-ability.ts";
+import {
+  SpaceCaslAction,
+  SpaceCaslSubject,
+} from "@/features/space/permissions/permissions.type.ts";
+
+type TemporaryNoteBannerProps = {
+  slugId: string;
+};
+
+/**
+ * Banner shown on an open temporary note ("structure or die"). Mirrors
+ * DeletedPageBanner: it reads the page from the shared query cache and offers
+ * the explicit rescue action — "Make permanent". Children ride along to trash
+ * with the note, which is noted in the copy.
+ */
+export function TemporaryNoteBanner({ slugId }: TemporaryNoteBannerProps) {
+  const { t } = useTranslation();
+  const { data: page } = usePageQuery({ pageId: slugId });
+  const { data: space } = useGetSpaceBySlugQuery(page?.space?.slug);
+  const spaceAbility = useSpaceAbility(space?.membership?.permissions);
+  const expiresTimeAgo = useTimeAgo(page?.temporaryExpiresAt);
+  const toggleTemporary = useToggleTemporaryMutation();
+
+  // Don't show on a note that is already in trash; the deleted-page banner
+  // owns that state.
+  if (!page?.temporaryExpiresAt || page?.deletedAt) return null;
+
+  const canEdit = spaceAbility.can(SpaceCaslAction.Edit, SpaceCaslSubject.Page);
+
+  const handleMakePermanent = async () => {
+    try {
+      const res = await toggleTemporary.mutateAsync({
+        pageId: page.id,
+        temporary: false,
+      });
+      syncTemporaryExpiresInCache(page, res.temporaryExpiresAt);
+    } catch {
+      // mutation surfaces the error via notifications
+    }
+  };
+
+  return (
+    <Paper radius="sm" mb="md" px="md" py="xs" bg="orange.0">
+      <Group justify="space-between" wrap="wrap" gap="sm">
+        <Group gap="xs" wrap="nowrap" style={{ flex: 1, minWidth: 0 }}>
+          <IconClockHour4
+            size={18}
+            stroke={1.5}
+            style={{
+              flexShrink: 0,
+              color: "var(--mantine-color-orange-7)",
+            }}
+          />
+          <Text size="sm">
+            <Trans
+              i18nKey="This temporary note moves to trash {{time}} (with its sub-pages) unless made permanent."
+              values={{ time: expiresTimeAgo }}
+            />
+          </Text>
+        </Group>
+        {canEdit && (
+          <Button
+            size="xs"
+            variant="light"
+            color="orange"
+            leftSection={<IconClockHour4 size={16} />}
+            onClick={handleMakePermanent}
+            loading={toggleTemporary.isPending}
+          >
+            {t("Make permanent")}
+          </Button>
+        )}
+      </Group>
+    </Paper>
+  );
+}

apps/client/src/features/page/tree/components/space-tree-node-menu.tsx

@@ -6,6 +6,7 @@ import { useDisclosure } from "@mantine/hooks";
 import { notifications } from "@mantine/notifications";
 import {
   IconArrowRight,
+  IconClockHour4,
   IconCopy,
   IconDotsVertical,
   IconFileExport,
@@ -30,7 +31,10 @@ import {
   useRemoveFavoriteMutation,
 } from "@/features/favorite/queries/favorite-query";
 
-import { useToggleTemplateMutation } from "@/features/page-embed/queries/page-embed-query";
+import {
+  useToggleTemplateMutation,
+  useToggleTemporaryMutation,
+} from "@/features/page-embed/queries/page-embed-query";
 import { treeDataAtom } from "@/features/page/tree/atoms/tree-data-atom.ts";
 import { treeModel } from "@/features/page/tree/model/tree-model";
 import { useTreeMutation } from "@/features/page/tree/hooks/use-tree-mutation.ts";
@@ -65,6 +69,8 @@ export function NodeMenu({ node, canEdit }: NodeMenuProps) {
   const isFavorited = favoriteIds.has(node.id);
   const toggleTemplate = useToggleTemplateMutation();
   const isTemplate = !!node.isTemplate;
+  const toggleTemporary = useToggleTemporaryMutation();
+  const isTemporary = !!node.temporaryExpiresAt;
 
   const handleToggleTemplate = async () => {
     const next = !isTemplate;
@@ -84,6 +90,29 @@ export function NodeMenu({ node, canEdit }: NodeMenuProps) {
     }
   };
 
+  const handleToggleTemporary = async () => {
+    const next = !isTemporary;
+    try {
+      const res = await toggleTemporary.mutateAsync({
+        pageId: node.id,
+        temporary: next,
+      });
+      // Reflect the new deadline locally so the icon/menu update immediately.
+      setData((prev) =>
+        treeModel.update(prev, node.id, {
+          temporaryExpiresAt: res.temporaryExpiresAt,
+        } as any),
+      );
+      notifications.show({
+        message: next
+          ? t("Note will move to trash unless made permanent")
+          : t("Note is now permanent"),
+      });
+    } catch {
+      // mutation surfaces the error via notifications
+    }
+  };
+
   const handleCopyLink = () => {
     const pageUrl =
       getAppUrl() + buildPageUrl(spaceSlug, node.slugId, node.name);
@@ -248,6 +277,17 @@ export function NodeMenu({ node, canEdit }: NodeMenuProps) {
                 {isTemplate ? t("Unset as template") : t("Make template")}
               </Menu.Item>
 
+              <Menu.Item
+                leftSection={<IconClockHour4 size={16} />}
+                onClick={(e) => {
+                  e.preventDefault();
+                  e.stopPropagation();
+                  handleToggleTemporary();
+                }}
+              >
+                {isTemporary ? t("Make permanent") : t("Make temporary")}
+              </Menu.Item>
+
               <Menu.Divider />
               <Menu.Item
                 c="red"

apps/client/src/features/page/tree/components/space-tree-row.tsx

@@ -6,6 +6,7 @@ import { ActionIcon, rem, Tooltip } from "@mantine/core";
 import {
   IconChevronDown,
   IconChevronRight,
+  IconClockHour4,
   IconFileDescription,
   IconPlus,
   IconPointFilled,
@@ -191,6 +192,28 @@ export function SpaceTreeRow({
         </Tooltip>
       )}
 
+      {node.temporaryExpiresAt && (
+        <Tooltip
+          // Children ride along to trash with the note (recursive removePage).
+          label={t("Temporary note — moves to trash unless made permanent")}
+          withArrow
+        >
+          <IconClockHour4
+            size={14}
+            stroke={1.5}
+            // Same visual-only indicator pattern as the template icon, but
+            // orange to flag the impending death timer.
+            style={{
+              flexShrink: 0,
+              marginLeft: rem(4),
+              color: "var(--mantine-color-orange-6)",
+            }}
+            aria-label={t("Temporary note")}
+            role="img"
+          />
+        </Tooltip>
+      )}
+
       <div className={classes.actions}>
         <NodeMenu node={node} canEdit={canEdit} />
 

apps/client/src/features/page/tree/hooks/use-tree-mutation.ts

@@ -22,7 +22,10 @@ import { getSpaceUrl } from "@/lib/config.ts";
 
 export type UseTreeMutation = {
   handleMove: (sourceId: string, op: DropOp) => Promise<void>;
-  handleCreate: (parentId: string | null) => Promise<void>;
+  handleCreate: (
+    parentId: string | null,
+    opts?: { temporary?: boolean },
+  ) => Promise<void>;
   handleRename: (id: string, name: string) => Promise<void>;
   handleDelete: (id: string) => Promise<void>;
 };
@@ -119,9 +122,15 @@ export function useTreeMutation(spaceId: string): UseTreeMutation {
   );
 
   const handleCreate = useCallback(
-    async (parentId: string | null) => {
-      const payload: { spaceId: string; parentPageId?: string } = { spaceId };
+    async (parentId: string | null, opts?: { temporary?: boolean }) => {
+      const payload: {
+        spaceId: string;
+        parentPageId?: string;
+        temporary?: boolean;
+      } = { spaceId };
       if (parentId) payload.parentPageId = parentId;
+      // Ask the server to arm the death timer for a "temporary note".
+      if (opts?.temporary) payload.temporary = true;
 
       let createdPage: IPage;
       try {
@@ -138,6 +147,8 @@ export function useTreeMutation(spaceId: string): UseTreeMutation {
         spaceId: createdPage.spaceId,
         parentPageId: createdPage.parentPageId,
         hasChildren: false,
+        // Show the temporary-note icon immediately on optimistic insert.
+        temporaryExpiresAt: createdPage.temporaryExpiresAt,
         children: [],
       };
 

apps/client/src/features/page/tree/model/tree-model.test.ts

@@ -752,6 +752,27 @@ describe("treeModel.placeByPosition", () => {
     });
     expect(t.map((n) => n.id)).toEqual(["r1", "child", "r2", "rp"]);
   });
+
+  it("returns same reference (no-op) when the destination parent is inside the source's own subtree (#206 ui-state-races-1)", () => {
+    // Moving `a` under its own descendant `b` is a cycle. Without the guard,
+    // remove(a) drops b too and insertByPosition can't re-place a -> the whole
+    // subtree silently vanishes. The guard refuses the move (same reference).
+    const cyclic: P[] = [
+      {
+        id: "a",
+        name: "A",
+        position: "a0",
+        children: [{ id: "b", name: "B", position: "a1" }],
+      },
+    ];
+    const t = treeModel.placeByPosition(cyclic, "a", {
+      parentId: "b",
+      position: "a5",
+    });
+    expect(t).toBe(cyclic);
+    expect(treeModel.find(t, "a")).not.toBeNull();
+    expect(treeModel.find(t, "b")).not.toBeNull();
+  });
 });
 
 describe("treeModel.move", () => {

apps/client/src/features/page/tree/model/tree-model.ts

@@ -294,6 +294,20 @@ export const treeModel = {
     const source = treeModel.find(tree, sourceId);
     if (!source) return tree;
     if (to.parentId !== null && !treeModel.find(tree, to.parentId)) return tree;
+    // Cycle guard, mirroring `move`'s `isDescendant` check (#206 ui-state-races-1).
+    // If the destination parent is INSIDE the moved node's own subtree (reachable
+    // when server-authoritative move events arrive out of order — e.g. X moved
+    // under Y, then Y under X, but on this receiver Y is still inside X), then
+    // `remove(sourceId)` would drop the future parent along with the whole subtree
+    // and `insertByPosition` could not find it again — the node and ALL its
+    // descendants would silently vanish. Refuse the move and return the same
+    // reference so callers can detect the no-op and reconcile (refetch) instead.
+    if (
+      to.parentId !== null &&
+      treeModel.isDescendant(tree, sourceId, to.parentId)
+    ) {
+      return tree;
+    }
     const removed = treeModel.remove(tree, sourceId);
     // Reuse the same position-ordered insertion as `insertByPosition` by
     // stamping the authoritative position onto the moved node first.

apps/client/src/features/page/tree/types.ts

@@ -9,5 +9,7 @@ export type SpaceTreeNode = {
   hasChildren: boolean;
   canEdit?: boolean;
   isTemplate?: boolean;
+  // Death-timer deadline. null/absent => permanent; ISO string => temporary note.
+  temporaryExpiresAt?: string | null;
   children: SpaceTreeNode[];
 };

apps/client/src/features/page/tree/utils/utils.ts

@@ -26,6 +26,7 @@ export function buildTree(pages: IPage[]): SpaceTreeNode[] {
       parentPageId: page.parentPageId,
       canEdit: page.canEdit ?? page.permissions?.canEdit,
       isTemplate: page.isTemplate,
+      temporaryExpiresAt: page.temporaryExpiresAt,
       children: [],
     };
   });

apps/client/src/features/page/types/page.types.ts

@@ -13,6 +13,10 @@ export interface IPage {
   workspaceId: string;
   isLocked: boolean;
   isTemplate?: boolean;
+  // Death-timer deadline. null/absent => permanent; ISO string => temporary note.
+  temporaryExpiresAt?: string | null;
+  // Create-only input flag: ask the server to arm the timer on a new page.
+  temporary?: boolean;
   lastUpdatedById: string;
   createdAt: Date;
   updatedAt: Date;

apps/client/src/features/share/components/share-alias-section.tsx

@@ -0,0 +1,237 @@
+import {
+  ActionIcon,
+  Button,
+  Group,
+  Modal,
+  Text,
+  TextInput,
+} from "@mantine/core";
+import { IconExternalLink } from "@tabler/icons-react";
+import { useEffect, useMemo, useRef, useState } from "react";
+import { useTranslation } from "react-i18next";
+import CopyTextButton from "@/components/common/copy.tsx";
+import { getAppUrl } from "@/lib/config.ts";
+import {
+  useRemoveShareAliasMutation,
+  useSetShareAliasMutation,
+  useShareAliasForPageQuery,
+} from "@/features/share/queries/share-query.ts";
+import { checkShareAliasAvailability } from "@/features/share/services/share-service.ts";
+import {
+  isValidShareAlias,
+  normalizeShareAlias,
+} from "@/features/share/share-alias.util.ts";
+
+interface ShareAliasSectionProps {
+  pageId: string;
+  readOnly: boolean;
+}
+
+// The prefix label shown next to the slug input, e.g. "docs.example.com/l/".
+function aliasPrefixLabel(): string {
+  const url = getAppUrl();
+  const host = url.replace(/^https?:\/\//, "").replace(/\/+$/, "");
+  return `${host}/l/`;
+}
+
+export default function ShareAliasSection({
+  pageId,
+  readOnly,
+}: ShareAliasSectionProps) {
+  const { t } = useTranslation();
+  const { data: currentAlias } = useShareAliasForPageQuery(pageId);
+  const setAliasMutation = useSetShareAliasMutation();
+  const removeAliasMutation = useRemoveShareAliasMutation();
+
+  const [value, setValue] = useState("");
+  const [availability, setAvailability] = useState<{
+    valid: boolean;
+    available: boolean;
+    currentPageId: string | null;
+  } | null>(null);
+  const [reassign, setReassign] = useState<{
+    alias: string;
+    currentPageTitle: string | null;
+  } | null>(null);
+
+  // Seed the input from the page's current alias (if any).
+  useEffect(() => {
+    setValue(currentAlias?.alias ?? "");
+  }, [currentAlias?.alias, pageId]);
+
+  const normalized = useMemo(() => normalizeShareAlias(value), [value]);
+  const isValid = isValidShareAlias(normalized);
+  const unchanged = currentAlias?.alias === normalized;
+
+  // Debounced availability probe (skips when invalid or unchanged).
+  const debounceRef = useRef<ReturnType<typeof setTimeout>>();
+  useEffect(() => {
+    setAvailability(null);
+    if (!isValid || unchanged) return;
+    debounceRef.current && clearTimeout(debounceRef.current);
+    debounceRef.current = setTimeout(async () => {
+      try {
+        const res = await checkShareAliasAvailability(normalized);
+        setAvailability({
+          valid: res.valid,
+          available: res.available,
+          currentPageId: res.currentPageId,
+        });
+      } catch {
+        setAvailability(null);
+      }
+    }, 400);
+    return () => {
+      debounceRef.current && clearTimeout(debounceRef.current);
+    };
+  }, [normalized, isValid, unchanged]);
+
+  const prettyLink = currentAlias?.alias
+    ? `${getAppUrl()}/l/${currentAlias.alias}`
+    : null;
+
+  const handleSave = async (confirmReassign = false) => {
+    try {
+      await setAliasMutation.mutateAsync({
+        pageId,
+        alias: normalized,
+        confirmReassign,
+      });
+      setReassign(null);
+    } catch (error: any) {
+      // The address already points at another page: prompt to move it here.
+      if (error?.status === 409 || error?.response?.status === 409) {
+        const data = error?.response?.data;
+        if (data?.code === "ALIAS_REASSIGN_REQUIRED") {
+          setReassign({
+            alias: normalized,
+            currentPageTitle: data?.currentPageTitle ?? null,
+          });
+        }
+      }
+    }
+  };
+
+  const handleRemove = async () => {
+    if (!currentAlias?.id) return;
+    await removeAliasMutation.mutateAsync(currentAlias.id);
+    setValue("");
+  };
+
+  const showInvalid = normalized.length > 0 && !isValid;
+  const showTaken =
+    isValid && !unchanged && availability && !availability.available;
+
+  return (
+    <>
+      <Text size="sm" fw={500} mt="md">
+        {t("Custom address")}
+      </Text>
+      <Text size="xs" c="dimmed" mb={4}>
+        {t("A short, memorable link you can point at any shared page.")}
+      </Text>
+
+      {prettyLink && (
+        <Group my="xs" gap={4} wrap="nowrap">
+          <TextInput
+            variant="filled"
+            value={prettyLink}
+            readOnly
+            rightSection={<CopyTextButton text={prettyLink} />}
+            style={{ width: "100%" }}
+          />
+          <ActionIcon
+            component="a"
+            variant="default"
+            target="_blank"
+            href={prettyLink}
+            size="sm"
+          >
+            <IconExternalLink size={16} />
+          </ActionIcon>
+        </Group>
+      )}
+
+      <TextInput
+        value={value}
+        onChange={(e) => setValue(e.currentTarget.value)}
+        // Show the canonical form once the user pauses so what they type maps
+        // visibly to what gets stored.
+        onBlur={() => setValue(normalized)}
+        leftSection={
+          <Text size="xs" c="dimmed" pl={4} style={{ whiteSpace: "nowrap" }}>
+            {aliasPrefixLabel()}
+          </Text>
+        }
+        leftSectionWidth={Math.min(aliasPrefixLabel().length * 7 + 12, 180)}
+        placeholder={t("my-page")}
+        disabled={readOnly}
+        error={
+          showInvalid
+            ? t("Use 2-60 lowercase letters, digits and hyphens")
+            : showTaken
+              ? t("This address is already in use")
+              : undefined
+        }
+      />
+
+      <Group mt="xs" gap="xs">
+        <Button
+          size="compact-sm"
+          onClick={() => handleSave(false)}
+          loading={setAliasMutation.isPending}
+          disabled={readOnly || !isValid || unchanged}
+        >
+          {t("Save")}
+        </Button>
+        {currentAlias?.id && (
+          <Button
+            size="compact-sm"
+            variant="default"
+            color="red"
+            onClick={handleRemove}
+            loading={removeAliasMutation.isPending}
+            disabled={readOnly}
+          >
+            {t("Remove")}
+          </Button>
+        )}
+      </Group>
+
+      <Modal
+        opened={!!reassign}
+        onClose={() => setReassign(null)}
+        title={t("Move custom address?")}
+        centered
+        size="sm"
+      >
+        <Text size="sm">
+          {reassign?.currentPageTitle
+            ? t(
+                'The address "{{alias}}" currently points to "{{title}}". Move it to this page?',
+                {
+                  alias: reassign?.alias,
+                  title: reassign?.currentPageTitle,
+                },
+              )
+            : t(
+                'The address "{{alias}}" is already in use. Move it to this page?',
+                { alias: reassign?.alias },
+              )}
+        </Text>
+        <Group justify="flex-end" mt="md">
+          <Button variant="default" onClick={() => setReassign(null)}>
+            {t("Cancel")}
+          </Button>
+          <Button
+            color="red"
+            onClick={() => handleSave(true)}
+            loading={setAliasMutation.isPending}
+          >
+            {t("Move here")}
+          </Button>
+        </Group>
+      </Modal>
+    </>
+  );
+}

apps/client/src/features/share/components/share-modal.tsx

@@ -25,6 +25,7 @@ import CopyTextButton from "@/components/common/copy.tsx";
 import { getAppUrl } from "@/lib/config.ts";
 import { buildPageUrl } from "@/features/page/page.utils.ts";
 import classes from "@/features/share/components/share.module.css";
+import ShareAliasSection from "@/features/share/components/share-alias-section.tsx";
 import { useAtom } from "jotai";
 import { workspaceAtom } from "@/features/user/atoms/current-user-atom.ts";
 import { useSpaceQuery } from "@/features/space/queries/space-query.ts";
@@ -253,6 +254,9 @@ export default function ShareModal({ readOnly }: ShareModalProps) {
                     disabled={readOnly}
                   />
                 </Group>
+                {pageId && (
+                  <ShareAliasSection pageId={pageId} readOnly={readOnly} />
+                )}
               </>
             )}
           </>

apps/client/src/features/share/queries/share-query.ts

@@ -10,6 +10,8 @@ import { useTranslation } from "react-i18next";
 import {
   ICreateShare,
   IShare,
+  IShareAlias,
+  ISetShareAlias,
   ISharedItem,
   ISharedPage,
   ISharedPageTree,
@@ -20,11 +22,14 @@ import {
 import {
   createShare,
   deleteShare,
+  getShareAliasForPage,
   getSharedPageTree,
   getShareForPage,
   getShareInfo,
   getSharePageInfo,
   getShares,
+  removeShareAlias,
+  setShareAlias,
   updateShare,
 } from "@/features/share/services/share-service.ts";
 import { IPagination, QueryParams } from "@/lib/types.ts";
@@ -170,6 +175,72 @@ export function useDeleteShareMutation() {
   });
 }
 
+export function useShareAliasForPageQuery(
+  pageId: string,
+): UseQueryResult<IShareAlias | null, Error> {
+  return useQuery({
+    // The endpoint resolves to null when the page has no alias; normalize the
+    // absence so React Query never sees `undefined`.
+    queryKey: ["share-alias-for-page", pageId],
+    queryFn: async () => (await getShareAliasForPage(pageId)) ?? null,
+    enabled: !!pageId,
+    staleTime: 60 * 1000,
+    retry: false,
+  });
+}
+
+export function useSetShareAliasMutation() {
+  const { t } = useTranslation();
+  const queryClient = useQueryClient();
+
+  return useMutation<IShareAlias, Error, ISetShareAlias>({
+    mutationFn: (data) => setShareAlias(data),
+    onSuccess: () => {
+      queryClient.invalidateQueries({
+        predicate: (item) =>
+          ["share-alias-for-page", "share-list"].includes(
+            item.queryKey[0] as string,
+          ),
+      });
+    },
+    onError: (error) => {
+      // A 409 reassign-required is handled inline by the modal (it shows the
+      // "move address here?" confirmation), so don't surface a generic toast.
+      if (error?.["status"] === 409) return;
+      notifications.show({
+        message:
+          error?.["response"]?.data?.message || t("Failed to set custom address"),
+        color: "red",
+      });
+    },
+  });
+}
+
+export function useRemoveShareAliasMutation() {
+  const { t } = useTranslation();
+  const queryClient = useQueryClient();
+
+  return useMutation<void, Error, string>({
+    mutationFn: (aliasId) => removeShareAlias(aliasId),
+    onSuccess: () => {
+      queryClient.invalidateQueries({
+        predicate: (item) =>
+          ["share-alias-for-page", "share-list"].includes(
+            item.queryKey[0] as string,
+          ),
+      });
+    },
+    onError: (error) => {
+      notifications.show({
+        message:
+          error?.["response"]?.data?.message ||
+          t("Failed to remove custom address"),
+        color: "red",
+      });
+    },
+  });
+}
+
 export function useGetSharedPageTreeQuery(
   shareId: string,
 ): UseQueryResult<ISharedPageTree, Error> {

apps/client/src/features/share/services/share-service.ts

@@ -4,6 +4,9 @@ import { IPage } from "@/features/page/types/page.types";
 import {
   ICreateShare,
   IShare,
+  IShareAlias,
+  IShareAliasAvailability,
+  ISetShareAlias,
   ISharedItem,
   ISharedPage,
   ISharedPageTree,
@@ -57,3 +60,33 @@ export async function getSharedPageTree(
   const req = await api.post<ISharedPageTree>("/shares/tree", { shareId });
   return req.data;
 }
+
+export async function getShareAliasForPage(
+  pageId: string,
+): Promise<IShareAlias | null> {
+  const req = await api.post<IShareAlias | null>("/share-aliases/for-page", {
+    pageId,
+  });
+  return req.data;
+}
+
+export async function setShareAlias(
+  data: ISetShareAlias,
+): Promise<IShareAlias> {
+  const req = await api.post<IShareAlias>("/share-aliases/set", data);
+  return req.data;
+}
+
+export async function removeShareAlias(aliasId: string): Promise<void> {
+  await api.post("/share-aliases/remove", { aliasId });
+}
+
+export async function checkShareAliasAvailability(
+  alias: string,
+): Promise<IShareAliasAvailability> {
+  const req = await api.post<IShareAliasAvailability>(
+    "/share-aliases/availability",
+    { alias },
+  );
+  return req.data;
+}

apps/client/src/features/share/share-alias.util.test.ts

@@ -0,0 +1,32 @@
+import { describe, it, expect } from "vitest";
+import {
+  isValidShareAlias,
+  normalizeShareAlias,
+} from "@/features/share/share-alias.util.ts";
+
+// Mirrors the server-side util so the modal's live feedback matches what the
+// server will accept/store.
+describe("normalizeShareAlias", () => {
+  it("lowercases, trims and maps separators to single hyphens", () => {
+    expect(normalizeShareAlias("  My  Cool_Page ")).toBe("my-cool-page");
+  });
+
+  it("collapses repeated hyphens and trims edges", () => {
+    expect(normalizeShareAlias("--a---b--")).toBe("a-b");
+  });
+});
+
+describe("isValidShareAlias", () => {
+  it("accepts ascii hyphen-separated slugs of length 2..60", () => {
+    expect(isValidShareAlias("hello-world")).toBe(true);
+    expect(isValidShareAlias("a".repeat(60))).toBe(true);
+  });
+
+  it("rejects too short, edge/double hyphens, uppercase and non-ascii", () => {
+    expect(isValidShareAlias("a")).toBe(false);
+    expect(isValidShareAlias("-a")).toBe(false);
+    expect(isValidShareAlias("a--b")).toBe(false);
+    expect(isValidShareAlias("Hello")).toBe(false);
+    expect(isValidShareAlias("привет")).toBe(false);
+  });
+});

apps/client/src/features/share/share-alias.util.ts

@@ -0,0 +1,26 @@
+/**
+ * Client copy of the vanity share-alias helpers. Kept in sync with the server
+ * (`apps/server/src/core/share/share-alias.util.ts`) so live input feedback
+ * matches what the server will store/accept. ASCII-only, lowercase, hyphen
+ * separated, length 2..60.
+ */
+
+// Normalize a user-provided vanity alias into canonical ASCII storage form.
+export function normalizeShareAlias(raw: string): string {
+  return (raw ?? "")
+    .trim()
+    .toLowerCase()
+    .replace(/[\s_]+/g, "-")
+    .replace(/-{2,}/g, "-")
+    .replace(/^-+|-+$/g, "");
+}
+
+const ALIAS_RE = /^[a-z0-9]+(?:-[a-z0-9]+)*$/;
+export function isValidShareAlias(alias: string): boolean {
+  return (
+    typeof alias === "string" &&
+    alias.length >= 2 &&
+    alias.length <= 60 &&
+    ALIAS_RE.test(alias)
+  );
+}

apps/client/src/features/share/types/share.types.ts

@@ -75,6 +75,30 @@ export interface IShareInfoInput {
   pageId: string;
 }
 
+// Vanity /l/:alias pointer.
+export interface IShareAlias {
+  id: string;
+  workspaceId: string;
+  alias: string;
+  pageId: string | null;
+  creatorId: string | null;
+  createdAt: string;
+  updatedAt: string;
+}
+
+export interface ISetShareAlias {
+  pageId: string;
+  alias: string;
+  confirmReassign?: boolean;
+}
+
+export interface IShareAliasAvailability {
+  alias: string;
+  valid: boolean;
+  available: boolean;
+  currentPageId: string | null;
+}
+
 export interface ISharedPageTree {
   share: IShare;
   pageTree: Partial<IPage[]>;

apps/client/src/features/space/components/sidebar/space-sidebar.tsx

@@ -13,6 +13,7 @@ import {
   IconEye,
   IconEyeOff,
   IconFileExport,
+  IconHourglass,
   IconPlus,
   IconSettings,
   IconStar,
@@ -71,6 +72,10 @@ export function SpaceSidebar() {
     handleCreate(null);
   }
 
+  function handleCreateTemporaryPage() {
+    handleCreate(null, { temporary: true });
+  }
+
   return (
     <>
       <div className={classes.navbar}>
@@ -111,16 +116,39 @@ export function SpaceSidebar() {
                 SpaceCaslAction.Manage,
                 SpaceCaslSubject.Page,
               ) && (
-                <Tooltip label={t("Create page")} withArrow position="right">
-                  <ActionIcon
-                    variant="default"
-                    size={18}
-                    onClick={handleCreatePage}
-                    aria-label={t("Create page")}
+                <>
+                  <Tooltip
+                    label={t("Create page")}
+                    withArrow
+                    position="right"
                   >
-                    <IconPlus />
-                  </ActionIcon>
-                </Tooltip>
+                    <ActionIcon
+                      variant="default"
+                      size={18}
+                      onClick={handleCreatePage}
+                      aria-label={t("Create page")}
+                    >
+                      <IconPlus />
+                    </ActionIcon>
+                  </Tooltip>
+
+                  {/* Standalone second button: a "temporary note" auto-moves to
+                      trash after the workspace lifetime unless made permanent. */}
+                  <Tooltip
+                    label={t("New temporary note")}
+                    withArrow
+                    position="right"
+                  >
+                    <ActionIcon
+                      variant="default"
+                      size={18}
+                      onClick={handleCreateTemporaryPage}
+                      aria-label={t("New temporary note")}
+                    >
+                      <IconHourglass />
+                    </ActionIcon>
+                  </Tooltip>
+                </>
               )}
             </Group>
           </Group>

apps/client/src/features/websocket/tree-socket-reducers.test.ts

@@ -183,6 +183,34 @@ describe("applyMoveTreeNode", () => {
     expect(moved?.hasChildren).toBe(true);
     expect(moved?.position).toBe("a4");
   });
+
+  it("does NOT drop a subtree on a cyclic/out-of-order move (parent inside source) (#206 ui-state-races-1)", () => {
+    // Locally `b` is still nested inside `a` (an earlier "a under b" echo hasn't
+    // applied yet). An out-of-order "move a under b" event now arrives — b is a
+    // descendant of a, so re-parenting would make placeByPosition remove a (and
+    // its whole subtree, incl. b) and fail to re-insert. Before the fix BOTH a
+    // and b silently vanished; now the reducer leaves the tree untouched.
+    const tree: SpaceTreeNode[] = [
+      node("a", {
+        position: "a0",
+        hasChildren: true,
+        children: [node("b", { position: "a1", parentPageId: "a" })],
+      }),
+    ];
+    const next = applyMoveTreeNode(tree, {
+      id: "a",
+      parentId: "b",
+      oldParentId: null,
+      index: 0,
+      position: "a4",
+      pageData: {},
+    });
+    // No silent data loss: both nodes survive.
+    expect(treeModel.find(next, "a")).not.toBeNull();
+    expect(treeModel.find(next, "b")).not.toBeNull();
+    // The cyclic move is refused as a no-op (same reference) pending reconcile.
+    expect(next).toBe(tree);
+  });
 });
 
 describe("applyDeleteTreeNode", () => {

apps/client/src/features/websocket/tree-socket-reducers.ts

@@ -76,6 +76,19 @@ export function applyMoveTreeNode(
   const oldParentId = (sourceBefore as SpaceTreeNode).parentPageId ?? null;
   const newParentId = payload.parentId as string | null;
 
+  // Cyclic / out-of-order move guard (#206 ui-state-races-1): if the
+  // authoritative new parent is currently INSIDE the moved node's own subtree on
+  // this client (e.g. server moved X under Y then Y under X and the events
+  // arrived such that Y is still nested in X here), re-parenting is impossible to
+  // represent locally. `placeByPosition` returns `prev` for this, but the
+  // `placed === prev` fallback below would then `remove` the source — dropping
+  // the node AND every descendant (incl. the would-be parent) silently. Leave the
+  // tree untouched instead; a later corrective event or a reconnect refetch
+  // reconciles it. Never delete a subtree we cannot safely re-place.
+  if (newParentId && treeModel.isDescendant(prev, payload.id, newParentId)) {
+    return prev;
+  }
+
   // Place the node by its fractional `position` among the new siblings — NOT by
   // the sender's absolute `index` (the sender computed that against its own
   // loaded set, which differs from this receiver's). Using the position keeps

apps/client/src/features/workspace/components/settings/components/ai-mcp-server-test-view.test.ts

@@ -0,0 +1,87 @@
+import { describe, expect, it } from "vitest";
+import { mcpTestButtonView } from "./ai-mcp-server-test-view";
+
+/**
+ * Pure-helper tests for the inline "Test" button presentation. Covers the four
+ * states (idle / loading is handled by the component's `isPending`, so here:
+ * idle / ok-with-tools / ok-without-tools / failed) and the tooltip text
+ * branches that are easiest to break silently.
+ */
+// Identity-ish translator that echoes the key and interpolates {{n}} so the
+// label/tooltip branches are observable without the real i18n bundle.
+const t = (key: string, options?: Record<string, unknown>): string =>
+  options && "n" in options
+    ? key.replace("{{n}}", String((options as { n: unknown }).n))
+    : key;
+
+describe("mcpTestButtonView", () => {
+  it("idle when there is no result", () => {
+    expect(mcpTestButtonView(undefined, t)).toEqual({
+      state: "idle",
+      color: undefined,
+      variant: "default",
+      label: "Test",
+      tooltip: "",
+    });
+  });
+
+  it("ok with tools lists them in the tooltip", () => {
+    expect(mcpTestButtonView({ ok: true, tools: ["a", "b"] }, t)).toEqual({
+      state: "ok",
+      color: "green",
+      variant: "light",
+      label: "OK · 2",
+      tooltip: "a, b",
+    });
+  });
+
+  it('ok with zero tools shows "No tools available"', () => {
+    expect(mcpTestButtonView({ ok: true, tools: [] }, t)).toEqual({
+      state: "ok",
+      color: "green",
+      variant: "light",
+      label: "OK · 0",
+      tooltip: "No tools available",
+    });
+  });
+
+  it("failed surfaces the error text in the tooltip", () => {
+    expect(
+      mcpTestButtonView({ ok: false, error: "402: nope" }, t),
+    ).toEqual({
+      state: "failed",
+      color: "red",
+      variant: "light",
+      label: "Failed",
+      tooltip: "402: nope",
+    });
+  });
+
+  it("failed when the request itself rejects (no result payload)", () => {
+    // 401/403/500/network: there is no { ok } body, only a thrown error. The
+    // row must still show a red "Failed" rather than reverting to idle "Test".
+    expect(
+      mcpTestButtonView(undefined, t, {
+        response: { data: { message: "Unauthorized" } },
+      }),
+    ).toEqual({
+      state: "failed",
+      color: "red",
+      variant: "light",
+      label: "Failed",
+      tooltip: "Unauthorized",
+    });
+  });
+
+  it("reject without a server message falls back to the generic label", () => {
+    // A bare network error (no response body) still surfaces as failed, using
+    // the i18n fallback for the tooltip.
+    expect(mcpTestButtonView(undefined, t, new Error("network down"))).toEqual({
+      state: "failed",
+      color: "red",
+      variant: "light",
+      label: "Failed",
+      tooltip: "Failed to update data",
+    });
+  });
+});

apps/client/src/features/workspace/components/settings/components/ai-mcp-server-test-view.ts

@@ -0,0 +1,90 @@
+import type { IAiMcpServerTestResult } from "@/features/workspace/services/ai-mcp-server-service.ts";
+
+/** Minimal translator shape (i18next `t`): key + optional interpolation. */
+type Translate = (key: string, options?: Record<string, unknown>) => string;
+
+/** Subset of an axios-style rejection we read for the reject tooltip. */
+type McpTestRequestError = {
+  response?: { data?: { message?: string } };
+};
+
+/**
+ * Best-effort extraction of a server-sent message from a rejected test request
+ * (axios stores it at `error.response.data.message`). Returns undefined for a
+ * bare/network error so the caller can fall back to a generic label.
+ */
+function readRequestErrorMessage(error: unknown): string | undefined {
+  if (error && typeof error === "object" && "response" in error) {
+    return (error as McpTestRequestError).response?.data?.message;
+  }
+  return undefined;
+}
+
+/**
+ * Presentation for the inline "Test" button, derived from the current test
+ * result tristate (no result yet / ok / failed). Color is never the only signal
+ * — the label and icon change too (a11y / colorblind-friendly). Kept as a single
+ * pure derivation (rather than two parallel if/else chains) so the button and
+ * tooltip can never drift apart, and so the text branches are unit-testable
+ * without rendering the row.
+ */
+export interface McpTestButtonView {
+  /** Tristate; the component maps this to the leftSection icon. */
+  state: "idle" | "ok" | "failed";
+  /** Mantine Button color; undefined = theme default (idle). */
+  color?: string;
+  /** Mantine Button variant. */
+  variant: string;
+  /** Translated button label. */
+  label: string;
+  /** Translated tooltip text; "" while there is no result (tooltip disabled). */
+  tooltip: string;
+}
+
+export function mcpTestButtonView(
+  result: IAiMcpServerTestResult | undefined,
+  t: Translate,
+  error?: unknown,
+): McpTestButtonView {
+  if (result?.ok) {
+    return {
+      state: "ok",
+      color: "green",
+      variant: "light",
+      label: t("OK · {{n}}", { n: result.tools.length }),
+      tooltip:
+        result.tools.length > 0
+          ? result.tools.join(", ")
+          : t("No tools available"),
+    };
+  }
+  if (result && result.ok === false) {
+    return {
+      state: "failed",
+      color: "red",
+      variant: "light",
+      label: t("Failed"),
+      tooltip: result.error,
+    };
+  }
+  if (error) {
+    // The test request itself rejected (401/403/500/network) — there is no
+    // `{ ok }` payload, so without this branch the row would silently revert to
+    // the idle "Test" instead of reporting the failure. Tooltip prefers the
+    // server-sent message, else the generic i18n fallback.
+    return {
+      state: "failed",
+      color: "red",
+      variant: "light",
+      label: t("Failed"),
+      tooltip: readRequestErrorMessage(error) ?? t("Failed to update data"),
+    };
+  }
+  return {
+    state: "idle",
+    color: undefined,
+    variant: "default",
+    label: t("Test"),
+    tooltip: "",
+  };
+}

apps/client/src/features/workspace/components/settings/components/ai-mcp-servers.tsx

@@ -1,4 +1,4 @@
-import { useState } from "react";
+import { useEffect, useState } from "react";
 import {
   ActionIcon,
   Badge,
@@ -10,18 +10,28 @@ import {
   Stack,
   Switch,
   Text,
+  Tooltip,
 } from "@mantine/core";
 import { useDisclosure } from "@mantine/hooks";
 import { modals } from "@mantine/modals";
-import { IconPencil, IconPlus, IconTrash } from "@tabler/icons-react";
+import {
+  IconCheck,
+  IconPencil,
+  IconPlugConnected,
+  IconPlus,
+  IconTrash,
+  IconX,
+} from "@tabler/icons-react";
 import { useTranslation } from "react-i18next";
 import useUserRole from "@/hooks/use-user-role.tsx";
 import {
   useAiMcpServersQuery,
   useDeleteAiMcpServerMutation,
+  useTestAiMcpServerMutation,
   useUpdateAiMcpServerMutation,
 } from "@/features/workspace/queries/ai-mcp-server-query.ts";
 import { IAiMcpServer } from "@/features/workspace/services/ai-mcp-server-service.ts";
+import { mcpTestButtonView } from "@/features/workspace/components/settings/components/ai-mcp-server-test-view.ts";
 import AiMcpServerForm from "./ai-mcp-server-form.tsx";
 
 /**
@@ -112,55 +122,15 @@ export default function AiMcpServers() {
 
       <Stack gap="xs" mt="sm">
         {servers?.map((server) => (
-          <Group key={server.id} justify="space-between" wrap="nowrap">
-            <Stack gap={2} style={{ minWidth: 0 }}>
-              <Group gap="xs">
-                <Text fw={500} truncate>
-                  {server.name}
-                </Text>
-                <Badge size="xs" variant="light">
-                  {server.transport.toUpperCase()}
-                </Badge>
-              </Group>
-              <Text
-                size="xs"
-                c="dimmed"
-                truncate
-                style={{ fontFamily: "ui-monospace, Menlo, monospace" }}
-              >
-                {server.url}
-              </Text>
-            </Stack>
-
-            <Group gap="xs" wrap="nowrap">
-              <Switch
-                size="sm"
-                checked={server.enabled}
-                aria-label={t("Enabled")}
-                onChange={(event) =>
-                  updateMutation.mutate({
-                    id: server.id,
-                    enabled: event.currentTarget.checked,
-                  })
-                }
-              />
-              <ActionIcon
-                variant="subtle"
-                aria-label={t("Edit")}
-                onClick={() => openEdit(server)}
-              >
-                <IconPencil size={16} />
-              </ActionIcon>
-              <ActionIcon
-                variant="subtle"
-                color="red"
-                aria-label={t("Delete")}
-                onClick={() => confirmDelete(server)}
-              >
-                <IconTrash size={16} />
-              </ActionIcon>
-            </Group>
-          </Group>
+          <AiMcpServerRow
+            key={server.id}
+            server={server}
+            onEdit={openEdit}
+            onDelete={confirmDelete}
+            onToggleEnabled={(enabled) =>
+              updateMutation.mutate({ id: server.id, enabled })
+            }
+          />
         ))}
       </Stack>
 
@@ -180,3 +150,127 @@ export default function AiMcpServers() {
     </Paper>
   );
 }
+
+interface AiMcpServerRowProps {
+  server: IAiMcpServer;
+  onEdit: (server: IAiMcpServer) => void;
+  onDelete: (server: IAiMcpServer) => void;
+  onToggleEnabled: (enabled: boolean) => void;
+}
+
+/**
+ * A single external MCP server row: name/badge/url on the left and the
+ * Test / Switch / Edit / Delete controls on the right. Each row owns its own
+ * `useTestAiMcpServerMutation()` so the inline Test result and loading state are
+ * independent per row (a shared mutation would make `isPending` global and make
+ * every row flicker).
+ */
+function AiMcpServerRow({
+  server,
+  onEdit,
+  onDelete,
+  onToggleEnabled,
+}: AiMcpServerRowProps) {
+  const { t } = useTranslation();
+  const testMutation = useTestAiMcpServerMutation();
+  const result = testMutation.data;
+
+  // The row is keyed by `server.id`, so editing the connection-relevant fields
+  // (url/transport/headers) does NOT remount it — an old success/failure result
+  // would otherwise stick. Clear the result when those fields change.
+  useEffect(() => {
+    testMutation.reset();
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [server.url, server.transport, server.hasHeaders]);
+
+  // Single derivation of the button/tooltip presentation from the test tristate
+  // (idle / ok / failed), so the two can never drift apart. Tooltip is "" while
+  // there is no result; the icon is mapped from `view.state` below. When the
+  // request itself rejects (401/403/500/network) there is no `data` payload, so
+  // we feed the mutation error in too — otherwise the row would silently revert
+  // to "Test" instead of showing a red "Failed".
+  const view = mcpTestButtonView(
+    result,
+    t,
+    testMutation.isError ? testMutation.error : undefined,
+  );
+  const tooltipLabel = view.tooltip;
+  const buttonColor = view.color;
+  const buttonVariant = view.variant;
+  const buttonLabel = view.label;
+  const buttonIcon =
+    view.state === "ok" ? (
+      <IconCheck size={16} />
+    ) : view.state === "failed" ? (
+      <IconX size={16} />
+    ) : (
+      <IconPlugConnected size={16} />
+    );
+
+  return (
+    <Group justify="space-between" wrap="nowrap">
+      <Stack gap={2} style={{ minWidth: 0 }}>
+        <Group gap="xs">
+          <Text fw={500} truncate>
+            {server.name}
+          </Text>
+          <Badge size="xs" variant="light">
+            {server.transport.toUpperCase()}
+          </Badge>
+        </Group>
+        <Text
+          size="xs"
+          c="dimmed"
+          truncate
+          style={{ fontFamily: "ui-monospace, Menlo, monospace" }}
+        >
+          {server.url}
+        </Text>
+      </Stack>
+
+      <Group gap="xs" wrap="nowrap">
+        {/* Always clickable: testing a disabled server before enabling it is useful. */}
+        <Tooltip
+          label={tooltipLabel}
+          disabled={view.state === "idle"}
+          multiline
+          maw={320}
+          withinPortal
+        >
+          <Button
+            size="xs"
+            miw={88}
+            color={buttonColor}
+            variant={buttonVariant}
+            leftSection={testMutation.isPending ? undefined : buttonIcon}
+            loading={testMutation.isPending}
+            onClick={() => testMutation.mutate(server.id)}
+          >
+            {buttonLabel}
+          </Button>
+        </Tooltip>
+        <Switch
+          size="sm"
+          checked={server.enabled}
+          aria-label={t("Enabled")}
+          onChange={(event) => onToggleEnabled(event.currentTarget.checked)}
+        />
+        <ActionIcon
+          variant="subtle"
+          aria-label={t("Edit")}
+          onClick={() => onEdit(server)}
+        >
+          <IconPencil size={16} />
+        </ActionIcon>
+        <ActionIcon
+          variant="subtle"
+          color="red"
+          aria-label={t("Delete")}
+          onClick={() => onDelete(server)}
+        >
+          <IconTrash size={16} />
+        </ActionIcon>
+      </Group>
+    </Group>
+  );
+}

apps/client/src/features/workspace/components/settings/components/ai-provider-settings.tsx

@@ -7,6 +7,7 @@ import {
   Button,
   Group,
   Modal,
+  NumberInput,
   Paper,
   PasswordInput,
   Select,
@@ -83,6 +84,9 @@ const STT_LANGUAGE_OPTIONS: { value: string; label: string }[] = [
 // (empty means "leave unchanged" unless explicitly cleared).
 const formSchema = z.object({
   chatModel: z.string(),
+  // Max context window in tokens shown in the chat header badge. A number, or ""
+  // when the NumberInput is empty (no limit).
+  chatContextWindow: z.union([z.number(), z.literal("")]),
   // Chat provider implementation (reasoning surfacing). Default openai-compatible.
   chatApiStyle: z.enum(["openai-compatible", "openai"]),
   // Cheap model id for the anonymous public-share assistant; empty = use chatModel.
@@ -311,6 +315,7 @@ export default function AiProviderSettings() {
     validate: zod4Resolver(formSchema),
     initialValues: {
       chatModel: "",
+      chatContextWindow: "",
       chatApiStyle: "openai-compatible" as ChatApiStyle,
       publicShareChatModel: "",
       publicShareAssistantRoleId: "",
@@ -334,6 +339,7 @@ export default function AiProviderSettings() {
     if (!settings) return;
     form.setValues({
       chatModel: settings.chatModel ?? "",
+      chatContextWindow: settings.chatContextWindow ?? "",
       chatApiStyle: settings.chatApiStyle ?? "openai-compatible",
       publicShareChatModel: settings.publicShareChatModel ?? "",
       publicShareAssistantRoleId: settings.publicShareAssistantRoleId ?? "",
@@ -364,6 +370,12 @@ export default function AiProviderSettings() {
       // Everything is OpenAI-compatible.
       driver: "openai",
       chatModel: values.chatModel,
+      // Max context window for the chat header badge; empty NumberInput ("") →
+      // 0, which clears the limit server-side (no denominator shown).
+      chatContextWindow:
+        typeof values.chatContextWindow === "number"
+          ? values.chatContextWindow
+          : 0,
       chatApiStyle: values.chatApiStyle,
       // Cheap model id for the anonymous public-share assistant; empty falls
       // back to chatModel server-side.
@@ -767,6 +779,18 @@ export default function AiProviderSettings() {
           {t("Resolves to {{url}}", { url: chatResolved })}
         </Text>
 
+        <NumberInput
+          mt="sm"
+          label={t("Context window (tokens)")}
+          description={t(
+            "Shown as used / total in the chat header. Leave empty to hide the limit.",
+          )}
+          min={0}
+          allowDecimal={false}
+          disabled={isLoading}
+          {...form.getInputProps("chatContextWindow")}
+        />
+
         <Select
           mt="sm"
           label={t("Protocol")}

apps/client/src/features/workspace/components/settings/components/temporary-note-settings.tsx

@@ -0,0 +1,86 @@
+import { useState } from "react";
+import { useAtom } from "jotai";
+import {
+  Button,
+  Group,
+  NumberInput,
+  Paper,
+  Stack,
+  Text,
+} from "@mantine/core";
+import { notifications } from "@mantine/notifications";
+import { useTranslation } from "react-i18next";
+import useUserRole from "@/hooks/use-user-role.tsx";
+import { workspaceAtom } from "@/features/user/atoms/current-user-atom.ts";
+import { updateWorkspace } from "@/features/workspace/services/workspace-service.ts";
+import { IWorkspace } from "@/features/workspace/types/workspace.types.ts";
+
+// Mirrors DEFAULT_TEMPORARY_NOTE_HOURS on the server. Shown when the workspace
+// has no explicit value configured yet.
+const DEFAULT_TEMPORARY_NOTE_HOURS = 24;
+
+/**
+ * Workspace-level editor for the temporary-note lifetime, in HOURS. The deadline
+ * is frozen per-note at creation, so changing this only affects notes created
+ * afterwards. `temporaryNoteHours` is a top-level workspace column (like
+ * trashRetentionDays), not a nested setting.
+ */
+export default function TemporaryNoteSettings() {
+  const { t } = useTranslation();
+  const [workspace, setWorkspace] = useAtom(workspaceAtom);
+  const { isAdmin } = useUserRole();
+  const [isLoading, setIsLoading] = useState(false);
+  const [value, setValue] = useState<number>(
+    workspace?.temporaryNoteHours ?? DEFAULT_TEMPORARY_NOTE_HOURS,
+  );
+
+  async function handleSave() {
+    if (!value || value < 1) return;
+    setIsLoading(true);
+    try {
+      const updated = await updateWorkspace({
+        temporaryNoteHours: value,
+      } as Partial<IWorkspace>);
+      setWorkspace({ ...updated, temporaryNoteHours: value });
+      notifications.show({ message: t("Updated successfully") });
+    } catch (err) {
+      notifications.show({
+        message:
+          (err as any)?.response?.data?.message ?? t("Failed to update data"),
+        color: "red",
+      });
+    } finally {
+      setIsLoading(false);
+    }
+  }
+
+  return (
+    <Stack mt="sm">
+      <Text fw={700} size="lg">
+        {t("Temporary notes")}
+      </Text>
+
+      <Paper withBorder radius="md" p="lg">
+        <Text size="xs" c="dimmed" mb="xs">
+          {t(
+            "A temporary note is automatically moved to trash after this many hours unless it is made permanent. The deadline is fixed when the note is created.",
+          )}
+        </Text>
+        <NumberInput
+          label={t("Temporary note lifetime (hours)")}
+          min={1}
+          allowDecimal={false}
+          value={value}
+          onChange={(v) => setValue(typeof v === "number" ? v : Number(v))}
+          disabled={!isAdmin || isLoading}
+          w={220}
+        />
+        <Group justify="flex-end" mt="md">
+          <Button onClick={handleSave} loading={isLoading} disabled={!isAdmin}>
+            {t("Save")}
+          </Button>
+        </Group>
+      </Paper>
+    </Stack>
+  );
+}

apps/client/src/features/workspace/services/ai-settings-service.ts

@@ -22,6 +22,8 @@ export type ChatApiStyle = "openai-compatible" | "openai";
 export interface IAiSettings {
   driver?: AiDriver;
   chatModel?: string;
+  // Max context window in tokens shown in the chat header badge; 0/unset = no limit.
+  chatContextWindow?: number;
   chatApiStyle?: ChatApiStyle;
   // Cheap model id for the anonymous public-share assistant; empty = chatModel.
   publicShareChatModel?: string;
@@ -56,6 +58,8 @@ export interface IAiSettings {
 export interface IAiSettingsUpdate {
   driver?: AiDriver;
   chatModel?: string;
+  // Max context window in tokens for the chat header badge; 0 = clear the limit.
+  chatContextWindow?: number;
   chatApiStyle?: ChatApiStyle;
   publicShareChatModel?: string;
   // Agent-role id whose persona the public-share assistant adopts; empty =

apps/client/src/features/workspace/types/workspace.types.ts

@@ -28,6 +28,8 @@ export interface IWorkspace {
   aiDictationStreaming?: boolean;
   aiPublicShareAssistant?: boolean;
   trashRetentionDays?: number;
+  // Default lifetime (HOURS) for new temporary notes; frozen per-note at creation.
+  temporaryNoteHours?: number;
   restrictApiToAdmins?: boolean;
   allowMemberTemplates?: boolean;
   isScimEnabled?: boolean;

apps/client/src/pages/settings/workspace/workspace-settings.tsx

@@ -3,6 +3,7 @@ import WorkspaceNameForm from "@/features/workspace/components/settings/componen
 import WorkspaceIcon from "@/features/workspace/components/settings/components/workspace-icon.tsx";
 import HtmlEmbedSettings from "@/features/workspace/components/settings/components/html-embed-settings.tsx";
 import TrackerSettings from "@/features/workspace/components/settings/components/tracker-settings.tsx";
+import TemporaryNoteSettings from "@/features/workspace/components/settings/components/temporary-note-settings.tsx";
 import { useTranslation } from "react-i18next";
 import { getAppName } from "@/lib/config.ts";
 import { Helmet } from "react-helmet-async";
@@ -19,6 +20,7 @@ export default function WorkspaceSettings() {
       <WorkspaceNameForm />
       <HtmlEmbedSettings />
       <TrackerSettings />
+      <TemporaryNoteSettings />
     </>
   );
 }

apps/server/package.json

@@ -1,6 +1,6 @@
 {
   "name": "server",
-  "version": "0.93.0",
+  "version": "0.94.1",
   "description": "",
   "author": "",
   "private": true,

apps/server/src/collaboration/extensions/persistence-store.spec.ts

@@ -182,4 +182,46 @@ describe('PersistenceExtension.onStoreDocument — Approach-A boundary snapshot'
     expect(pageHistoryRepo.saveHistory).not.toHaveBeenCalled();
     expect(historyQueue.add).not.toHaveBeenCalled();
   });
+
+  // persist-1 — a transient DB failure during store must not silently lose the
+  // edit. hocuspocus unloads (destroys) the in-memory Y.Doc right after this
+  // hook resolves, so the store has to retry while it still holds the only copy.
+  it('retries a transient DB failure and still persists the edit (persist-1)', async () => {
+    const document = ydocFor(doc('NEW HUMAN CONTENT'));
+    pageRepo.findById.mockResolvedValue(persistedHumanPage('NEW HUMAN CONTENT'));
+    let attempts = 0;
+    pageRepo.updatePage.mockImplementation(async () => {
+      attempts += 1;
+      if (attempts === 1) throw new Error('deadlock detected'); // transient
+      callOrder.push('updatePage');
+    });
+
+    await ext.onStoreDocument(buildData(document, 'user') as any);
+
+    // First attempt failed and rolled back; the retry persisted the edit.
+    expect(pageRepo.updatePage).toHaveBeenCalledTimes(2);
+    // The edit WAS saved, so the post-store success path runs as normal.
+    expect((document as any).broadcastStateless).toHaveBeenCalledTimes(1);
+    expect(historyQueue.add).toHaveBeenCalledTimes(1);
+  });
+
+  // persist-1 — when every attempt fails the hook must NOT report a phantom
+  // success: no "page.updated" badge broadcast and no history snapshot for
+  // content that was never written.
+  it('does not run post-store side effects when every store attempt fails (persist-1)', async () => {
+    const document = ydocFor(doc('NEW HUMAN CONTENT'));
+    pageRepo.findById.mockResolvedValue(persistedHumanPage('NEW HUMAN CONTENT'));
+    pageRepo.updatePage.mockRejectedValue(new Error('connection reset'));
+
+    await expect(
+      ext.onStoreDocument(buildData(document, 'user') as any),
+    ).resolves.toBeUndefined();
+
+    // Bounded retry exhausted (MAX_STORE_ATTEMPTS).
+    expect(pageRepo.updatePage).toHaveBeenCalledTimes(3);
+    // No false-success: nothing downstream fires for the unsaved content.
+    expect((document as any).broadcastStateless).not.toHaveBeenCalled();
+    expect(historyQueue.add).not.toHaveBeenCalled();
+    expect(aiQueue.add).not.toHaveBeenCalled();
+  });
 });

apps/server/src/collaboration/extensions/persistence.extension.ts

@@ -181,83 +181,113 @@ export class PersistenceExtension implements Extension {
       context?.actor,
     );
 
-    try {
-      await executeTx(this.db, async (trx) => {
-        page = await this.pageRepo.findById(pageId, {
-          withLock: true,
-          includeContent: true,
-          trx,
-        });
+    // Persist with a small bounded retry. The in-memory Y.Doc is the ONLY copy
+    // of the latest edit until this hook returns: hocuspocus destroys/unloads the
+    // doc right after onStoreDocument resolves (see storeDocumentHooks' finally
+    // -> unloadDocument). If a transient DB error (deadlock, serialization
+    // failure, dropped connection) is merely logged and swallowed, the function
+    // resolves "successfully", the doc is unloaded, and the edit is lost silently
+    // (#206 persist-1). Retrying here re-attempts the write while we still hold
+    // the doc; on total failure we clear `page` so the post-store side effects
+    // (badge broadcast, history snapshot) never report a save that didn't happen.
+    const MAX_STORE_ATTEMPTS = 3;
+    for (let attempt = 1; attempt <= MAX_STORE_ATTEMPTS; attempt++) {
+      try {
+        await executeTx(this.db, async (trx) => {
+          page = await this.pageRepo.findById(pageId, {
+            withLock: true,
+            includeContent: true,
+            trx,
+          });
 
-        if (!page) {
-          this.logger.error(`Page with id ${pageId} not found`);
-          return;
-        }
-
-        if (isDeepStrictEqual(tiptapJson, page.content)) {
-          page = null;
-          return;
-        }
-
-        let contributorIds = undefined;
-        try {
-          const existingContributors = page.contributorIds || [];
-          contributorIds = Array.from(
-            new Set([
-              ...existingContributors,
-              ...editingUserIds,
-              page.creatorId,
-            ]),
-          );
-        } catch (err) {
-          //this.logger.debug('Contributors error:' + err?.['message']);
-        }
-
-        // Approach A — boundary snapshot before the agent's first edit.
-        // When this store is the agent's and the page's currently persisted
-        // state was authored by a human, pin that human state as its own
-        // history version BEFORE the agent overwrites it. `page` still holds the
-        // OLD content/provenance here, so saveHistory(page) captures the
-        // pre-agent state tagged 'user'. The agent's new content is snapshotted
-        // later by the debounced PAGE_HISTORY job ('agent'). Skip if the prior
-        // state is already agent-authored (boundary already pinned on the
-        // user->agent transition), if the page is effectively empty, or if the
-        // latest existing snapshot already equals this human state (avoid
-        // duplicates).
-        if (lastUpdatedSource === 'agent' && page.lastUpdatedSource !== 'agent') {
-          const lastHistory = await this.pageHistoryRepo.findPageLastHistory(
-            pageId,
-            { includeContent: true, trx },
-          );
-          const humanBaselineMissing =
-            !lastHistory || !isDeepStrictEqual(lastHistory.content, page.content);
-          if (!isEmptyParagraphDoc(page.content as any) && humanBaselineMissing) {
-            await this.pageHistoryRepo.saveHistory(page, {
-              contributorIds: page.contributorIds ?? undefined,
-              trx,
-            });
+          if (!page) {
+            this.logger.error(`Page with id ${pageId} not found`);
+            return;
           }
-        }
 
-        await this.pageRepo.updatePage(
-          {
-            content: tiptapJson,
-            textContent: textContent,
-            ydoc: ydocState,
-            lastUpdatedById: context.user.id,
-            // Human stays the responsible author; these annotate the source.
-            lastUpdatedSource,
-            lastUpdatedAiChatId: context?.aiChatId ?? null,
-            contributorIds: contributorIds,
-          },
-          pageId,
-          trx,
+          if (isDeepStrictEqual(tiptapJson, page.content)) {
+            page = null;
+            return;
+          }
+
+          let contributorIds = undefined;
+          try {
+            const existingContributors = page.contributorIds || [];
+            contributorIds = Array.from(
+              new Set([
+                ...existingContributors,
+                ...editingUserIds,
+                page.creatorId,
+              ]),
+            );
+          } catch (err) {
+            //this.logger.debug('Contributors error:' + err?.['message']);
+          }
+
+          // Approach A — boundary snapshot before the agent's first edit.
+          // When this store is the agent's and the page's currently persisted
+          // state was authored by a human, pin that human state as its own
+          // history version BEFORE the agent overwrites it. `page` still holds
+          // the OLD content/provenance here, so saveHistory(page) captures the
+          // pre-agent state tagged 'user'. The agent's new content is
+          // snapshotted later by the debounced PAGE_HISTORY job ('agent'). Skip
+          // if the prior state is already agent-authored (boundary already
+          // pinned on the user->agent transition), if the page is effectively
+          // empty, or if the latest existing snapshot already equals this human
+          // state (avoid duplicates).
+          if (
+            lastUpdatedSource === 'agent' &&
+            page.lastUpdatedSource !== 'agent'
+          ) {
+            const lastHistory = await this.pageHistoryRepo.findPageLastHistory(
+              pageId,
+              { includeContent: true, trx },
+            );
+            const humanBaselineMissing =
+              !lastHistory ||
+              !isDeepStrictEqual(lastHistory.content, page.content);
+            if (
+              !isEmptyParagraphDoc(page.content as any) &&
+              humanBaselineMissing
+            ) {
+              await this.pageHistoryRepo.saveHistory(page, {
+                contributorIds: page.contributorIds ?? undefined,
+                trx,
+              });
+            }
+          }
+
+          await this.pageRepo.updatePage(
+            {
+              content: tiptapJson,
+              textContent: textContent,
+              ydoc: ydocState,
+              lastUpdatedById: context.user.id,
+              // Human stays the responsible author; these annotate the source.
+              lastUpdatedSource,
+              lastUpdatedAiChatId: context?.aiChatId ?? null,
+              contributorIds: contributorIds,
+            },
+            pageId,
+            trx,
+          );
+
+          this.logger.debug(`Page updated: ${pageId} - SlugId: ${page.slugId}`);
+        });
+        break;
+      } catch (err) {
+        this.logger.error(
+          `Failed to update page ${pageId} (attempt ${attempt}/${MAX_STORE_ATTEMPTS})`,
+          err,
         );
-
-        this.logger.debug(`Page updated: ${pageId} - SlugId: ${page.slugId}`);
-      });
-    } catch (err) {
-      this.logger.error(`Failed to update page ${pageId}`, err);
+        // The write failed and rolled back; clear the partially-assigned `page`
+        // so the post-store success branch below is skipped (no false "saved"
+        // broadcast / history snapshot for content that was never persisted).
+        page = null;
+        if (attempt < MAX_STORE_ATTEMPTS) {
+          await new Promise((resolve) => setTimeout(resolve, attempt * 50));
+        }
+      }
     }
 
     if (page) {

apps/server/src/core/ai-chat/ai-chat.controller.export.spec.ts

@@ -0,0 +1,159 @@
+import { ForbiddenException } from '@nestjs/common';
+import { AiChatController } from './ai-chat.controller';
+import {
+  planFinalizeAssistant,
+  applyFinalize,
+  flushAssistant,
+  type AssistantFlush,
+} from './ai-chat.service';
+import type { User, Workspace } from '@docmost/db/types/entity.types';
+
+/**
+ * Wiring spec for the #183 `POST /ai-chat/export` endpoint. It must: own-gate via
+ * the chat lookup (workspace-scoped + creator-owned), load the FULL transcript
+ * via findAllByChat, render server-side, and return `{ markdown }`. Exercised by
+ * instantiating the controller with hand-rolled mocks — no Nest graph, no DB.
+ */
+describe('AiChatController.export', () => {
+  const user = { id: 'u1' } as User;
+  const workspace = { id: 'ws1' } as Workspace;
+
+  function makeController(
+    over: {
+      chat?: unknown;
+      rows?: unknown[];
+    } = {},
+  ) {
+    const chat =
+      'chat' in over
+        ? over.chat
+        : { id: 'c1', creatorId: 'u1', title: 'My chat' };
+    const aiChatRepo = {
+      findById: jest.fn().mockResolvedValue(chat),
+    };
+    const aiChatMessageRepo = {
+      findAllByChat: jest.fn().mockResolvedValue(
+        over.rows ?? [
+          {
+            id: 'm1',
+            role: 'user',
+            content: 'hi',
+            metadata: null,
+            status: null,
+          },
+          {
+            id: 'm2',
+            role: 'assistant',
+            content: 'hello',
+            metadata: null,
+            status: 'completed',
+          },
+        ],
+      ),
+    };
+    const controller = new AiChatController(
+      {} as never,
+      aiChatRepo as never,
+      aiChatMessageRepo as never,
+      {} as never,
+    );
+    return { controller, aiChatRepo, aiChatMessageRepo };
+  }
+
+  it('renders the full transcript and returns { markdown }', async () => {
+    const { controller, aiChatMessageRepo } = makeController();
+    const res = await controller.export({ chatId: 'c1' }, user, workspace);
+    expect(aiChatMessageRepo.findAllByChat).toHaveBeenCalledWith('c1', 'ws1');
+    expect(res.markdown).toContain('# My chat');
+    expect(res.markdown).toContain('## 1. You');
+    expect(res.markdown).toContain('## 2. AI agent');
+  });
+
+  it('forbids a chat the user does not own', async () => {
+    const { controller } = makeController({
+      chat: { id: 'c1', creatorId: 'someone-else', title: 'X' },
+    });
+    await expect(
+      controller.export({ chatId: 'c1' }, user, workspace),
+    ).rejects.toBeInstanceOf(ForbiddenException);
+  });
+
+  it('forbids a missing / foreign-workspace chat', async () => {
+    const { controller } = makeController({ chat: null });
+    await expect(
+      controller.export({ chatId: 'c1' }, user, workspace),
+    ).rejects.toBeInstanceOf(ForbiddenException);
+  });
+
+  it('localizes labels when lang=ru is passed', async () => {
+    const { controller } = makeController();
+    const res = await controller.export(
+      { chatId: 'c1', lang: 'ru' },
+      user,
+      workspace,
+    );
+    expect(res.markdown).toContain('## 1. Вы');
+    expect(res.markdown).toContain('## 2. ИИ-агент');
+  });
+});
+
+/**
+ * The terminal-finalize dispatch (#183): the assistant row is INSERTed upfront
+ * as 'streaming' and finalized once on the terminal callback. When the upfront
+ * insert SUCCEEDED (we hold an id) finalize UPDATEs that row; when it FAILED
+ * (assistantId is undefined) finalize falls back to INSERTing the terminal row
+ * so the turn is not lost — the only safety against losing the turn entirely.
+ *
+ * `planFinalizeAssistant` is the pure decision; `applyFinalize` is the REAL
+ * dispatch the service uses, exercised here over a mock repo (not a copy of the
+ * logic) so a production drift would fail the test (#186 review).
+ */
+describe('finalizeAssistant dispatch (planFinalizeAssistant + applyFinalize)', () => {
+  const workspaceId = 'ws1';
+
+  // Drive the SAME applyFinalize the service calls (no duplicated logic).
+  async function dispatchFinalize(
+    repo: { insert: jest.Mock; update: jest.Mock },
+    assistantId: string | undefined,
+    flushed: AssistantFlush,
+  ): Promise<void> {
+    await applyFinalize(
+      repo,
+      planFinalizeAssistant(assistantId),
+      { chatId: 'c1', workspaceId, userId: 'u1' },
+      flushed,
+    );
+  }
+
+  it('plan: update when the upfront insert returned an id', () => {
+    expect(planFinalizeAssistant('a1')).toEqual({ kind: 'update', id: 'a1' });
+  });
+
+  it('plan: insert (fallback) when there is no upfront id', () => {
+    expect(planFinalizeAssistant(undefined)).toEqual({ kind: 'insert' });
+  });
+
+  it('(a) upfront insert succeeded -> finalize UPDATEs the row by id', async () => {
+    const repo = { insert: jest.fn(), update: jest.fn() };
+    const flushed = flushAssistant([], 'final answer', 'completed', {
+      finishReason: 'stop',
+    });
+    await dispatchFinalize(repo, 'a1', flushed);
+    expect(repo.update).toHaveBeenCalledWith('a1', workspaceId, flushed);
+    expect(repo.insert).not.toHaveBeenCalled();
+  });
+
+  it('(b) upfront insert failed -> finalize INSERTs the terminal payload', async () => {
+    const repo = { insert: jest.fn(), update: jest.fn() };
+    const flushed = flushAssistant([], 'partial', 'error', { error: 'boom' });
+    await dispatchFinalize(repo, undefined, flushed);
+    expect(repo.update).not.toHaveBeenCalled();
+    expect(repo.insert).toHaveBeenCalledTimes(1);
+    const arg = repo.insert.mock.calls[0][0];
+    // The fallback insert carries the terminal content/status/metadata.
+    expect(arg.role).toBe('assistant');
+    expect(arg.content).toBe('partial');
+    expect(arg.status).toBe('error');
+    expect((arg.metadata as { error?: string }).error).toBe('boom');
+  });
+});

apps/server/src/core/ai-chat/ai-chat.controller.ts

@@ -20,7 +20,7 @@ import { JwtAuthGuard } from '../../common/guards/jwt-auth.guard';
 import { AuthUser } from '../../common/decorators/auth-user.decorator';
 import { AuthWorkspace } from '../../common/decorators/auth-workspace.decorator';
 import { SkipTransform } from '../../common/decorators/skip-transform.decorator';
-import { User, Workspace } from '@docmost/db/types/entity.types';
+import { AiChat, User, Workspace } from '@docmost/db/types/entity.types';
 import { PaginationOptions } from '@docmost/db/pagination/pagination-options';
 import { AiChatRepo } from '@docmost/db/repos/ai-chat/ai-chat.repo';
 import { AiChatMessageRepo } from '@docmost/db/repos/ai-chat/ai-chat-message.repo';
@@ -31,10 +31,12 @@ import { AiChatService, AiChatStreamBody } from './ai-chat.service';
 import { AiTranscriptionService } from './ai-transcription.service';
 import {
   ChatIdDto,
+  ExportChatDto,
   GetChatMessagesDto,
   RenameChatDto,
 } from './dto/ai-chat.dto';
 import { describeProviderError } from '../../integrations/ai/ai-error.util';
+import { buildChatMarkdown } from './chat-markdown.util';
 
 /**
  * Per-user AI chat API (§6.1). Routes are POST to match this codebase's
@@ -81,6 +83,36 @@ export class AiChatController {
     );
   }
 
+  /**
+   * Export a chat to Markdown (#183). The DB is the single source of truth: the
+   * whole transcript is loaded (oldest -> newest) and rendered server-side. Now
+   * that the assistant row is persisted upfront and per step, an interrupted
+   * turn is included up to its last finished step. Workspace-scoped and owner-
+   * gated via assertOwnedChat (same as the other read endpoints). Returns
+   * `{ markdown }`. `lang` localizes the few fixed labels (default English).
+   */
+  @HttpCode(HttpStatus.OK)
+  @Post('export')
+  async export(
+    @Body() dto: ExportChatDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ): Promise<{ markdown: string }> {
+    const chat = await this.assertOwnedChat(dto.chatId, user, workspace);
+    const rows = await this.aiChatMessageRepo.findAllByChat(
+      dto.chatId,
+      workspace.id,
+    );
+    const markdown = buildChatMarkdown({
+      title: chat.title ?? null,
+      chatId: dto.chatId,
+      rows,
+      // normalizeLang(undefined) already yields 'en', so no `?? 'en'` is needed.
+      lang: dto.lang,
+    });
+    return { markdown };
+  }
+
   /** Rename a chat. */
   @HttpCode(HttpStatus.OK)
   @Post('rename')
@@ -90,7 +122,11 @@ export class AiChatController {
     @AuthWorkspace() workspace: Workspace,
   ) {
     await this.assertOwnedChat(dto.chatId, user, workspace);
-    await this.aiChatRepo.update(dto.chatId, { title: dto.title }, workspace.id);
+    await this.aiChatRepo.update(
+      dto.chatId,
+      { title: dto.title },
+      workspace.id,
+    );
     return { success: true };
   }
 
@@ -145,7 +181,10 @@ export class AiChatController {
     // Resolve the agent role for this turn BEFORE hijack: existing chats read it
     // from ai_chats.role_id (authoritative), a new chat from body.roleId. The
     // role drives both the persona and the optional model override below.
-    const role = await this.aiChatService.resolveRoleForRequest(workspace, body);
+    const role = await this.aiChatService.resolveRoleForRequest(
+      workspace,
+      body,
+    );
 
     // Resolve the model (applying the role's optional override) BEFORE hijack so
     // an unconfigured provider — including a role pointing at an unconfigured
@@ -232,7 +271,9 @@ export class AiChatController {
     let file = null;
     try {
       // Whisper hard-caps uploads at 25MB; allow a single file.
-      file = await req.file({ limits: { fileSize: 25 * 1024 * 1024, files: 1 } });
+      file = await req.file({
+        limits: { fileSize: 25 * 1024 * 1024, files: 1 },
+      });
     } catch (err: any) {
       if (err?.statusCode === 413) {
         throw new BadRequestException('Audio file too large (max 25MB)');
@@ -283,11 +324,12 @@ export class AiChatController {
     chatId: string,
     user: User,
     workspace: Workspace,
-  ): Promise<void> {
+  ): Promise<AiChat> {
     const chat = await this.aiChatRepo.findById(chatId, workspace.id);
     if (!chat || chat.creatorId !== user.id) {
       throw new ForbiddenException();
     }
+    return chat;
   }
 }
 

apps/server/src/core/ai-chat/ai-chat.service.lifecycle.spec.ts

@@ -0,0 +1,61 @@
+import { Logger } from '@nestjs/common';
+import { AiChatService } from './ai-chat.service';
+
+/**
+ * Lifecycle unit tests for AiChatService.onModuleInit (#183 crash-recovery
+ * sweep). The sweep is BEST-EFFORT: a failure must be logged (warn) but must
+ * NEVER throw out of onModuleInit and block server startup. Exercised with a
+ * hand-rolled mock repo — no Nest graph, no DB. Only `aiChatMessageRepo` is
+ * touched by onModuleInit, so the other constructor deps are stubbed as never.
+ */
+describe('AiChatService.onModuleInit (startup sweep)', () => {
+  function makeService(sweepStreaming: jest.Mock) {
+    const aiChatMessageRepo = { sweepStreaming };
+    const service = new AiChatService(
+      {} as never, // ai
+      {} as never, // aiChatRepo
+      aiChatMessageRepo as never,
+      {} as never, // aiSettings
+      {} as never, // tools
+      {} as never, // mcpClients
+      {} as never, // aiAgentRoleRepo
+      {} as never, // pageRepo
+      {} as never, // pageAccess
+    );
+    return { service, aiChatMessageRepo };
+  }
+
+  afterEach(() => jest.restoreAllMocks());
+
+  it('happy path: calls sweepStreaming and resolves', async () => {
+    const sweepStreaming = jest.fn().mockResolvedValue(0);
+    const { service } = makeService(sweepStreaming);
+    await expect(service.onModuleInit()).resolves.toBeUndefined();
+    expect(sweepStreaming).toHaveBeenCalledTimes(1);
+  });
+
+  it('logs how many rows were swept when > 0', async () => {
+    const sweepStreaming = jest.fn().mockResolvedValue(3);
+    const logSpy = jest
+      .spyOn(Logger.prototype, 'log')
+      .mockImplementation(() => undefined);
+    const { service } = makeService(sweepStreaming);
+    await service.onModuleInit();
+    expect(logSpy).toHaveBeenCalledTimes(1);
+    expect(String(logSpy.mock.calls[0][0])).toContain('3');
+  });
+
+  it('sweepStreaming throws -> onModuleInit resolves (does NOT throw) and warns', async () => {
+    const sweepStreaming = jest
+      .fn()
+      .mockRejectedValue(new Error('db unavailable'));
+    const warnSpy = jest
+      .spyOn(Logger.prototype, 'warn')
+      .mockImplementation(() => undefined);
+    const { service } = makeService(sweepStreaming);
+    // Must not throw — a sweep failure may never block startup.
+    await expect(service.onModuleInit()).resolves.toBeUndefined();
+    expect(warnSpy).toHaveBeenCalledTimes(1);
+    expect(String(warnSpy.mock.calls[0][0])).toContain('db unavailable');
+  });
+});

apps/server/src/core/ai-chat/ai-chat.service.spec.ts

@@ -6,7 +6,7 @@ import {
   serializeSteps,
   rowToUiMessage,
   prepareAgentStep,
-  buildPartialAssistantRecord,
+  flushAssistant,
   chatStreamMetadata,
   accumulateStepUsage,
   MAX_AGENT_STEPS,
@@ -233,101 +233,126 @@ describe('prepareAgentStep', () => {
     // The synthesis instruction is appended.
     expect(result?.system).toContain(FINAL_STEP_INSTRUCTION);
   });
-
-  it('pins the off-by-one boundary (MAX-2 is not final, MAX-1 is)', () => {
-    // Boundary expressed via the constant, not a hardcoded 18/19, so the test
-    // tracks MAX_AGENT_STEPS if the cap ever changes.
-    expect(prepareAgentStep(MAX_AGENT_STEPS - 2, 'SYS')).toBeUndefined();
-    const atBoundary = prepareAgentStep(MAX_AGENT_STEPS - 1, 'SYS');
-    expect(atBoundary).toBeDefined();
-    expect(atBoundary?.toolChoice).toBe('none');
-  });
 });
 
 /**
- * Unit test for buildPartialAssistantRecord: the pure helper that shapes the
- * assistant-message record persisted on a partial/failed turn (the streamText
- * onError / onAbort paths). It captures the PARTIAL answer the user already saw
- * (finished steps' text + tool parts, plus the in-progress step's text) so a
- * provider error / disconnect no longer throws the streamed answer away. Pinning
- * the record shape here covers the persist-partial logic without seaming
- * streamText itself.
+ * flushAssistant (#183): the PURE row builder behind the step-granular durable
+ * write path. It runs identically for the upfront insert (empty steps,
+ * 'streaming'), every per-step update, and the terminal finalize — so a future
+ * background worker can call the same function. These tests pin the four status
+ * shapes and the `metadata.parts` shape that rowToUiMessage/findRecent depend on
+ * (per-step text + tool parts via assistantParts, in-progress text appended).
  */
-describe('buildPartialAssistantRecord', () => {
+describe('flushAssistant', () => {
   type AnyPart = Record<string, unknown>;
 
-  it('records an empty turn with the error text (preserves old behavior)', () => {
-    const rec = buildPartialAssistantRecord(
-      [],
-      '',
-      'error',
-      '401: Unauthorized',
-    );
-    expect(rec).toEqual({
-      text: '',
-      toolCalls: null,
-      metadata: {
-        finishReason: 'error',
-        parts: [],
-        error: '401: Unauthorized',
-      },
-    });
+  const toolStep = {
+    text: 'looked it up',
+    toolCalls: [{ toolCallId: 'c1', toolName: 'getPage', input: { id: 'p1' } }],
+    toolResults: [
+      { toolCallId: 'c1', toolName: 'getPage', output: { title: 'T' } },
+    ],
+  };
+
+  it('upfront seed: empty streaming row (no content, no toolCalls, empty parts)', () => {
+    const f = flushAssistant([], '', 'streaming');
+    expect(f.status).toBe('streaming');
+    expect(f.content).toBe('');
+    expect(f.toolCalls).toBeNull();
+    expect(f.metadata.parts).toEqual([]);
+    // No finishReason while streaming (it is not a terminal state).
+    expect('finishReason' in f.metadata).toBe(false);
   });
 
-  it('persists in-progress text (no finished steps) as the partial answer', () => {
-    const rec = buildPartialAssistantRecord(
-      [],
-      'partial answer',
-      'error',
-      'boom',
-    );
-    expect(rec.text).toBe('partial answer');
-    expect(rec.metadata.parts).toEqual([
+  it('streaming update folds in finished steps but keeps status streaming', () => {
+    const f = flushAssistant([toolStep], '', 'streaming');
+    expect(f.status).toBe('streaming');
+    expect(f.content).toBe('looked it up');
+    const parts = f.metadata.parts as AnyPart[];
+    expect(parts).toContainEqual({ type: 'text', text: 'looked it up' });
+    const toolPart = parts.find((p) => p.type === 'tool-getPage');
+    expect(toolPart!.state).toBe('output-available');
+    expect(f.toolCalls).not.toBeNull();
+  });
+
+  it('completed: attaches finishReason + normalized usage + contextTokens + maxContextTokens', () => {
+    const f = flushAssistant([toolStep], '', 'completed', {
+      finishReason: 'stop',
+      usage: { inputTokens: 10, outputTokens: 5, totalTokens: 15 },
+      contextTokens: 15,
+      maxContextTokens: 200000,
+    });
+    expect(f.status).toBe('completed');
+    expect(f.metadata.finishReason).toBe('stop');
+    expect(f.metadata.usage).toEqual({
+      inputTokens: 10,
+      outputTokens: 5,
+      totalTokens: 15,
+      reasoningTokens: undefined,
+    });
+    expect(f.metadata.contextTokens).toBe(15);
+    expect(f.metadata.maxContextTokens).toBe(200000);
+  });
+
+  it('completed: omits maxContextTokens when unset or 0', () => {
+    // No maxContextTokens in the extra (admin set no context window).
+    const f = flushAssistant([toolStep], '', 'completed', {
+      finishReason: 'stop',
+      contextTokens: 15,
+    });
+    expect('maxContextTokens' in f.metadata).toBe(false);
+    // Explicit 0 is treated the same as unset (no limit -> key omitted).
+    const f0 = flushAssistant([toolStep], '', 'completed', {
+      finishReason: 'stop',
+      contextTokens: 15,
+      maxContextTokens: 0,
+    });
+    expect('maxContextTokens' in f0.metadata).toBe(false);
+  });
+
+  it('error: records the error and a derived finishReason', () => {
+    const f = flushAssistant([], 'partial answer', 'error', { error: 'boom' });
+    expect(f.status).toBe('error');
+    expect(f.content).toBe('partial answer');
+    expect(f.metadata.error).toBe('boom');
+    // Derives finishReason from the terminal status when none is supplied.
+    expect(f.metadata.finishReason).toBe('error');
+    expect(f.metadata.parts).toEqual([
       { type: 'text', text: 'partial answer' },
     ]);
-    expect(rec.metadata.error).toBe('boom');
   });
 
-  it('combines a finished tool step with trailing in-progress text', () => {
-    const steps = [
-      {
-        text: 'looked it up',
-        toolCalls: [
-          { toolCallId: 'c1', toolName: 'getPage', input: { id: 'p1' } },
-        ],
-        toolResults: [
-          { toolCallId: 'c1', toolName: 'getPage', output: { title: 'T' } },
-        ],
-      },
-    ];
-    const rec = buildPartialAssistantRecord(
-      steps,
-      ' and then',
-      'error',
-      'boom',
-    );
-    const parts = rec.metadata.parts as AnyPart[];
-    // The finished step's text part is present.
-    expect(parts).toContainEqual({ type: 'text', text: 'looked it up' });
-    // The paired tool call+result becomes an output-available part.
-    const toolPart = parts.find((p) => p.type === 'tool-getPage');
-    expect(toolPart).toBeDefined();
-    expect(toolPart!.state).toBe('output-available');
-    // The in-progress text is appended LAST so the parts match the stream order.
+  it('aborted: in-progress text appended last, no error key', () => {
+    const f = flushAssistant([toolStep], ' and then', 'aborted');
+    expect(f.status).toBe('aborted');
+    expect(f.metadata.finishReason).toBe('aborted');
+    expect('error' in f.metadata).toBe(false);
+    expect(f.content).toBe('looked it up and then');
+    const parts = f.metadata.parts as AnyPart[];
     expect(parts[parts.length - 1]).toEqual({
       type: 'text',
       text: ' and then',
     });
-    expect(rec.text).toBe('looked it up and then');
-    expect(rec.toolCalls).not.toBeNull();
-    expect(rec.metadata.error).toBe('boom');
   });
 
-  it('omits the error key on the abort path (no errorText)', () => {
-    const rec = buildPartialAssistantRecord([], 'half', 'aborted');
-    expect(rec.metadata.finishReason).toBe('aborted');
-    expect('error' in rec.metadata).toBe(false);
-    expect(rec.text).toBe('half');
+  it('combines a finished tool step with trailing in-progress text (error path)', () => {
+    // The error path captures the PARTIAL answer the user already saw: each
+    // finished step's text + tool parts, then the in-progress step's text last.
+    const flushed = flushAssistant([toolStep], ' and then', 'error', {
+      error: 'boom',
+    });
+    const parts = flushed.metadata.parts as AnyPart[];
+    expect(parts).toContainEqual({ type: 'text', text: 'looked it up' });
+    const toolPart = parts.find((p) => p.type === 'tool-getPage');
+    expect(toolPart!.state).toBe('output-available');
+    // In-progress text appended LAST so the parts match the stream order.
+    expect(parts[parts.length - 1]).toEqual({
+      type: 'text',
+      text: ' and then',
+    });
+    expect(flushed.content).toBe('looked it up and then');
+    expect(flushed.toolCalls).not.toBeNull();
+    expect(flushed.metadata.error).toBe('boom');
   });
 });
 

apps/server/src/core/ai-chat/ai-chat.service.ts

@@ -1,4 +1,9 @@
-import { ForbiddenException, Injectable, Logger } from '@nestjs/common';
+import {
+  ForbiddenException,
+  Injectable,
+  Logger,
+  OnModuleInit,
+} from '@nestjs/common';
 import { FastifyReply } from 'fastify';
 import {
   streamText,
@@ -124,7 +129,7 @@ export interface AiChatStreamArgs {
  *                    can be rebuilt for `convertToModelMessages`.
  */
 @Injectable()
-export class AiChatService {
+export class AiChatService implements OnModuleInit {
   private readonly logger = new Logger(AiChatService.name);
 
   constructor(
@@ -139,6 +144,32 @@ export class AiChatService {
     private readonly pageAccess: PageAccessService,
   ) {}
 
+  /**
+   * Crash-recovery sweep on server start (#183): any assistant row left in the
+   * 'streaming' state is the relic of a turn whose process died before it
+   * reached a terminal status. Flip those to 'aborted' so history/export show
+   * them settled (with whatever finished steps were already persisted) instead
+   * of perpetually "streaming". Best-effort: a sweep failure is logged but must
+   * never block server startup.
+   */
+  async onModuleInit(): Promise<void> {
+    try {
+      const swept = await this.aiChatMessageRepo.sweepStreaming();
+      if (swept > 0) {
+        this.logger.log(
+          `Startup sweep: marked ${swept} dangling 'streaming' assistant ` +
+            `message(s) as 'aborted'.`,
+        );
+      }
+    } catch (err) {
+      this.logger.warn(
+        `Startup sweep of dangling 'streaming' messages failed: ${
+          err instanceof Error ? err.message : 'unknown error'
+        }`,
+      );
+    }
+  }
+
   /**
    * Resolve the agent role that applies to this stream request, scoped to the
    * workspace and soft-delete aware. For an EXISTING chat the role is read from
@@ -395,31 +426,6 @@ export class AiChatService {
 
     const tools = { ...external.tools, ...docmostTools };
 
-    // Persist the assistant message. Used by onFinish (full result) and the
-    // abort/error paths (partial result). Guarded so we persist at most once.
-    let persisted = false;
-    const persistAssistant = async (data: {
-      text: string;
-      toolCalls: unknown;
-      metadata: Record<string, unknown>;
-    }): Promise<void> => {
-      if (persisted) return;
-      persisted = true;
-      try {
-        await this.aiChatMessageRepo.insert({
-          chatId,
-          workspaceId: workspace.id,
-          userId: user.id,
-          role: 'assistant',
-          content: data.text ?? '',
-          toolCalls: (data.toolCalls ?? null) as never,
-          metadata: data.metadata as never,
-        });
-      } catch (err) {
-        this.logger.error('Failed to persist assistant message', err as Error);
-      }
-    };
-
     // Accumulate the turn's streamed output so a provider error / disconnect can
     // persist the PARTIAL answer the user already saw — the SDK's onError/onAbort
     // callbacks don't hand us the in-progress text. `capturedSteps` holds finished
@@ -428,6 +434,101 @@ export class AiChatService {
     const capturedSteps: StepLike[] = [];
     let inProgressText = '';
 
+    // Step-granular durability (#183): create the assistant row UPFRONT in the
+    // 'streaming' state (before any token), then UPDATE it as each step finishes
+    // and finalize it once on the terminal callback. If the process dies
+    // mid-turn the row survives with every finished step already persisted; the
+    // startup sweep (sweepStreaming) later flips a dangling 'streaming' row to
+    // 'aborted'. The DB is now the single source of truth for the turn — the
+    // socket is never required for the write path. A failed upfront insert is
+    // logged and leaves assistantId undefined; the per-step/terminal updates then
+    // no-op (guarded below) so the turn still streams to the user.
+    let assistantId: string | undefined;
+    try {
+      const seed = flushAssistant([], '', 'streaming');
+      const seeded = await this.aiChatMessageRepo.insert({
+        chatId,
+        workspaceId: workspace.id,
+        userId: user.id,
+        role: 'assistant',
+        content: seed.content,
+        // jsonb columns: cast through never (same as the user insert above).
+        toolCalls: (seed.toolCalls ?? null) as never,
+        metadata: seed.metadata as never,
+        status: seed.status,
+      });
+      assistantId = seeded?.id;
+    } catch (err) {
+      this.logger.error(
+        `Failed to insert upfront assistant row (chat ${chatId}, workspace ${workspace.id})`,
+        err as Error,
+      );
+    }
+
+    // Per-step (non-terminal) update: persist the finished steps the moment a
+    // step ends. Tolerant — a failed update is logged and swallowed so it never
+    // throws into the stream. Keeps status 'streaming'.
+    const updateStreaming = async (): Promise<void> => {
+      if (!assistantId) return;
+      // Cheap short-circuit once the turn is finalized (see `finalized` below).
+      // The AUTHORITATIVE guard is `onlyIfStreaming` on the UPDATE: a late
+      // fire-and-forget step update could still be in flight on another pool
+      // connection when finalize runs, so the SQL `WHERE status='streaming'`
+      // (not this flag) is what prevents it clobbering the terminal row.
+      if (finalized) return;
+      try {
+        await this.aiChatMessageRepo.update(
+          assistantId,
+          workspace.id,
+          flushAssistant(capturedSteps, '', 'streaming'),
+          { onlyIfStreaming: true },
+        );
+      } catch (err) {
+        this.logger.warn(
+          `Failed to update streaming assistant row: ${
+            err instanceof Error ? err.message : 'unknown error'
+          }`,
+        );
+      }
+    };
+
+    // Serialize the per-step updates (#183 review): onStepFinish fires them
+    // without await, so two could otherwise commit out of order on different pool
+    // connections (step N landing after N+1). Chaining each onto the previous
+    // keeps the persisted row monotonic with step order; each link short-circuits
+    // on `finalized`, so a tail of late updates is cheap.
+    let stepUpdateChain: Promise<void> = Promise.resolve();
+
+    // Terminal finalize: write the completed/error/aborted row exactly once
+    // across the (mutually-exclusive, at-most-once) onFinish/onError/onAbort
+    // callbacks — mirroring the pre-#183 persist-at-most-once guard for the
+    // TERMINAL status (the row may be updated many times with 'streaming' before
+    // this fires once).
+    let finalized = false;
+    const finalizeAssistant = async (
+      flushed: AssistantFlush,
+    ): Promise<void> => {
+      if (finalized) return;
+      finalized = true;
+      const plan = planFinalizeAssistant(assistantId);
+      try {
+        // Shared dispatch (see applyFinalize): UPDATE the upfront row, or — when
+        // the upfront insert failed (kind 'insert') — INSERT the terminal row as
+        // the only safety against losing the turn entirely.
+        await applyFinalize(
+          this.aiChatMessageRepo,
+          plan,
+          { chatId, workspaceId: workspace.id, userId: user.id },
+          flushed,
+        );
+      } catch (err) {
+        this.logger.error(
+          `Failed to finalize assistant message (kind=${plan.kind})`,
+          err as Error,
+        );
+      }
+    };
+
     // DIAGNOSTIC (Safari stream-drop investigation) — temporary. Measure
     // first-chunk latency, the model-silent gap right before a disconnect, and
     // how many SSE heartbeats were written, so a Safari drop can be classified
@@ -476,6 +577,12 @@ export class AiChatService {
           // the in-progress accumulator for the next step.
           capturedSteps.push(step as StepLike);
           inProgressText = '';
+          // Step-granular durability (#183): persist this finished step (its text +
+          // tool calls + tool RESULTS) the moment it ends, so a process death after
+          // this point still recovers the step. Not awaited here (never block the
+          // stream), but SERIALIZED via stepUpdateChain so the writes commit in
+          // step order; updateStreaming is error-tolerant (logs + swallows).
+          stepUpdateChain = stepUpdateChain.then(() => updateStreaming());
         },
         onFinish: async ({ text, finishReason, totalUsage, usage, steps }) => {
           // DIAGNOSTIC (Safari stream-drop investigation) — temporary: success
@@ -486,30 +593,35 @@ export class AiChatService {
               `firstChunkLatency=${firstModelChunkAt ? firstModelChunkAt - streamStartedAt : 'none'}ms ` +
               `heartbeatsSent=${heartbeatsSent} steps=${steps.length}`,
           );
-          await persistAssistant({
-            text,
-            toolCalls: serializeSteps(steps),
-            metadata: {
-              finishReason,
-              // Persist the turn's cumulative usage WITH reasoning tokens resolved
-              // from either the new `outputTokenDetails` or the deprecated top-level
-              // field, so reopened history / the Markdown export show the thinking
-              // token cost too.
-              usage:
-                normalizeStreamUsage(totalUsage as StreamUsage) ?? totalUsage,
-              // Final-step usage = the context actually fed to the model on the last LLM
-              // call (full history + tool results) plus the answer it just generated.
-              // input+output of the FINAL step ≈ the conversation's CURRENT context size,
-              // distinct from totalUsage which sums every step (cumulative tokens spent).
+          // Finalize the assistant row (#183): the upfront 'streaming' row is
+          // UPDATEd to 'completed' with the turn's final text, cumulative usage and
+          // full UIMessage parts. We pass the SDK `steps` (which carry the final
+          // step's text) as the captured steps so metadata.parts matches the
+          // pre-#183 onFinish record exactly; `inProgressText` is '' here (the last
+          // step already finished). Final-step usage (usage.input+output) ≈ the
+          // conversation's CURRENT context size, distinct from totalUsage.
+          //
+          // COLUMN-SEMANTICS NOTE (#183): `content` is built by flushAssistant as
+          // the CONCATENATION of every step's text (stepsText), whereas pre-#183
+          // it stored only the FINAL step's text. This is a deliberate, harmless
+          // change: the UI and the Markdown export render from `metadata.parts`
+          // (per-step text + tool parts), not from `content`; `content` is the
+          // plain-text projection (full-text search / fallback). A multi-step
+          // turn's `content` therefore now holds all steps' prose, not just the
+          // last block.
+          await finalizeAssistant(
+            flushAssistant(steps as StepLike[], '', 'completed', {
+              finishReason: finishReason as string,
+              usage: totalUsage as StreamUsage,
               contextTokens:
                 (usage?.inputTokens ?? 0) + (usage?.outputTokens ?? 0) ||
                 undefined,
-              // Persist the FULL set of UIMessage parts for the turn (text +
-              // tool-call/result), so the rebuilt history replays prior tool
-              // context to the model on later turns.
-              parts: assistantParts(steps, text),
-            },
-          });
+              // Max context window for the chat header badge denominator;
+              // resolved from the admin-configured provider settings (in
+              // closure scope here). Omitted/0 = no limit.
+              maxContextTokens: resolved?.chatContextWindow,
+            }),
+          );
           // Lifecycle: release the external MCP clients leased for this turn.
           await closeExternalClients();
 
@@ -545,16 +657,14 @@ export class AiChatService {
               `firstChunkLatency=${firstModelChunkAt ? firstModelChunkAt - streamStartedAt : 'none'}ms ` +
               `silentGapBeforeDrop=${diagNow - lastModelChunkAt}ms heartbeatsSent=${heartbeatsSent}`,
           );
-          // Persist the PARTIAL answer streamed before the failure (text + any
+          // Finalize the PARTIAL answer streamed before the failure (text + any
           // finished tool steps) WITH the error in metadata, so the turn shows what
-          // the user already saw plus the cause — not just a bare error.
-          await persistAssistant(
-            buildPartialAssistantRecord(
-              capturedSteps,
-              inProgressText,
-              'error',
-              errorText,
-            ),
+          // the user already saw plus the cause — not just a bare error. Status
+          // 'error' (#183).
+          await finalizeAssistant(
+            flushAssistant(capturedSteps, inProgressText, 'error', {
+              error: errorText,
+            }),
           );
           await closeExternalClients();
         },
@@ -578,12 +688,8 @@ export class AiChatService {
               `silentGapBeforeDrop=${diagNow - lastModelChunkAt}ms heartbeatsSent=${heartbeatsSent} ` +
               `steps=${steps.length}`,
           );
-          await persistAssistant(
-            buildPartialAssistantRecord(
-              capturedSteps,
-              inProgressText,
-              'aborted',
-            ),
+          await finalizeAssistant(
+            flushAssistant(capturedSteps, inProgressText, 'aborted'),
           );
           await closeExternalClients();
         },
@@ -1032,38 +1138,136 @@ export function rowToUiMessage(row: AiChatMessage): Omit<UIMessage, 'id'> & {
 }
 
 /**
- * Build the assistant-message record persisted on a partial/failed turn (the
- * streamText onError / onAbort paths). Captures the partial answer the user
- * already saw: each finished step's text + tool parts (via assistantParts),
- * then the in-progress step's text appended last. When `errorText` is provided
- * it is recorded in metadata.error so the cause shows in history; an aborted
- * turn passes none. Pure, so the partial-recording shape is unit-testable
- * without seaming streamText.
+ * The persisted-row patch shape produced by {@link flushAssistant}. It is the
+ * SAME shape the assistant repo insert/update consume (content + toolCalls +
+ * metadata) plus the lifecycle `status` column added in #183.
  */
-export function buildPartialAssistantRecord(
-  steps: ReadonlyArray<StepLike> | undefined,
+export interface AssistantFlush {
+  content: string;
+  toolCalls: unknown;
+  metadata: Record<string, unknown>;
+  status: 'streaming' | 'completed' | 'error' | 'aborted';
+}
+
+/**
+ * Pure decision for the terminal finalize (#183): given whether the upfront
+ * assistant row exists (`assistantId`), choose whether the terminal payload is
+ * written by UPDATEing that row or — when the upfront insert failed and there is
+ * no id — by INSERTing a fresh terminal row so the turn is not lost entirely.
+ * Returns `{ kind: 'update', id }` or `{ kind: 'insert' }`. Extracted so the
+ * fallback-insert branch (the only safety against losing a turn whose upfront
+ * insert failed) is unit-testable without seaming streamText.
+ */
+export function planFinalizeAssistant(
+  assistantId: string | undefined,
+): { kind: 'update'; id: string } | { kind: 'insert' } {
+  return assistantId ? { kind: 'update', id: assistantId } : { kind: 'insert' };
+}
+
+/** The repo surface the terminal finalize needs (structural — the real repo and
+ *  a test mock both satisfy it). */
+export interface FinalizeRepo {
+  insert(insertable: Record<string, unknown>): Promise<unknown>;
+  update(
+    id: string,
+    workspaceId: string,
+    patch: AssistantFlush,
+  ): Promise<unknown>;
+}
+
+/**
+ * Apply a finalize `plan` to the repo with the terminal `flushed` payload (#183):
+ * UPDATE the upfront row, or INSERT a fresh terminal row as the fallback when the
+ * upfront insert failed. The SINGLE dispatch shared by the service's
+ * finalizeAssistant and its test, so the test exercises the real path instead of
+ * a copy (#186 review). Pure of error handling — the caller wraps it.
+ */
+export async function applyFinalize(
+  repo: FinalizeRepo,
+  plan: { kind: 'update'; id: string } | { kind: 'insert' },
+  base: { chatId: string; workspaceId: string; userId: string },
+  flushed: AssistantFlush,
+): Promise<void> {
+  if (plan.kind === 'update') {
+    await repo.update(plan.id, base.workspaceId, flushed);
+    return;
+  }
+  await repo.insert({
+    chatId: base.chatId,
+    workspaceId: base.workspaceId,
+    userId: base.userId,
+    role: 'assistant',
+    content: flushed.content,
+    toolCalls: flushed.toolCalls ?? null,
+    metadata: flushed.metadata,
+    status: flushed.status,
+  });
+}
+
+/**
+ * PURE assistant-row builder (#183 step-granular durability). Given the turn's
+ * accumulated steps + the in-progress (not-yet-finished) text + the lifecycle
+ * status, it returns the row patch to persist. The SAME path runs for the
+ * upfront insert (empty steps, status 'streaming'), every per-step update, and
+ * the terminal finalize (completed/error/aborted) — and a future background
+ * worker can call it identically, so it must stay a pure function of its inputs
+ * (NO `this`, no IO).
+ *
+ * `metadata.parts` is built by assistantParts over the finished steps, then the
+ * in-progress text appended as a trailing text part, so rowToUiMessage /
+ * findRecent keep replaying the turn unchanged. `metadata.finishReason`,
+ * `metadata.error`, `metadata.usage`, `metadata.contextTokens` and
+ * `metadata.maxContextTokens` are attached only when provided/relevant, matching
+ * the pre-#183 onFinish/onError records.
+ */
+export function flushAssistant(
+  capturedSteps: ReadonlyArray<StepLike> | undefined,
   inProgressText: string,
-  finishReason: 'error' | 'aborted',
-  errorText?: string,
-): { text: string; toolCalls: unknown; metadata: Record<string, unknown> } {
-  const finished = steps ?? [];
+  status: 'streaming' | 'completed' | 'error' | 'aborted',
+  extra?: {
+    finishReason?: string;
+    usage?: ChatStreamUsage | StreamUsage | undefined;
+    contextTokens?: number;
+    maxContextTokens?: number;
+    error?: string;
+  },
+): AssistantFlush {
+  const finished = capturedSteps ?? [];
   const stepsText = finished.map((s) => s.text ?? '').join('');
   const trailing = inProgressText ?? '';
   // assistantParts emits text parts only for FINISHED steps; append the
-  // in-progress step's text (the answer cut off by the error) as the last text
-  // part so the persisted parts match what streamed to the client.
+  // in-progress step's text (the partial answer cut off by an error/abort, or
+  // simply not yet flushed mid-stream) as the last text part so the persisted
+  // parts match what streamed to the client.
   const parts = assistantParts(finished, '') as unknown as Array<
     Record<string, unknown>
   >;
   if (trailing) parts.push({ type: 'text', text: trailing });
+
+  const metadata: Record<string, unknown> = {
+    parts: parts as unknown as UIMessage['parts'],
+  };
+  // finishReason: prefer an explicit one; else derive a sensible value from the
+  // terminal status (so onError/onAbort records keep their historical reason).
+  if (extra?.finishReason) {
+    metadata.finishReason = extra.finishReason;
+  } else if (status === 'error' || status === 'aborted') {
+    metadata.finishReason = status;
+  }
+  if (extra?.usage !== undefined) {
+    metadata.usage =
+      normalizeStreamUsage(extra.usage as StreamUsage) ?? extra.usage;
+  }
+  if (extra?.contextTokens) metadata.contextTokens = extra.contextTokens;
+  if (extra?.maxContextTokens)
+    metadata.maxContextTokens = extra.maxContextTokens;
+  if (extra?.error) metadata.error = extra.error;
+
   return {
-    text: stepsText + trailing,
+    content: stepsText + trailing,
     toolCalls: serializeSteps(finished),
-    metadata: {
-      finishReason,
-      parts: parts as unknown as UIMessage['parts'],
-      ...(errorText ? { error: errorText } : {}),
-    },
+    metadata,
+    status,
   };
 }
 

apps/server/src/core/ai-chat/chat-markdown.util.spec.ts

@@ -0,0 +1,295 @@
+import { buildChatMarkdown, normalizeLang } from './chat-markdown.util';
+import type { AiChatMessage } from '@docmost/db/types/entity.types';
+
+/**
+ * normalizeLang: the client sends `i18n.language` — a FULL locale tag like
+ * 'en-US' / 'ru-RU', NOT a bare 'en'/'ru'. A `@IsIn(['en','ru'])` DTO rejected
+ * that with a 400 (caught in real-browser testing); the export now accepts any
+ * string and normalizes here. Guards that regression.
+ */
+describe('normalizeLang', () => {
+  it("maps any 'ru…' locale tag to ru", () => {
+    expect(normalizeLang('ru')).toBe('ru');
+    expect(normalizeLang('ru-RU')).toBe('ru');
+    expect(normalizeLang('RU-ru')).toBe('ru');
+  });
+
+  it('maps everything else (incl. region-qualified English) to en', () => {
+    expect(normalizeLang('en')).toBe('en');
+    expect(normalizeLang('en-US')).toBe('en');
+    expect(normalizeLang('fr-FR')).toBe('en');
+    expect(normalizeLang(undefined)).toBe('en');
+    expect(normalizeLang('')).toBe('en');
+  });
+});
+
+/**
+ * Unit tests for the SERVER Markdown export (#183). Mirrors the coverage of the
+ * (now-removed) client chat-markdown tests: heading/metadata, role labels, text
+ * + tool blocks, token footers, the interrupted-turn note, and NULL-status
+ * (legacy) rows. The export embeds a live `new Date().toISOString()` timestamp;
+ * we never assert it, only the deterministic structure.
+ */
+
+function row(partial: Partial<AiChatMessage>): AiChatMessage {
+  return {
+    id: partial.id ?? 'id',
+    chatId: partial.chatId ?? 'chat-1',
+    workspaceId: partial.workspaceId ?? 'ws-1',
+    userId: partial.userId ?? null,
+    role: partial.role ?? 'user',
+    content: partial.content ?? null,
+    toolCalls: partial.toolCalls ?? null,
+    metadata: partial.metadata ?? null,
+    status: partial.status ?? null,
+    createdAt: partial.createdAt ?? ('2026-06-21T00:00:00.000Z' as never),
+    updatedAt: partial.updatedAt ?? ('2026-06-21T00:00:00.000Z' as never),
+    deletedAt: partial.deletedAt ?? null,
+  } as AiChatMessage;
+}
+
+describe('buildChatMarkdown (server) — structure', () => {
+  it('emits the title heading, chat id and message count', () => {
+    const md = buildChatMarkdown({
+      title: 'My chat',
+      chatId: 'chat-123',
+      rows: [],
+    });
+    expect(md).toContain('# My chat');
+    expect(md).toContain('- Chat ID: `chat-123`');
+    expect(md).toContain('- Messages: 0');
+  });
+
+  it('falls back to "Untitled chat" with no title (en)', () => {
+    const md = buildChatMarkdown({ title: null, chatId: 'c', rows: [] });
+    expect(md).toContain('# Untitled chat');
+  });
+
+  it('localizes fixed labels with lang=ru (structure stays English)', () => {
+    const md = buildChatMarkdown({
+      title: null,
+      chatId: 'c',
+      lang: 'ru',
+      rows: [row({ role: 'assistant', content: 'hi' })],
+    });
+    expect(md).toContain('# Без названия');
+    expect(md).toContain('## 1. ИИ-агент');
+    // Structural words remain English.
+    expect(md).toContain('- Chat ID:');
+  });
+
+  it('numbers messages and labels roles (You / AI agent)', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({ role: 'user', content: 'question' }),
+        row({ role: 'assistant', content: 'answer' }),
+      ],
+    });
+    expect(md).toContain('## 1. You');
+    expect(md).toContain('question');
+    expect(md).toContain('## 2. AI agent');
+    expect(md).toContain('answer');
+  });
+
+  it('renders a tool part with fenced input/output and the friendly label', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({
+          role: 'assistant',
+          content: 'done',
+          metadata: {
+            parts: [
+              {
+                type: 'tool-getPage',
+                state: 'output-available',
+                input: { id: 'p1' },
+                output: { title: 'Hello' },
+              },
+              { type: 'text', text: 'done' },
+            ],
+          } as never,
+        }),
+      ],
+    });
+    expect(md).toContain('**Tool: Read page** (`getPage`) — done');
+    expect(md).toContain('Input:');
+    expect(md).toContain('"id": "p1"');
+    expect(md).toContain('Output:');
+    expect(md).toContain('"title": "Hello"');
+  });
+
+  // #186 re-review pt 1: restore the parity coverage of the removed client spec —
+  // error state, unknown-tool fallback (en + ru), and the circular-stringify catch.
+  it('renders a tool part in the error state with its errorText', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({
+          role: 'assistant',
+          metadata: {
+            parts: [
+              {
+                type: 'tool-getPage',
+                state: 'output-error',
+                input: { id: 'p1' },
+                errorText: 'page not found',
+              },
+            ],
+          } as never,
+        }),
+      ],
+    });
+    expect(md).toContain('**Tool: Read page** (`getPage`) — error');
+    expect(md).toContain('**Error:** page not found');
+  });
+
+  it('falls back to "Ran tool <name>" for an unknown tool (en) and the ru variant', () => {
+    const parts = [
+      {
+        type: 'tool-mysteryTool',
+        state: 'output-available',
+        output: { ok: 1 },
+      },
+    ];
+    const en = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [row({ role: 'assistant', metadata: { parts } as never })],
+    });
+    expect(en).toContain('**Tool: Ran tool mysteryTool** (`mysteryTool`)');
+    const ru = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      lang: 'ru',
+      rows: [row({ role: 'assistant', metadata: { parts } as never })],
+    });
+    expect(ru).toContain('Выполнил инструмент mysteryTool');
+  });
+
+  it('does not throw on a circular tool output (falls back to String)', () => {
+    const circular: Record<string, unknown> = {};
+    circular.self = circular;
+    expect(() =>
+      buildChatMarkdown({
+        title: 'T',
+        chatId: 'c',
+        rows: [
+          row({
+            role: 'assistant',
+            metadata: {
+              parts: [
+                {
+                  type: 'tool-getPage',
+                  state: 'output-available',
+                  output: circular,
+                },
+              ],
+            } as never,
+          }),
+        ],
+      }),
+    ).not.toThrow();
+  });
+
+  it('emits a token footer + total when usage is present', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({
+          role: 'assistant',
+          content: 'a',
+          metadata: {
+            usage: {
+              inputTokens: 100,
+              outputTokens: 20,
+              totalTokens: 120,
+              reasoningTokens: 8,
+            },
+          } as never,
+        }),
+      ],
+    });
+    expect(md).toContain('- Total tokens: 120');
+    expect(md).toContain(
+      '_Tokens — in: 100, out: 20, reasoning: 8, total: 120_',
+    );
+  });
+
+  it('flags a still-streaming (interrupted) row', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({ role: 'assistant', content: 'partial', status: 'streaming' }),
+      ],
+    });
+    expect(md).toContain('still being generated');
+  });
+
+  it('does NOT flag a completed row', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [row({ role: 'assistant', content: 'final', status: 'completed' })],
+    });
+    expect(md).not.toContain('still being generated');
+  });
+
+  it('renders a legacy NULL-status row (no parts) from plain content', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({ role: 'assistant', content: 'legacy answer', status: null }),
+      ],
+    });
+    expect(md).toContain('legacy answer');
+    expect(md).not.toContain('still being generated');
+  });
+
+  it('renders a persisted error', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({
+          role: 'assistant',
+          content: '',
+          status: 'error',
+          metadata: { error: '401: Unauthorized' } as never,
+        }),
+      ],
+    });
+    expect(md).toContain('**⚠️ Error:** 401: Unauthorized');
+  });
+
+  it('escapes embedded triple-backtick fences with a longer delimiter', () => {
+    const md = buildChatMarkdown({
+      title: 'T',
+      chatId: 'c',
+      rows: [
+        row({
+          role: 'assistant',
+          content: 'x',
+          metadata: {
+            parts: [
+              {
+                type: 'tool-getPage',
+                state: 'output-available',
+                output: '```inner```',
+              },
+            ],
+          } as never,
+        }),
+      ],
+    });
+    // A 4-backtick fence wraps content that itself contains a 3-backtick run.
+    expect(md).toContain('````');
+  });
+});

apps/server/src/core/ai-chat/chat-markdown.util.ts

@@ -0,0 +1,299 @@
+/**
+ * Server-side Markdown export for an AI agent chat (#183). The DB is the single
+ * source of truth: this renders a chat purely from its persisted message rows
+ * (`AiChatMessage[]` — role / content / metadata.parts / toolCalls / usage).
+ * Because the assistant row is now persisted UPFRONT and updated per step, an
+ * interrupted turn is included up to its last finished step.
+ *
+ * Ported from the client `utils/chat-markdown.ts`. It is a PURE function (apart
+ * from `new Date()` for the export timestamp), so it is straightforward to
+ * unit-test and a future background worker can reuse it.
+ *
+ * Only a few fixed role/tool labels are localized via the `lang` param; the
+ * structural document words (Input/Output/Error/Tokens/...) stay English because
+ * the output is a technical artifact.
+ */
+
+import type { AiChatMessage } from '@docmost/db/types/entity.types';
+
+/** Supported export label languages. Defaults to English. */
+export type ExportLang = 'en' | 'ru';
+
+/**
+ * Normalize an arbitrary client locale code to a supported export language. The
+ * client sends `i18n.language`, which is a FULL locale tag (e.g. `en-US`,
+ * `ru-RU`), not a bare `en`/`ru` — so match on the language subtag and fall back
+ * to English for anything non-Russian.
+ */
+export function normalizeLang(lang?: string): ExportLang {
+  return lang?.toLowerCase().startsWith('ru') ? 'ru' : 'en';
+}
+
+/** A single AI SDK UIMessage part (text part or a tool part). */
+interface ExportPart {
+  type: string;
+  text?: string;
+  state?: string;
+  toolName?: string;
+  input?: unknown;
+  output?: unknown;
+  errorText?: string;
+}
+
+/** Authoritative per-turn usage the server attaches to a message row. */
+interface UsageLike {
+  inputTokens?: number;
+  outputTokens?: number;
+  totalTokens?: number;
+  reasoningTokens?: number;
+}
+
+/** Localized label table. The client-side Markdown builder was removed by #183
+ *  (the export is now server-side only), so this no longer mirrors a second
+ *  exporter — instead the tool-action labels are kept in parity with the
+ *  on-screen action-log labels in the client's `tool-parts.tsx` (`toolLabelKey`)
+ *  so the export reads the same as the UI. Only role + tool-action labels are
+ *  localized; everything structural is an English constant in the renderer. */
+const LABELS: Record<
+  ExportLang,
+  {
+    untitled: string;
+    aiAgent: string;
+    you: string;
+    tools: Record<string, string>;
+    ranTool: (name: string) => string;
+    stillGenerating: string;
+  }
+> = {
+  en: {
+    untitled: 'Untitled chat',
+    aiAgent: 'AI agent',
+    you: 'You',
+    tools: {
+      searchPages: 'Searched pages',
+      getPage: 'Read page',
+      createPage: 'Created page',
+      updatePageContent: 'Updated page',
+      renamePage: 'Renamed page',
+      movePage: 'Moved page',
+      deletePage: 'Deleted page (to trash)',
+      createComment: 'Commented',
+      resolveComment: 'Resolved comment',
+    },
+    ranTool: (name) => `Ran tool ${name}`,
+    stillGenerating:
+      'This message is still being generated — the export captured a partial, in-progress response.',
+  },
+  ru: {
+    untitled: 'Без названия',
+    aiAgent: 'ИИ-агент',
+    you: 'Вы',
+    tools: {
+      searchPages: 'Искал по страницам',
+      getPage: 'Прочитал страницу',
+      createPage: 'Создал страницу',
+      updatePageContent: 'Обновил страницу',
+      renamePage: 'Переименовал страницу',
+      movePage: 'Переместил страницу',
+      deletePage: 'Удалил страницу (в корзину)',
+      createComment: 'Прокомментировал',
+      resolveComment: 'Закрыл комментарий',
+    },
+    ranTool: (name) => `Выполнил инструмент ${name}`,
+    stillGenerating:
+      'Это сообщение всё ещё генерируется — экспорт захватил частичный, незавершённый ответ.',
+  },
+};
+
+/** True for AI SDK tool parts (static `tool-*` or `dynamic-tool`). */
+function isToolPart(type: string): boolean {
+  return type.startsWith('tool-') || type === 'dynamic-tool';
+}
+
+/** Extract the tool name from a part `type` of `tool-${name}` (or dynamic). */
+function getToolName(part: ExportPart): string {
+  if (part.type === 'dynamic-tool') return part.toolName ?? '';
+  return part.type.startsWith('tool-')
+    ? part.type.slice('tool-'.length)
+    : part.type;
+}
+
+/** Map an AI SDK tool-part state to the 3 states the action-log renders. */
+function toolRunState(state: string | undefined): 'running' | 'done' | 'error' {
+  if (state === 'output-error' || state === 'output-denied') return 'error';
+  if (state === 'output-available') return 'done';
+  return 'running';
+}
+
+/** Resolve a tool's friendly action-log label (localized) from its name. */
+function toolLabel(name: string, lang: ExportLang): string {
+  return LABELS[lang].tools[name] ?? LABELS[lang].ranTool(name);
+}
+
+/**
+ * Stringify an arbitrary tool input/output value for a fenced block. Strings
+ * pass through as-is; everything else is pretty-printed JSON, falling back to
+ * `String(value)` if serialization throws (e.g. a circular structure).
+ */
+function stringify(value: unknown): string {
+  if (typeof value === 'string') return value;
+  try {
+    return JSON.stringify(value, null, 2);
+  } catch {
+    return String(value);
+  }
+}
+
+/**
+ * Wrap `code` in a fenced code block whose backtick delimiter is LONGER than the
+ * longest backtick run inside the content, so embedded backticks (or a literal
+ * ``` fence) never break out of the block. Minimum 3 backticks.
+ */
+function fence(code: string, lang = ''): string {
+  const runs: string[] = code.match(/`+/g) ?? [];
+  const longest = runs.reduce((m, s) => Math.max(m, s.length), 0);
+  const delim = '`'.repeat(Math.max(3, longest + 1));
+  return `${delim}${lang}\n${code}\n${delim}`;
+}
+
+/** Per-row token count, mirroring the header sum in the client window. */
+function rowTokens(usage: UsageLike): number {
+  return (
+    usage.totalTokens ?? (usage.inputTokens ?? 0) + (usage.outputTokens ?? 0)
+  );
+}
+
+/** Render one message's UIMessage parts into an array of Markdown blocks
+ *  (text blocks + tool blocks). Mirrors the client renderer / MessageItem. */
+function renderMessageParts(parts: ExportPart[], lang: ExportLang): string[] {
+  const out: string[] = [];
+
+  for (const part of parts) {
+    if (part.type === 'text') {
+      const text = (part.text ?? '').trim();
+      if (text.length > 0) out.push(text);
+      continue;
+    }
+
+    if (!isToolPart(part.type)) continue;
+
+    const name = getToolName(part);
+    const label = toolLabel(name, lang);
+    const state = toolRunState(part.state);
+
+    const toolLines: string[] = [`**Tool: ${label}** (\`${name}\`) — ${state}`];
+    if (part.input !== undefined) {
+      toolLines.push('Input:');
+      toolLines.push(fence(stringify(part.input), 'json'));
+    }
+    if (part.output !== undefined) {
+      toolLines.push('Output:');
+      toolLines.push(fence(stringify(part.output), 'json'));
+    }
+    if (part.errorText) {
+      toolLines.push(`**Error:** ${part.errorText}`);
+    }
+    out.push(toolLines.join('\n\n'));
+  }
+
+  return out;
+}
+
+/** Resolve a persisted row's parts: prefer the rich persisted parts, else a
+ *  single text part built from the plain-text content (mirrors rowToUiMessage). */
+function rowParts(row: AiChatMessage): ExportPart[] {
+  const meta = (row.metadata ?? {}) as { parts?: ExportPart[] };
+  return Array.isArray(meta.parts) && meta.parts.length > 0
+    ? meta.parts
+    : [{ type: 'text', text: row.content ?? '' }];
+}
+
+/**
+ * Serialize a chat to a Markdown string from its persisted rows. Source = DB
+ * ONLY (no live client state). A row whose `status` is still 'streaming' is an
+ * interrupted turn that the export captured mid-flight; it is rendered up to its
+ * last finished step and flagged "still generating".
+ */
+export function buildChatMarkdown(args: {
+  title: string | null;
+  chatId: string;
+  rows: AiChatMessage[];
+  // Accepts a full client locale tag (e.g. 'en-US'/'ru-RU'); normalized below.
+  lang?: string;
+}): string {
+  const { title, chatId, rows } = args;
+  const lang: ExportLang = normalizeLang(args.lang);
+  const L = LABELS[lang];
+  const blocks: string[] = [];
+
+  const heading = (title ?? '').trim() || L.untitled;
+  blocks.push(`# ${heading}`);
+
+  const usageOf = (row: AiChatMessage): UsageLike | undefined => {
+    const meta = (row.metadata ?? {}) as { usage?: UsageLike };
+    return meta.usage;
+  };
+  const errorOf = (row: AiChatMessage): string | undefined => {
+    const meta = (row.metadata ?? {}) as { error?: string };
+    return meta.error;
+  };
+
+  // Metadata bullet list. Total tokens is only shown when there is a sum.
+  const totalTokens = rows.reduce((sum, row) => {
+    const usage = usageOf(row);
+    return usage ? sum + rowTokens(usage) : sum;
+  }, 0);
+  const meta = [
+    `- Chat ID: \`${chatId}\``,
+    `- Exported: ${new Date().toISOString()}`,
+    `- Messages: ${rows.length}`,
+  ];
+  if (totalTokens > 0) meta.push(`- Total tokens: ${totalTokens}`);
+  blocks.push(meta.join('\n'));
+
+  rows.forEach((row, index) => {
+    blocks.push('---');
+
+    const roleLabel = row.role === 'assistant' ? L.aiAgent : L.you;
+    blocks.push(`## ${index + 1}. ${roleLabel}`);
+
+    // Created-at kept in source as an HTML comment (out of the rendered prose).
+    if (row.createdAt) {
+      const iso =
+        row.createdAt instanceof Date
+          ? row.createdAt.toISOString()
+          : String(row.createdAt);
+      blocks.push(`<!-- ${iso} -->`);
+    }
+
+    blocks.push(...renderMessageParts(rowParts(row), lang));
+
+    // A still-'streaming' row is an interrupted/in-progress turn captured by the
+    // export; record that so the partial answer is not mistaken for complete.
+    if (row.status === 'streaming') {
+      blocks.push(`_⏳ ${L.stillGenerating}_`);
+    }
+
+    const error = errorOf(row);
+    if (error) {
+      blocks.push(`**⚠️ Error:** ${error}`);
+    }
+
+    const usage = usageOf(row);
+    if (usage) {
+      const total = usage.totalTokens ?? rowTokens(usage);
+      const reasoning =
+        usage.reasoningTokens && usage.reasoningTokens > 0
+          ? `, reasoning: ${usage.reasoningTokens}`
+          : '';
+      blocks.push(
+        `_Tokens — in: ${usage.inputTokens ?? '?'}, out: ${
+          usage.outputTokens ?? '?'
+        }${reasoning}, total: ${total}_`,
+      );
+    }
+  });
+
+  // Blank line between blocks so the Markdown renders cleanly.
+  return blocks.join('\n\n');
+}

apps/server/src/core/ai-chat/dto/ai-chat.dto.ts

@@ -26,3 +26,17 @@ export class GetChatMessagesDto {
   @IsString()
   cursor?: string;
 }
+
+/** Export a chat to Markdown (#183). `lang` localizes the few fixed
+ *  role/tool-action labels; defaults to English server-side. */
+export class ExportChatDto {
+  @IsString()
+  chatId: string;
+
+  // A full client locale tag (e.g. 'en-US', 'ru-RU') — normalized server-side to
+  // a supported export language (see normalizeLang). Accept any string so a
+  // region-qualified locale is not rejected (the 400 that broke the real client).
+  @IsOptional()
+  @IsString()
+  lang?: string;
+}

apps/server/src/core/ai-chat/public-share-chat.controller.spec.ts

@@ -34,6 +34,7 @@ describe('resolveShareAssistantRequest (extracted controller funnel)', () => {
     resolveShareRole?: jest.Mock;
     getShareChatModel?: jest.Mock;
     tryConsumeWorkspaceQuota?: jest.Mock;
+    withinShareTokenBudget?: jest.Mock;
   } = {}) {
     const aiSettings = {
       isPublicShareAssistantEnabled: jest
@@ -65,6 +66,8 @@ describe('resolveShareAssistantRequest (extracted controller funnel)', () => {
         over.getShareChatModel ?? jest.fn().mockResolvedValue('MODEL'),
       tryConsumeWorkspaceQuota:
         over.tryConsumeWorkspaceQuota ?? jest.fn().mockResolvedValue(true),
+      withinShareTokenBudget:
+        over.withinShareTokenBudget ?? jest.fn().mockResolvedValue(true),
     };
     const deps: ShareAssistantDeps = {
       aiSettings: aiSettings as never,
@@ -191,6 +194,39 @@ describe('resolveShareAssistantRequest (extracted controller funnel)', () => {
     expect(publicShareChat.tryConsumeWorkspaceQuota).toHaveBeenCalledWith('ws-1');
   });
 
+  it('withinShareTokenBudget false => 429 thrown BEFORE any stream (cost cap, #159 #5)', async () => {
+    const { deps, publicShareChat } = makeDeps({
+      withinShareTokenBudget: jest.fn().mockResolvedValue(false),
+    });
+    expect(await statusOf(deps, body())).toBe(429);
+    expect(publicShareChat.withinShareTokenBudget).toHaveBeenCalledWith('ws-1');
+    // The token budget is the COST backstop: an over-budget workspace must be
+    // rejected WITHOUT consuming a request slot, so the request cap never runs.
+    expect(publicShareChat.tryConsumeWorkspaceQuota).not.toHaveBeenCalled();
+  });
+
+  it('the token budget is checked BEFORE the request cap (over-budget wins, no slot spent)', async () => {
+    // Over budget AND the request cap would also reject: the read-only budget
+    // gate must win so the (mutating) request-slot consume is never reached.
+    const { deps, publicShareChat } = makeDeps({
+      withinShareTokenBudget: jest.fn().mockResolvedValue(false),
+      tryConsumeWorkspaceQuota: jest.fn().mockResolvedValue(false),
+    });
+    expect(await statusOf(deps, body())).toBe(429);
+    expect(publicShareChat.tryConsumeWorkspaceQuota).not.toHaveBeenCalled();
+  });
+
+  it('the token-budget gate is checked BEFORE the payload caps (429 wins over 413)', async () => {
+    const { deps } = makeDeps({
+      withinShareTokenBudget: jest.fn().mockResolvedValue(false),
+    });
+    const huge = {
+      role: 'user',
+      parts: [{ type: 'text', text: 'x'.repeat(MAX_SHARE_MESSAGE_CHARS + 1) }],
+    };
+    expect(await statusOf(deps, body({ messages: [huge] }))).toBe(429);
+  });
+
   it('messages over MAX_SHARE_MESSAGES => 413', async () => {
     const { deps } = makeDeps();
     const tooMany = Array.from({ length: MAX_SHARE_MESSAGES + 1 }, () => ({

apps/server/src/core/ai-chat/public-share-chat.controller.ts

@@ -151,6 +151,7 @@ export interface ShareAssistantDeps {
     | 'resolveShareRole'
     | 'getShareChatModel'
     | 'tryConsumeWorkspaceQuota'
+    | 'withinShareTokenBudget'
   >;
 }
 
@@ -267,9 +268,21 @@ export async function resolveShareAssistantRequest(
     throw new NotFoundException('Not found');
   }
 
-  // 5. Per-WORKSPACE anti-abuse cap (IP-independent; defense in depth). Checked
-  //    BEFORE res.hijack(), so an over-cap workspace gets a clean 429 and spends
-  //    nothing.
+  // 5a. Per-WORKSPACE rolling-day TOKEN budget (the COST backstop). Read-only and
+  //     checked FIRST so a workspace that has already burned its day's token
+  //     budget gets a clean 429 WITHOUT consuming a request slot, and spends
+  //     nothing. Counting requests alone does not bound the owner's provider
+  //     bill (issue #159, finding #5).
+  if (!(await deps.publicShareChat.withinShareTokenBudget(workspaceId))) {
+    throw new HttpException(
+      'This documentation assistant has reached its usage budget. Please try again later.',
+      HttpStatus.TOO_MANY_REQUESTS,
+    );
+  }
+
+  // 5b. Per-WORKSPACE anti-abuse request cap (IP-independent; defense in depth).
+  //     Checked BEFORE res.hijack(), so an over-cap workspace gets a clean 429
+  //     and spends nothing.
   if (!(await deps.publicShareChat.tryConsumeWorkspaceQuota(workspaceId))) {
     throw new HttpException(
       'This documentation assistant is temporarily busy. Please try again later.',

apps/server/src/core/ai-chat/public-share-chat.service.ts

@@ -17,7 +17,9 @@ import { buildShareSystemPrompt } from './public-share-chat.prompt';
 import { roleModelOverride } from './roles/role-model-config';
 import {
   PublicShareWorkspaceLimiter,
+  PublicShareWorkspaceTokenBudget,
   createPublicShareWorkspaceLimiter,
+  createPublicShareWorkspaceTokenBudget,
 } from './public-share-workspace-limiter';
 import { describeProviderError } from '../../integrations/ai/ai-error.util';
 import {
@@ -125,6 +127,16 @@ export class PublicShareChatService {
    */
   private readonly workspaceLimiter: PublicShareWorkspaceLimiter;
 
+  /**
+   * COST contour two: a per-workspace TOKEN budget over a rolling day. The
+   * request-count limiter above bounds how many anonymous calls run; this bounds
+   * how many provider TOKENS they spend (input re-sent per step + output),
+   * which is what the owner is actually billed for (issue #159, finding #5).
+   * Checked read-only before a turn streams; the real usage is recorded once the
+   * turn finishes (`onFinish`).
+   */
+  private readonly tokenBudget: PublicShareWorkspaceTokenBudget;
+
   constructor(
     private readonly ai: AiService,
     private readonly aiSettings: AiSettingsService,
@@ -133,6 +145,7 @@ export class PublicShareChatService {
     private readonly aiAgentRoleRepo: AiAgentRoleRepo,
   ) {
     this.workspaceLimiter = createPublicShareWorkspaceLimiter(redisService);
+    this.tokenBudget = createPublicShareWorkspaceTokenBudget(redisService);
   }
 
   /**
@@ -144,6 +157,48 @@ export class PublicShareChatService {
     return this.workspaceLimiter.tryConsume(workspaceId);
   }
 
+  /**
+   * Read-only pre-stream COST gate: true while the workspace is under its
+   * rolling-day token budget, false once the trailing-day token spend has
+   * reached it (the controller must then 429 BEFORE starting the stream). This
+   * bounds the owner's actual provider bill, which counting requests alone does
+   * not (issue #159, finding #5).
+   */
+  async withinShareTokenBudget(workspaceId: string): Promise<boolean> {
+    return this.tokenBudget.withinBudget(workspaceId);
+  }
+
+  /**
+   * Record a finished turn's real token spend against the rolling-day budget.
+   * Best-effort (the turn already ran): failures are swallowed by the budget.
+   */
+  async recordShareTokens(workspaceId: string, tokens: number): Promise<void> {
+    return this.tokenBudget.record(workspaceId, tokens);
+  }
+
+  /**
+   * `streamText` onFinish hook body: account a finished turn's REAL token spend
+   * (input re-sent per step + output, summed across all steps) against the
+   * per-workspace rolling-day budget, so a future turn over budget is rejected up
+   * front (issue #159, finding #5). `totalUsage` fields are `number | undefined`;
+   * fall back to the sum of input+output when the provider omits `totalTokens`.
+   * Fire-and-forget: the turn already streamed, so a record failure must not
+   * break it.
+   */
+  recordTurnUsage(
+    workspaceId: string,
+    totalUsage: {
+      totalTokens?: number;
+      inputTokens?: number;
+      outputTokens?: number;
+    },
+  ): void {
+    const tokens =
+      totalUsage.totalTokens ??
+      (totalUsage.inputTokens ?? 0) + (totalUsage.outputTokens ?? 0);
+    void this.recordShareTokens(workspaceId, tokens);
+  }
+
   /**
    * Resolve the admin-selected agent role for the anonymous public-share
    * assistant, scoped to the workspace and soft-delete aware. Returns null when
@@ -231,6 +286,8 @@ export class PublicShareChatService {
         // bill even if the per-IP throttle is evaded; worst case = steps × this.
         maxOutputTokens: resolveShareAiMaxOutputTokens(),
         abortSignal: signal,
+        onFinish: ({ totalUsage }) =>
+          this.recordTurnUsage(workspaceId, totalUsage),
         onError: ({ error }) => {
           // Reuse the shared formatter so provider error formatting stays
           // unified (statusCode + body) with the authenticated path.

apps/server/src/core/ai-chat/public-share-chat.spec.ts

@@ -11,8 +11,11 @@ import {
 import { PublicShareChatToolsService } from './tools/public-share-chat-tools.service';
 import {
   PublicShareWorkspaceLimiter,
+  PublicShareWorkspaceTokenBudget,
   resolveShareAiWorkspaceMax,
+  resolveShareAiWorkspaceTokenBudget,
   SHARE_AI_WORKSPACE_MAX_PER_WINDOW,
+  SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT,
 } from './public-share-workspace-limiter';
 
 /**
@@ -546,6 +549,228 @@ describe('PublicShareWorkspaceLimiter (cluster-wide sliding-window per-workspace
   });
 });
 
+/**
+ * In-memory fake of the ioredis slice the TOKEN budget uses. Unlike the request
+ * limiter (one Lua), the budget runs TWO scripts over the same sorted set:
+ *  - the read-only CHECK (sums the token counts encoded as each member's leading
+ *    integer, admits while the sum is under budget, never mutates), and
+ *  - the RECORD (ZADDs a finished turn's `<tokens>:<unique>` member).
+ * The fake faithfully reproduces both (branching on the script body) so the spec
+ * exercises the REAL budget math, not a re-implementation.
+ */
+class FakeTokenRedis {
+  private sets = new Map<string, Array<{ score: number; member: string }>>();
+
+  async eval(
+    script: string,
+    _numKeys: number,
+    key: string,
+    nowStr: string,
+    windowMsStr: string,
+    arg3: string,
+  ): Promise<number> {
+    const now = Number(nowStr);
+    const windowMs = Number(windowMsStr);
+    const cutoff = now - windowMs;
+    const arr = (this.sets.get(key) ?? []).filter((e) => e.score > cutoff);
+    if (script.includes('ZADD')) {
+      // RECORD: arg3 is the `<tokens>:<unique>` member; append at score=now.
+      arr.push({ score: now, member: arg3 });
+      this.sets.set(key, arr);
+      return 1;
+    }
+    // CHECK: arg3 is the budget; sum the leading integer of each survivor.
+    const budget = Number(arg3);
+    this.sets.set(key, arr);
+    const total = arr.reduce((sum, e) => {
+      const m = /^(\d+)/.exec(e.member);
+      return sum + (m ? Number(m[1]) : 0);
+    }, 0);
+    return total >= budget ? 0 : 1;
+  }
+}
+
+function makeTokenBudget(budget: number, windowMs: number, clock: () => number) {
+  const redis = new FakeTokenRedis() as unknown as import('ioredis').Redis;
+  return new PublicShareWorkspaceTokenBudget(redis, budget, windowMs, clock);
+}
+
+describe('resolveShareAiWorkspaceTokenBudget (env-overridable per-day token budget)', () => {
+  const KEY = 'SHARE_AI_WORKSPACE_TOKEN_BUDGET_PER_DAY';
+  const saved = process.env[KEY];
+  afterEach(() => {
+    if (saved === undefined) delete process.env[KEY];
+    else process.env[KEY] = saved;
+  });
+
+  it('falls back to the default when unset', () => {
+    delete process.env[KEY];
+    expect(resolveShareAiWorkspaceTokenBudget()).toBe(
+      SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT,
+    );
+  });
+
+  it('honors a positive override', () => {
+    process.env[KEY] = '250000';
+    expect(resolveShareAiWorkspaceTokenBudget()).toBe(250000);
+  });
+
+  it('ignores a non-positive / unparseable value (uses the default)', () => {
+    for (const bad of ['0', '-5', 'nope', '']) {
+      process.env[KEY] = bad;
+      expect(resolveShareAiWorkspaceTokenBudget()).toBe(
+        SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT,
+      );
+    }
+  });
+});
+
+describe('PublicShareWorkspaceTokenBudget (cluster-wide rolling-day token cap)', () => {
+  it('admits while under budget and rejects once the recorded spend reaches it', async () => {
+    const budget = makeTokenBudget(1000, 60_000, () => 1_000);
+    expect(await budget.withinBudget('ws-1')).toBe(true); // nothing spent yet
+    await budget.record('ws-1', 600);
+    expect(await budget.withinBudget('ws-1')).toBe(true); // 600 < 1000
+    await budget.record('ws-1', 400);
+    // 1000 >= 1000: the budget is exhausted, so the next turn is rejected up front.
+    expect(await budget.withinBudget('ws-1')).toBe(false);
+  });
+
+  it('counts TOKENS, not requests: one fat turn can exhaust the budget alone', async () => {
+    const budget = makeTokenBudget(1000, 60_000, () => 1_000);
+    // A single accepted turn re-sends the whole transcript across 5 steps; here
+    // it lands as 1200 tokens — already over the day budget on its own.
+    await budget.record('ws-1', 1200);
+    expect(await budget.withinBudget('ws-1')).toBe(false);
+  });
+
+  it('ages out spend older than the window so the budget recovers', async () => {
+    let now = 0;
+    const budget = makeTokenBudget(1000, 60_000, () => now);
+    await budget.record('ws-1', 1000); // at budget
+    now += 59_999; // still inside the day window
+    expect(await budget.withinBudget('ws-1')).toBe(false);
+    now += 2; // the spend is now strictly older than windowMs
+    expect(await budget.withinBudget('ws-1')).toBe(true);
+  });
+
+  it('ignores non-positive / non-finite usage (never records phantom spend)', async () => {
+    const budget = makeTokenBudget(1000, 60_000, () => 1_000);
+    await budget.record('ws-1', 0);
+    await budget.record('ws-1', -50);
+    await budget.record('ws-1', Number.NaN);
+    await budget.record('ws-1', Infinity);
+    expect(await budget.withinBudget('ws-1')).toBe(true); // nothing accumulated
+  });
+
+  it('keeps separate budgets per workspace', async () => {
+    const budget = makeTokenBudget(500, 60_000, () => 1_000);
+    await budget.record('ws-a', 500); // ws-a exhausted
+    expect(await budget.withinBudget('ws-a')).toBe(false);
+    expect(await budget.withinBudget('ws-b')).toBe(true); // ws-b untouched
+  });
+
+  it('FAILS CLOSED on the read-only check when Redis rejects', async () => {
+    const failingRedis = {
+      eval: () => Promise.reject(new Error('redis down')),
+    } as unknown as import('ioredis').Redis;
+    const budget = new PublicShareWorkspaceTokenBudget(
+      failingRedis,
+      1000,
+      60_000,
+      () => 1_000,
+    );
+    const errSpy = jest
+      .spyOn(Logger.prototype, 'error')
+      .mockImplementation(() => undefined);
+    expect(await budget.withinBudget('ws-1')).toBe(false);
+    expect(errSpy).toHaveBeenCalled();
+    errSpy.mockRestore();
+  });
+
+  it('SWALLOWS a record failure (best-effort post-accounting, never throws)', async () => {
+    // The turn already streamed; a record failure must not surface to the caller.
+    const failingRedis = {
+      eval: () => Promise.reject(new Error('redis down')),
+    } as unknown as import('ioredis').Redis;
+    const budget = new PublicShareWorkspaceTokenBudget(
+      failingRedis,
+      1000,
+      60_000,
+      () => 1_000,
+    );
+    const errSpy = jest
+      .spyOn(Logger.prototype, 'error')
+      .mockImplementation(() => undefined);
+    await expect(budget.record('ws-1', 100)).resolves.toBeUndefined();
+    expect(errSpy).toHaveBeenCalled();
+    errSpy.mockRestore();
+  });
+});
+
+describe('PublicShareChatService.withinShareTokenBudget / recordShareTokens', () => {
+  it('delegates the cost gate + accounting to the redis-backed token budget', async () => {
+    const redis = new FakeTokenRedis();
+    const redisService = { getOrThrow: () => redis } as never;
+    const service = new PublicShareChatService(
+      {} as never,
+      {} as never,
+      {} as never,
+      redisService,
+      {} as never,
+    );
+    // Default budget is large, so a fresh workspace is under budget; recording a
+    // modest spend keeps it under budget (asserts the wiring the controller +
+    // onFinish rely on).
+    expect(await service.withinShareTokenBudget('ws-1')).toBe(true);
+    await service.recordShareTokens('ws-1', 1234);
+    expect(await service.withinShareTokenBudget('ws-1')).toBe(true);
+  });
+});
+
+describe('PublicShareChatService.recordTurnUsage (streamText onFinish accounting)', () => {
+  function makeService() {
+    const redisService = { getOrThrow: () => new FakeTokenRedis() } as never;
+    const service = new PublicShareChatService(
+      {} as never,
+      {} as never,
+      {} as never,
+      redisService,
+      {} as never,
+    );
+    const recordSpy = jest
+      .spyOn(service, 'recordShareTokens')
+      .mockResolvedValue(undefined);
+    return { service, recordSpy };
+  }
+
+  it('sums input+output when the provider omits totalTokens', () => {
+    const { service, recordSpy } = makeService();
+    // The onFinish payload shape: a totalUsage with per-component counts but no
+    // authoritative total (provider omitted it).
+    service.recordTurnUsage('ws-1', { inputTokens: 1200, outputTokens: 300 });
+    expect(recordSpy).toHaveBeenCalledWith('ws-1', 1500);
+  });
+
+  it('treats missing input/output components as 0 in the fallback sum', () => {
+    const { service, recordSpy } = makeService();
+    service.recordTurnUsage('ws-1', { outputTokens: 42 });
+    expect(recordSpy).toHaveBeenCalledWith('ws-1', 42);
+  });
+
+  it('prefers the authoritative totalTokens when present (not the sum)', () => {
+    const { service, recordSpy } = makeService();
+    // totalTokens is the provider's authoritative figure and may differ from a
+    // naive input+output sum (e.g. cached/ reasoning tokens); it must win.
+    service.recordTurnUsage('ws-1', {
+      totalTokens: 5000,
+      inputTokens: 1200,
+      outputTokens: 300,
+    });
+    expect(recordSpy).toHaveBeenCalledWith('ws-1', 5000);
+  });
+});
+
 describe('PublicShareChatService.tryConsumeWorkspaceQuota', () => {
   it('delegates to the redis-backed per-workspace limiter', async () => {
     const redis = new FakeRedis();

apps/server/src/core/ai-chat/public-share-workspace-limiter.ts

@@ -136,6 +136,177 @@ export class PublicShareWorkspaceLimiter {
   }
 }
 
+/**
+ * SECOND cost contour: a per-workspace TOKEN budget over a rolling DAY.
+ *
+ * The request-count cap above bounds how MANY anonymous calls a workspace
+ * admits, but NOT how expensive each one is: one accepted call runs the agent
+ * loop up to `stepCountIs(5)`, and every step re-sends the WHOLE client-held
+ * transcript (~hundreds of KB) as input, so the provider input alone can be tens
+ * of thousands of tokens PER step while `maxOutputTokens` only caps the output.
+ * The request cap is also hourly with no daily ceiling, so a steady stream at
+ * the hourly cap sustains ~24x its count per day. Counting requests therefore
+ * does not bound the owner's actual LLM bill (issue #159, finding #5).
+ *
+ * This contour caps the SPEND directly: the actual tokens consumed (input +
+ * output, summed across all steps of every accepted turn) over the trailing
+ * `windowMs` (one rolling day) must stay under `budget`. It is checked BEFORE a
+ * turn streams (read-only) and the turn's real usage is recorded AFTER it
+ * finishes (`streamText` onFinish). Like the request cap it is cluster-wide
+ * (shared Redis) and uses a sliding-window LOG so the day boundary cannot be
+ * gamed for a 2x burst.
+ *
+ * Pre-check is read-only, so a turn already over budget is rejected, but the
+ * tokens of an in-flight turn are not yet known and are accounted only once it
+ * finishes. The worst-case overshoot past the budget is therefore one turn
+ * (bounded by steps x (maxOutputTokens + transcript size)) — acceptable for a
+ * cost backstop on an optional anonymous assistant.
+ */
+
+/** Default per-workspace token budget over the rolling day. */
+export const SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT = 1_000_000;
+/** Default token-budget window length: one rolling day. */
+export const SHARE_AI_WORKSPACE_TOKEN_WINDOW_MS = 24 * 60 * 60 * 1000;
+
+/** Redis key namespace for the per-workspace token-spend sliding-window log. */
+const TOKEN_KEY_PREFIX = 'share-ai:ws-tokens:';
+
+/**
+ * Read-only sliding-window token-budget check.
+ *
+ * KEYS[1] = the per-workspace token sorted-set key
+ * ARGV[1] = now (epoch ms)
+ * ARGV[2] = windowMs
+ * ARGV[3] = budget (max tokens in the trailing window)
+ *
+ * Drops entries older than the window, then sums the token counts encoded as the
+ * leading integer of each surviving member. Returns 1 if the running total is
+ * still UNDER budget (admit), 0 once it has reached/exceeded the budget. Does NOT
+ * add anything — the turn's real usage is recorded separately once it finishes.
+ */
+const TOKEN_BUDGET_CHECK_LUA = `
+local key = KEYS[1]
+local now = tonumber(ARGV[1])
+local windowMs = tonumber(ARGV[2])
+local budget = tonumber(ARGV[3])
+redis.call('ZREMRANGEBYSCORE', key, 0, now - windowMs)
+local members = redis.call('ZRANGE', key, 0, -1)
+local total = 0
+for i = 1, #members do
+  local t = tonumber(string.match(members[i], '^(%d+)'))
+  if t then total = total + t end
+end
+if total >= budget then
+  return 0
+end
+return 1
+`;
+
+/**
+ * Record one finished turn's token spend in the sliding-window log.
+ *
+ * KEYS[1] = the per-workspace token sorted-set key
+ * ARGV[1] = now (epoch ms) — the entry score
+ * ARGV[2] = windowMs
+ * ARGV[3] = member (`<tokens>:<unique>`; the leading integer is the token count)
+ *
+ * Always ZADDs (the turn already ran and spent the tokens) and refreshes the
+ * key TTL so idle workspaces cost no memory. Trims expired entries first so the
+ * set never grows unbounded for a busy workspace.
+ */
+const TOKEN_RECORD_LUA = `
+local key = KEYS[1]
+local now = tonumber(ARGV[1])
+local windowMs = tonumber(ARGV[2])
+local member = ARGV[3]
+redis.call('ZREMRANGEBYSCORE', key, 0, now - windowMs)
+redis.call('ZADD', key, now, member)
+redis.call('PEXPIRE', key, windowMs)
+return 1
+`;
+
+/**
+ * Cluster-wide, sliding-window per-workspace TOKEN budget backed by Redis.
+ * `withinBudget(key)` is a read-only pre-stream gate; `record(key, tokens)`
+ * accounts a finished turn's real usage. Decoupled from NestJS so it is testable
+ * against a mocked/real ioredis client, mirroring the request-count limiter.
+ */
+export class PublicShareWorkspaceTokenBudget {
+  private readonly logger = new Logger(PublicShareWorkspaceTokenBudget.name);
+  private counter = 0;
+
+  constructor(
+    private readonly redis: Redis,
+    private readonly budget: number = SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT,
+    private readonly windowMs: number = SHARE_AI_WORKSPACE_TOKEN_WINDOW_MS,
+    private readonly now: () => number = Date.now,
+  ) {}
+
+  /**
+   * Read-only pre-stream check. Returns true while the workspace is under its
+   * rolling-day token budget, false once the trailing-window spend has reached
+   * it (caller must then 429 BEFORE streaming any tokens).
+   *
+   * FAILS CLOSED (false) on a Redis error: identical reasoning to the request
+   * limiter — when we cannot prove the workspace is under budget we DENY rather
+   * than admit an unmetered billable call. The assistant is optional, so a
+   * transient Redis blip briefly disabling it beats an unbounded provider bill.
+   */
+  async withinBudget(key: string): Promise<boolean> {
+    const t = this.now();
+    try {
+      const admitted = await this.redis.eval(
+        TOKEN_BUDGET_CHECK_LUA,
+        1,
+        TOKEN_KEY_PREFIX + key,
+        String(t),
+        String(this.windowMs),
+        String(this.budget),
+      );
+      return admitted === 1;
+    } catch (err) {
+      this.logger.error(
+        `share-ai token budget Redis failure for key "${key}"; failing closed`,
+        err as Error,
+      );
+      return false;
+    }
+  }
+
+  /**
+   * Record a finished turn's token spend. Best-effort: the turn already ran, so
+   * a Redis failure here is logged but not propagated — it would only cause a
+   * slight under-count of the running budget, never a wrong answer to the
+   * caller. Non-positive / non-finite usage is ignored.
+   */
+  async record(key: string, tokens: number): Promise<void> {
+    if (!Number.isFinite(tokens) || tokens <= 0) return;
+    const spend = Math.floor(tokens);
+    const t = this.now();
+    // Member: `<tokens>:<unique>` — the check Lua sums the leading integer, and
+    // the unique suffix keeps distinct turns in the same ms from colliding on
+    // the sorted-set member (which would drop one entry and under-count).
+    const member = `${spend}:${t}-${this.counter++}-${Math.random()
+      .toString(36)
+      .slice(2)}`;
+    try {
+      await this.redis.eval(
+        TOKEN_RECORD_LUA,
+        1,
+        TOKEN_KEY_PREFIX + key,
+        String(t),
+        String(this.windowMs),
+        member,
+      );
+    } catch (err) {
+      this.logger.error(
+        `share-ai token budget record failure for key "${key}" (${spend} tokens); ignoring`,
+        err as Error,
+      );
+    }
+  }
+}
+
 /**
  * Read the per-workspace cap from the environment (overridable seam), falling
  * back to the sane default. A non-positive / unparseable value uses the default.
@@ -162,3 +333,31 @@ export function createPublicShareWorkspaceLimiter(
     SHARE_AI_WORKSPACE_WINDOW_MS,
   );
 }
+
+/**
+ * Read the per-workspace rolling-day token budget from the environment
+ * (overridable seam), falling back to the sane default. A non-positive /
+ * unparseable value uses the default.
+ */
+export function resolveShareAiWorkspaceTokenBudget(): number {
+  const raw = Number(process.env.SHARE_AI_WORKSPACE_TOKEN_BUDGET_PER_DAY);
+  return Number.isFinite(raw) && raw > 0
+    ? Math.floor(raw)
+    : SHARE_AI_WORKSPACE_TOKEN_BUDGET_DEFAULT;
+}
+
+/**
+ * Build the per-workspace token budget from the injected RedisService (the same
+ * global ioredis client used by the request-count limiter). Tiny factory so the
+ * service constructor stays declarative and the budget stays unit-testable with
+ * a hand-rolled fake redis.
+ */
+export function createPublicShareWorkspaceTokenBudget(
+  redisService: RedisService,
+): PublicShareWorkspaceTokenBudget {
+  return new PublicShareWorkspaceTokenBudget(
+    redisService.getOrThrow(),
+    resolveShareAiWorkspaceTokenBudget(),
+    SHARE_AI_WORKSPACE_TOKEN_WINDOW_MS,
+  );
+}

apps/server/src/core/ai-chat/tools/ai-chat-tools.service.spec.ts

@@ -120,18 +120,25 @@ describe('AiChatToolsService deletePage guardrail (H4)', () => {
     const tools = await buildTools();
     const deletePage = tools.deletePage;
 
-    // The Zod input schema only allows `pageId`; parsing strips/ignores extra
-    // keys, so a permanent/force flag is never part of the validated input.
+    // The wrapped input schema (modelFriendlyInput) only allows `pageId`;
+    // validation strips/ignores extra keys, so a permanent/force flag is never
+    // part of the validated input handed to execute.
     const schema = (deletePage as unknown as { inputSchema: unknown })
       .inputSchema as {
-      parse: (v: unknown) => Record<string, unknown>;
+      validate: (
+        v: unknown,
+      ) =>
+        | { success: boolean; value?: Record<string, unknown> }
+        | Promise<{ success: boolean; value?: Record<string, unknown> }>;
     };
-    const parsed = schema.parse({
+    const result = await schema.validate({
       pageId: 'page-789',
       permanentlyDelete: true,
       forceDelete: true,
     });
 
+    expect(result.success).toBe(true);
+    const parsed = result.value as Record<string, unknown>;
     expect(parsed).toHaveProperty('pageId', 'page-789');
     expect(parsed).not.toHaveProperty('permanentlyDelete');
     expect(parsed).not.toHaveProperty('forceDelete');
@@ -207,19 +214,26 @@ describe('AiChatToolsService expanded toolset guardrails', () => {
     const tools = await buildTools();
     const transformPage = tools.transformPage;
 
-    // The Zod input schema only allows pageId/transformJs/dryRun; parsing
-    // strips unknown keys, so deleteComments can never reach the client.
+    // The wrapped input schema only allows pageId/transformJs/dryRun;
+    // validation strips unknown keys, so deleteComments can never reach the
+    // client.
     const schema = (transformPage as unknown as { inputSchema: unknown })
       .inputSchema as {
-      parse: (v: unknown) => Record<string, unknown>;
+      validate: (
+        v: unknown,
+      ) =>
+        | { success: boolean; value?: Record<string, unknown> }
+        | Promise<{ success: boolean; value?: Record<string, unknown> }>;
     };
-    const parsed = schema.parse({
+    const result = await schema.validate({
       pageId: 'p',
       transformJs: '(d)=>d',
       dryRun: true,
       deleteComments: true,
     });
 
+    expect(result.success).toBe(true);
+    const parsed = result.value as Record<string, unknown>;
     expect(parsed).toHaveProperty('pageId', 'p');
     expect(parsed).not.toHaveProperty('deleteComments');
   });
@@ -395,3 +409,95 @@ describe('AiChatToolsService node-arg JSON-string coercion', () => {
     expect(updatePageJsonCalls).toHaveLength(0);
   });
 });
+
+/**
+ * Model-friendly tool-call validation (#190): when the model drops a required
+ * `pageId` in a parallel/batch tool call, the built-in input schema must return
+ * a CLEAR, actionable message (naming the parameter, reminding it not to drop
+ * ids in batches) instead of zod's raw "expected string, received undefined" —
+ * while a valid call still validates. This is wired centrally via
+ * modelFriendlyInput, so it applies to every in-app tool; createComment (the
+ * tool from the bug report) and a sharedTool-built tool (getPage's sibling
+ * getOutline) are exercised here end-to-end through forUser().
+ */
+describe('AiChatToolsService model-friendly input validation (#190)', () => {
+  const fakeClient: Partial<DocmostClientLike> = {};
+  const tokenServiceStub = {
+    generateAccessToken: jest.fn().mockResolvedValue('access-token'),
+    generateCollabToken: jest.fn().mockResolvedValue('collab-token'),
+  };
+  let service: AiChatToolsService;
+
+  beforeEach(() => {
+    jest.spyOn(loader, 'loadDocmostMcp').mockResolvedValue(
+      mockLoaded(function () {
+        return fakeClient as DocmostClientLike;
+      } as unknown as loader.DocmostClientCtor),
+    );
+    service = new AiChatToolsService(
+      tokenServiceStub as never,
+      {} as never,
+      {} as never,
+      {} as never,
+      {} as never,
+    );
+  });
+
+  afterEach(() => jest.restoreAllMocks());
+
+  function buildTools() {
+    return service.forUser(
+      { id: 'user-1', email: 'u@example.com', workspaceId: 'ws-1' } as never,
+      'session-1',
+      'ws-1',
+      'chat-1',
+    );
+  }
+
+  // The AI SDK Schema produced by modelFriendlyInput exposes `validate`.
+  type ValidatableSchema = {
+    validate: (
+      v: unknown,
+    ) =>
+      | { success: boolean; value?: unknown; error?: Error }
+      | Promise<{ success: boolean; value?: unknown; error?: Error }>;
+  };
+  const inputSchemaOf = (t: unknown) =>
+    (t as { inputSchema: unknown }).inputSchema as ValidatableSchema;
+
+  it('createComment: a dropped pageId yields a clear, model-actionable message', async () => {
+    const tools = await buildTools();
+    // The exact failing shape from the bug report's second parallel batch:
+    // content + selection, but pageId silently dropped.
+    const result = await inputSchemaOf(tools.createComment).validate({
+      content: 'A remark',
+      selection: 'титановый проводник',
+    });
+    expect(result.success).toBe(false);
+    expect(result.error?.message).toContain('parameter "pageId": missing (required)');
+    expect(result.error?.message).toContain('parallel/batch tool calls');
+    // Not the raw zod text the model previously received.
+    expect(result.error?.message).not.toContain('received undefined');
+  });
+
+  it('createComment: a valid call with pageId validates successfully', async () => {
+    const tools = await buildTools();
+    const result = await inputSchemaOf(tools.createComment).validate({
+      pageId: '019efe44-0000-0000-0000-000000000000',
+      content: 'A remark',
+      selection: 'титановый проводник',
+    });
+    expect(result.success).toBe(true);
+    expect(result.value).toMatchObject({
+      pageId: '019efe44-0000-0000-0000-000000000000',
+      content: 'A remark',
+    });
+  });
+
+  it('sharedTool-built tools (getOutline) also get the friendly message on a dropped pageId', async () => {
+    const tools = await buildTools();
+    const result = await inputSchemaOf(tools.getOutline).validate({});
+    expect(result.success).toBe(false);
+    expect(result.error?.message).toContain('parameter "pageId": missing (required)');
+  });
+});

apps/server/src/core/ai-chat/tools/ai-chat-tools.service.ts

@@ -15,6 +15,7 @@ import {
 } from './docmost-client.loader';
 import { resolveCurrentPageResult } from './current-page.util';
 import { parseNodeArg } from './parse-node-arg';
+import { modelFriendlyInput } from './model-friendly-input';
 
 /**
  * Per-user, per-request adapter that exposes Docmost READ operations to the
@@ -102,9 +103,13 @@ export class AiChatToolsService {
     ): Tool =>
       tool({
         description: spec.description,
-        inputSchema: spec.buildShape
-          ? z.object(spec.buildShape(z) as z.ZodRawShape)
-          : z.object({}),
+        // Wrap via modelFriendlyInput so a dropped/invalid parameter (e.g. a
+        // pageId omitted in a parallel batch, #190) yields a clear, actionable
+        // tool error instead of zod's raw text. No-arg specs still get an empty
+        // object schema.
+        inputSchema: modelFriendlyInput(
+          spec.buildShape ? (spec.buildShape(z) as z.ZodRawShape) : {},
+        ),
         execute,
       });
 
@@ -118,7 +123,7 @@ export class AiChatToolsService {
           'and entities), not a full sentence. If the first results look weak ' +
           'or incomplete, search again with different wording or synonyms ' +
           'before answering.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           query: z.string().describe('The search query.'),
           limit: z
             .number()
@@ -227,7 +232,7 @@ export class AiChatToolsService {
           '"the current page", or "here" refers to. Returns the page id and title, ' +
           'or null if the user is not currently on a page. Call this first whenever ' +
           'the user refers to the current page without giving an explicit id.',
-        inputSchema: z.object({}),
+        inputSchema: modelFriendlyInput({}),
         execute: async () => resolveCurrentPageResult(openedPage),
       }),
 
@@ -235,7 +240,7 @@ export class AiChatToolsService {
         description:
           'Fetch a single page as Markdown by its page id. Returns the page ' +
           'title and its Markdown content.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id (or slugId) of the page.'),
         }),
         execute: async ({ pageId }) => {
@@ -259,7 +264,7 @@ export class AiChatToolsService {
           'Create a new page with a Markdown body in a space, optionally under ' +
           'a parent page. Returns the new page id and title. Reversible: a page ' +
           'can be moved to trash later.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           title: z.string().describe('The title of the new page.'),
           content: z
             .string()
@@ -294,7 +299,7 @@ export class AiChatToolsService {
         description:
           "Replace a page's body with new Markdown content (and optionally its " +
           'title). Reversible: the previous version is kept in page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to update.'),
           content: z.string().describe('The new page body as Markdown.'),
           title: z
@@ -316,7 +321,7 @@ export class AiChatToolsService {
         description:
           "Rename a page (change its title only; the body is untouched). " +
           'Reversible: rename back at any time.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to rename.'),
           title: z.string().describe('The new title.'),
         }),
@@ -331,7 +336,7 @@ export class AiChatToolsService {
         description:
           'Move a page under a new parent page, or to the space root when no ' +
           'parent is given. Reversible: move it back at any time.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to move.'),
           parentPageId: z
             .string()
@@ -353,7 +358,7 @@ export class AiChatToolsService {
         description:
           'Move a page to the trash (SOFT delete only — fully reversible; the ' +
           'page can be restored from trash). This NEVER permanently deletes.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to move to trash.'),
         }),
         // GUARDRAIL (§14 H4): the only field ever passed to the client is
@@ -379,7 +384,7 @@ export class AiChatToolsService {
           '"selection not found" error, retry with a corrected EXACT selection ' +
           'copied verbatim from a single paragraph/block. Reversible via the ' +
           'comment UI.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to comment on.'),
           content: z.string().describe('The comment body as Markdown.'),
           selection: z
@@ -428,7 +433,7 @@ export class AiChatToolsService {
         description:
           'Resolve or reopen a top-level comment thread (reversible — toggle ' +
           'the resolved flag). Only top-level comments can be resolved.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           commentId: z
             .string()
             .describe('The id of the top-level comment to resolve/reopen.'),
@@ -460,7 +465,7 @@ export class AiChatToolsService {
           'List the most recent pages, optionally scoped to a single space. ' +
           'Returns a bounded list (default 50, max 100). Pass tree:true (with ' +
           "spaceId) to instead get the space's full page hierarchy as a nested tree.",
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           spaceId: z
             .string()
             .optional()
@@ -488,7 +493,7 @@ export class AiChatToolsService {
           'List sidebar pages for a space. With no pageId, returns the ' +
           "space's ROOT pages; with a pageId, returns that page's direct " +
           'CHILDREN.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           spaceId: z.string().describe('The id of the space.'),
           pageId: z
             .string()
@@ -520,7 +525,7 @@ export class AiChatToolsService {
         description:
           'Read a table as a matrix of cell texts (plus a parallel cellIds ' +
           'matrix so cells can be addressed for rich edits).',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           tableRef: z
             .string()
@@ -536,7 +541,7 @@ export class AiChatToolsService {
       listComments: tool({
         description:
           'List all comments on a page (content as Markdown).',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
         }),
         execute: async ({ pageId }) => await client.listComments(pageId),
@@ -544,7 +549,7 @@ export class AiChatToolsService {
 
       getComment: tool({
         description: 'Fetch a single comment by id (content as Markdown).',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           commentId: z.string().describe('The id of the comment.'),
         }),
         execute: async ({ commentId }) => await client.getComment(commentId),
@@ -554,7 +559,7 @@ export class AiChatToolsService {
         description:
           'Find new comments across a space (optionally scoped to a subtree) ' +
           'created after a given timestamp.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           spaceId: z.string().describe('The id of the space to scan.'),
           since: z
             .string()
@@ -586,7 +591,7 @@ export class AiChatToolsService {
         description:
           'Fetch a single page-history version including its lossless ' +
           'ProseMirror content.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           historyId: z.string().describe('The id of the history version.'),
         }),
         execute: async ({ historyId }) =>
@@ -604,7 +609,7 @@ export class AiChatToolsService {
           'Export a page to a single self-contained Docmost-flavoured ' +
           'Markdown file (meta + body + comment threads). Lossless round-trip ' +
           'with importPageMarkdown.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to export.'),
         }),
         execute: async ({ pageId }) => {
@@ -630,7 +635,7 @@ export class AiChatToolsService {
           '{"type":"text","text":"x","marks":[{"type":"bold"}]}. The node arg ' +
           'may be a JSON object or a JSON string (both accepted). Reversible: ' +
           'the previous version is kept in page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           nodeId: z
             .string()
@@ -663,7 +668,7 @@ export class AiChatToolsService {
           '{"type":"text","text":"x","marks":[{"type":"bold"}]}. The node arg ' +
           'may be a JSON object or a JSON string (both accepted). Reversible ' +
           'via page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           node: z
             .any()
@@ -722,7 +727,7 @@ export class AiChatToolsService {
           'object or a JSON string (both accepted). Omit content for a ' +
           'title-only update. Reversible: the previous version is kept in page ' +
           'history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to update.'),
           content: z
             .any()
@@ -753,7 +758,7 @@ export class AiChatToolsService {
         description:
           'Insert a row of plain-text cells into a table. Reversible via ' +
           'page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           tableRef: z
             .string()
@@ -772,7 +777,7 @@ export class AiChatToolsService {
       tableDeleteRow: tool({
         description:
           'Delete a table row at a 0-based index. Reversible via page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           tableRef: z
             .string()
@@ -787,7 +792,7 @@ export class AiChatToolsService {
         description:
           'Set the plain-text content of a table cell at [row, col] (0-based). ' +
           'Reversible via page history.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page.'),
           tableRef: z
             .string()
@@ -817,7 +822,7 @@ export class AiChatToolsService {
           'Make a page PUBLICLY accessible and return its public URL. ' +
           'Reversible via unsharePage. Only share when the user explicitly ' +
           'asked, since this exposes the page to anyone with the link.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to share.'),
           searchIndexing: z
             .boolean()
@@ -844,7 +849,7 @@ export class AiChatToolsService {
           "page's ProseMirror document for complex/scripted rewrites. dryRun " +
           '(default true) previews a diff WITHOUT writing; set dryRun:false to ' +
           'apply. Reversible: applying creates a new page-history snapshot.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z.string().describe('The id of the page to transform.'),
           transformJs: z
             .string()

apps/server/src/core/ai-chat/tools/model-friendly-input.spec.ts

@@ -0,0 +1,101 @@
+import { z } from 'zod';
+import {
+  modelFriendlyInput,
+  buildModelFriendlyMessage,
+} from './model-friendly-input';
+
+/**
+ * Unit tests for the centralized in-app tool input wrapper (#190). A dropped or
+ * invalid parameter must surface a clear, model-actionable message (naming the
+ * parameter and reminding the model not to drop ids in parallel batches), while
+ * a valid call validates cleanly and strips unknown keys — and the advertised
+ * JSON Schema keeps the unchanged required/description contract.
+ */
+describe('modelFriendlyInput', () => {
+  // Mirrors createComment's shape: pageId is the required id the model drops in
+  // parallel batches; selection is optional with a min length.
+  const shape = {
+    pageId: z.string().describe('The id of the page to comment on.'),
+    content: z.string().describe('The comment body as Markdown.'),
+    selection: z.string().min(1).max(250).optional(),
+  };
+
+  // Loose return type: the AI SDK ValidationResult is a discriminated union, but
+  // these tests assert on both branches, so a flat optional shape is simpler.
+  async function validate(
+    value: unknown,
+  ): Promise<{ success: boolean; value?: unknown; error?: Error }> {
+    const schema = modelFriendlyInput(shape);
+    return await schema.validate!(value);
+  }
+
+  it('rejects a dropped required pageId with a clear, actionable message', async () => {
+    const result = await validate({
+      content: 'Looks off here',
+      selection: 'титановый проводник',
+    });
+    expect(result.success).toBe(false);
+    const msg = result.error?.message ?? '';
+    // Names the dropped parameter...
+    expect(msg).toContain('parameter "pageId": missing (required)');
+    // ...and gives an explicit, non-raw instruction (not zod's raw text).
+    expect(msg).toContain('parallel/batch tool calls');
+    expect(msg).not.toContain('expected string, received undefined');
+  });
+
+  it('distinguishes a present-but-invalid parameter from a missing one', async () => {
+    // selection is present but too short (invalid), pageId is missing.
+    const result = await validate({ content: 'x', selection: '' });
+    expect(result.success).toBe(false);
+    const msg = result.error?.message ?? '';
+    expect(msg).toContain('parameter "pageId": missing (required)');
+    expect(msg).toContain('parameter "selection": invalid');
+  });
+
+  it('accepts a valid call and strips unknown keys from the validated value', async () => {
+    const result = await validate({
+      pageId: 'page-1',
+      content: 'A comment',
+      selection: 'anchor text',
+      bogus: true,
+    });
+    expect(result.success).toBe(true);
+    if (!result.success) throw new Error('expected success');
+    expect(result.value).toEqual({
+      pageId: 'page-1',
+      content: 'A comment',
+      selection: 'anchor text',
+    });
+    expect(result.value).not.toHaveProperty('bogus');
+  });
+
+  it('preserves the required/description contract in the advertised JSON Schema', async () => {
+    const schema = modelFriendlyInput(shape);
+    const json = (await schema.jsonSchema) as {
+      required?: string[];
+      properties?: Record<string, { description?: string }>;
+    };
+    // pageId + content stay required; selection stays optional.
+    expect(json.required).toEqual(expect.arrayContaining(['pageId', 'content']));
+    expect(json.required).not.toContain('selection');
+    expect(json.properties?.pageId.description).toBe(
+      'The id of the page to comment on.',
+    );
+  });
+
+  it('handles a no-arg tool (empty shape) without error', async () => {
+    const schema = modelFriendlyInput({});
+    const result = await schema.validate!({});
+    expect(result.success).toBe(true);
+  });
+});
+
+describe('buildModelFriendlyMessage', () => {
+  it('falls back to a generic message when issues carry an empty path', () => {
+    // safeParse on a non-object yields a root-level issue (empty path).
+    const error = z.object({ a: z.string() }).safeParse('not-an-object');
+    if (error.success) throw new Error('expected failure');
+    const msg = buildModelFriendlyMessage(error.error, 'not-an-object');
+    expect(msg).toContain('parameter "input"');
+  });
+});

apps/server/src/core/ai-chat/tools/model-friendly-input.ts

@@ -0,0 +1,93 @@
+import { jsonSchema, type Schema } from 'ai';
+import type { JSONSchema7 } from '@ai-sdk/provider';
+import { z } from 'zod';
+
+/**
+ * Centralized input-schema wrapper for every in-app AI-chat tool.
+ *
+ * THE PROBLEM (#190): when the model issues PARALLEL / batch tool calls it
+ * sometimes drops an "obvious" repeated required argument (typically `pageId`)
+ * from some of the calls. zod v4 correctly rejects the missing value, but the
+ * AI SDK forwards zod's RAW message ("Invalid input: expected string, received
+ * undefined") straight back to the model, which is not actionable — the model
+ * cannot tell WHICH parameter it dropped or that it must re-send it.
+ *
+ * THE FIX: keep the exact same validation, but replace the raw zod text with a
+ * model-friendly message that names every problematic parameter and tells the
+ * model to re-issue the call with all required parameters present. We do NOT
+ * guess/backfill the value (a silently-assumed "current page" could comment on
+ * the wrong page — cf. #159); the model is simply told to retry correctly.
+ *
+ * HOW IT WORKS: we build the tool's JSON Schema from the zod shape via
+ * `z.toJSONSchema(..., { target: 'draft-7' })` (so the advertised contract —
+ * `required` / `description` / field constraints — is unchanged) and hand the
+ * AI SDK a custom `validate` that runs `z.object(shape).safeParse(value)`. On
+ * failure the AI SDK wraps our returned `Error` in `InvalidToolInputError`, so
+ * our clear text is what reaches the model as the tool error.
+ */
+export function modelFriendlyInput<T extends z.ZodRawShape>(
+  shape: T,
+): Schema<z.output<z.ZodObject<T>>> {
+  const objectSchema = z.object(shape);
+  // draft-07 keeps required/description/constraints intact, matching what the
+  // model already saw — the tool contract does not change.
+  const json = z.toJSONSchema(objectSchema, {
+    target: 'draft-7',
+  }) as JSONSchema7;
+
+  return jsonSchema<z.output<z.ZodObject<T>>>(json, {
+    validate: (value) => {
+      const result = objectSchema.safeParse(value);
+      if (result.success) {
+        return { success: true, value: result.data };
+      }
+      return {
+        success: false,
+        error: new Error(buildModelFriendlyMessage(result.error, value)),
+      };
+    },
+  });
+}
+
+/**
+ * Turn a zod validation failure into a clear, model-actionable message naming
+ * each problematic parameter (and whether it is missing vs. invalid), plus an
+ * explicit reminder not to drop required ids in parallel/batch tool calls.
+ */
+export function buildModelFriendlyMessage(
+  error: z.ZodError,
+  value: unknown,
+): string {
+  const seen = new Set<string>();
+  const parts: string[] = [];
+  for (const issue of error.issues) {
+    const name = issue.path.length ? issue.path.map(String).join('.') : 'input';
+    // A parameter the model omitted entirely reads as `undefined` at its path;
+    // anything else is present-but-invalid (wrong type, too short, etc.).
+    const missing = valueAtPath(value, issue.path) === undefined;
+    const part = `parameter "${name}": ${missing ? 'missing (required)' : 'invalid'}`;
+    if (seen.has(part)) continue;
+    seen.add(part);
+    parts.push(part);
+  }
+  if (parts.length === 0) {
+    // Defensive: a ZodError always has issues, but never emit an empty list.
+    parts.push('input: invalid');
+  }
+  return (
+    `Invalid input for this tool — ${parts.join('; ')}. ` +
+    'Re-issue the call with EVERY required parameter present and valid. ' +
+    "Do not drop ids like pageId, even when making parallel/batch tool calls — " +
+    'each tool call must carry its own pageId.'
+  );
+}
+
+/** Read the value at a zod issue path; returns undefined if any hop is absent. */
+function valueAtPath(value: unknown, path: ReadonlyArray<PropertyKey>): unknown {
+  let current: unknown = value;
+  for (const key of path) {
+    if (current === null || typeof current !== 'object') return undefined;
+    current = (current as Record<PropertyKey, unknown>)[key];
+  }
+  return current;
+}

apps/server/src/core/ai-chat/tools/public-share-chat-tools.service.ts

@@ -5,6 +5,7 @@ import { ShareService } from '../../share/share.service';
 import { SearchService } from '../../search/search.service';
 import { PageRepo } from '@docmost/db/repos/page/page.repo';
 import { jsonToMarkdown } from '../../../collaboration/collaboration.util';
+import { modelFriendlyInput } from './model-friendly-input';
 
 /**
  * Isolated, READ-ONLY toolset for the ANONYMOUS public-share assistant.
@@ -52,7 +53,7 @@ export class PublicShareChatToolsService {
           '(key terms and entities), not a full sentence. If the first ' +
           'results look weak, search again with different wording before ' +
           'answering. Only pages inside this share are ever returned.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           query: z.string().describe('The search query.'),
           limit: z
             .number()
@@ -87,7 +88,7 @@ export class PublicShareChatToolsService {
           'Markdown, by its page id. Returns the page title and its Markdown ' +
           'content. Only pages inside this share can be read; reading any ' +
           'other page fails.',
-        inputSchema: z.object({
+        inputSchema: modelFriendlyInput({
           pageId: z
             .string()
             .describe('The id (or slugId) of a page within this share.'),
@@ -142,7 +143,7 @@ export class PublicShareChatToolsService {
           'List the pages (titles + ids) that make up THIS published ' +
           'documentation share, so you can orient yourself before reading or ' +
           'searching. Only pages inside this share are listed.',
-        inputSchema: z.object({}),
+        inputSchema: modelFriendlyInput({}),
         execute: async () => {
           // Reuse the same share-tree logic the public /shares/tree route uses:
           // it validates the share + workspace, excludes restricted subtrees,

apps/server/src/core/page/constants/temporary-note.constants.ts

@@ -0,0 +1,5 @@
+// Default lifetime for a temporary note, in HOURS, used when the workspace has
+// no `temporaryNoteHours` configured (NULL). Mirrors the trash-cleanup
+// DEFAULT_RETENTION_DAYS fallback. After this many hours a temporary note is
+// auto-moved to trash unless it was made permanent first.
+export const DEFAULT_TEMPORARY_NOTE_HOURS = 24;

apps/server/src/core/page/dto/create-page.dto.ts

@@ -1,4 +1,5 @@
 import {
+  IsBoolean,
   IsIn,
   IsOptional,
   IsString,
@@ -32,4 +33,10 @@ export class CreatePageDto {
   @Transform(({ value }) => value?.toLowerCase() ?? 'json')
   @IsIn(['json', 'markdown', 'html'])
   format?: ContentFormat;
+
+  // When true, create the page as a temporary note: arm its death timer
+  // (now + workspace temporaryNoteHours) at creation.
+  @IsOptional()
+  @IsBoolean()
+  temporary?: boolean;
 }

apps/server/src/core/page/page.module.ts

@@ -3,6 +3,7 @@ import { PageService } from './services/page.service';
 import { PageController } from './page.controller';
 import { PageHistoryService } from './services/page-history.service';
 import { TrashCleanupService } from './services/trash-cleanup.service';
+import { TemporaryNoteCleanupService } from './services/temporary-note-cleanup.service';
 import { BacklinkService } from './services/backlink.service';
 import { StorageModule } from '../../integrations/storage/storage.module';
 import { CollaborationModule } from '../../collaboration/collaboration.module';
@@ -16,6 +17,7 @@ import { LabelModule } from '../label/label.module';
     PageService,
     PageHistoryService,
     TrashCleanupService,
+    TemporaryNoteCleanupService,
     BacklinkService,
   ],
   exports: [PageService, PageHistoryService],

apps/server/src/core/page/services/page.service.spec.ts

@@ -2,6 +2,7 @@ import { BadRequestException } from '@nestjs/common';
 import { PageService } from './page.service';
 import { MovePageDto } from '../dto/move-page.dto';
 import { Page } from '@docmost/db/types/entity.types';
+import { DEFAULT_TEMPORARY_NOTE_HOURS } from '../constants/temporary-note.constants';
 
 // Direct instantiation with stub deps. The Test.createTestingModule form failed
 // to resolve the @InjectKysely()/@InjectQueue() tokens at compile(), and this
@@ -57,11 +58,28 @@ describe('PageService', () => {
 
       const eventEmitter = { emit: jest.fn() };
 
+      // movePage now runs the cycle-check + UPDATE inside executeTx(this.db),
+      // i.e. this.db.transaction().execute(fn => fn(trx)). A permissive chainable
+      // Proxy stands in for the Kysely trx so the per-space advisory-lock
+      // `sql``.execute(trx)` resolves; a thrown BadRequestException still
+      // propagates out of the transaction unchanged.
+      const trxStub: any = new Proxy(function () {}, {
+        get: (_t, p) =>
+          p === 'then'
+            ? undefined
+            : p === 'execute' || p === 'executeTakeFirst'
+              ? () => Promise.resolve([])
+              : () => trxStub,
+      });
+      const db = {
+        transaction: () => ({ execute: (fn: any) => fn(trxStub) }),
+      };
+
       const svc = new PageService(
         pageRepo as any, // pageRepo
         {} as any, // pagePermissionRepo
         {} as any, // attachmentRepo
-        {} as any, // db
+        db as any, // db
         {} as any, // storageService
         {} as any, // attachmentQueue
         {} as any, // aiQueue
@@ -268,9 +286,23 @@ describe('PageService', () => {
           }),
           updatePage: jest.fn().mockResolvedValue({ numUpdatedRows: 1n }),
         };
+        // movePage now runs the cycle-check + UPDATE inside executeTx(this.db),
+        // which calls this.db.transaction().execute(fn => fn(trx)). A permissive
+        // chainable Proxy stands in for the Kysely trx so the per-space
+        // advisory-lock `sql``.execute(trx)` resolves and updatePage receives it.
+        const trxStub: any = new Proxy(function () {}, {
+          get: (_t, p) =>
+            p === 'then'
+              ? undefined
+              : p === 'execute' || p === 'executeTakeFirst'
+                ? () => Promise.resolve([])
+                : () => trxStub,
+        });
         const svc = makeSvc({
           pageRepo,
-          db: {} as any,
+          db: {
+            transaction: () => ({ execute: (fn: any) => fn(trxStub) }),
+          } as any,
         });
         // Legitimate move: destination ancestors do NOT include the moved page.
         jest
@@ -389,4 +421,79 @@ describe('PageService', () => {
       });
     });
   });
+
+  describe('create() temporary deadline (#201)', () => {
+    // db stub for the workspaces.temporaryNoteHours lookup:
+    // selectFrom('workspaces').select(['temporaryNoteHours']).where(...).executeTakeFirst()
+    const makeDb = (workspaceRow: any) => {
+      const builder: any = {
+        selectFrom: jest.fn(() => builder),
+        select: jest.fn(() => builder),
+        where: jest.fn(() => builder),
+        executeTakeFirst: jest.fn().mockResolvedValue(workspaceRow),
+      };
+      return builder;
+    };
+
+    const makeGeneralQueue = () =>
+      ({ add: jest.fn().mockReturnValue({ catch: jest.fn() }) }) as any;
+
+    const run = async (dto: any, workspaceRow: any) => {
+      const pageRepo = {
+        insertPage: jest.fn().mockResolvedValue({ id: 'p1' }),
+      };
+      const db = makeDb(workspaceRow);
+      const svc = new PageService(
+        pageRepo as any, // pageRepo
+        {} as any, // pagePermissionRepo
+        {} as any, // attachmentRepo
+        db as any, // db
+        {} as any, // storageService
+        {} as any, // attachmentQueue
+        {} as any, // aiQueue
+        makeGeneralQueue(), // generalQueue
+        {} as any, // eventEmitter
+        {} as any, // collaborationGateway
+        {} as any, // watcherService
+        {} as any, // transclusionService
+      );
+      // nextPagePosition runs a real db query; stub it out.
+      jest.spyOn(svc, 'nextPagePosition').mockResolvedValue('a0' as any);
+      await svc.create('u1', 'w1', dto, undefined);
+      return { payload: pageRepo.insertPage.mock.calls[0][0], db };
+    };
+
+    afterEach(() => jest.useRealTimers());
+
+    it('freezes temporaryExpiresAt at now + workspace hours when temporary', async () => {
+      jest.useFakeTimers().setSystemTime(new Date('2026-06-26T00:00:00.000Z'));
+      const { payload } = await run(
+        { title: 't', spaceId: 's1', temporary: true },
+        { temporaryNoteHours: 5 },
+      );
+      expect(payload.temporaryExpiresAt).toEqual(
+        new Date(Date.now() + 5 * 60 * 60 * 1000),
+      );
+    });
+
+    it('falls back to DEFAULT_TEMPORARY_NOTE_HOURS when the workspace hours are null', async () => {
+      jest.useFakeTimers().setSystemTime(new Date('2026-06-26T00:00:00.000Z'));
+      const { payload } = await run(
+        { title: 't', spaceId: 's1', temporary: true },
+        { temporaryNoteHours: null },
+      );
+      expect(payload.temporaryExpiresAt).toEqual(
+        new Date(Date.now() + DEFAULT_TEMPORARY_NOTE_HOURS * 60 * 60 * 1000),
+      );
+    });
+
+    it('leaves temporaryExpiresAt undefined and skips the workspace lookup for a non-temporary page', async () => {
+      const { payload, db } = await run(
+        { title: 't', spaceId: 's1' },
+        { temporaryNoteHours: 5 },
+      );
+      expect(payload.temporaryExpiresAt).toBeUndefined();
+      expect(db.selectFrom).not.toHaveBeenCalled();
+    });
+  });
 });

apps/server/src/core/page/services/page.service.ts

@@ -15,13 +15,13 @@ import {
   executeWithCursorPagination,
 } from '@docmost/db/pagination/cursor-pagination';
 import { InjectKysely } from 'nestjs-kysely';
-import { KyselyDB } from '@docmost/db/types/kysely.types';
+import { KyselyDB, KyselyTransaction } from '@docmost/db/types/kysely.types';
 import { generateJitteredKeyBetween } from 'fractional-indexing-jittered';
 import { MovePageDto } from '../dto/move-page.dto';
 import { shapeSidebarPagesTree } from './sidebar-pages-tree.util';
 import { generateSlugId } from '../../../common/helpers';
 import { getPageTitle } from '../../../common/helpers';
-import { executeTx } from '@docmost/db/utils';
+import { dbOrTx, executeTx } from '@docmost/db/utils';
 import { AttachmentRepo } from '@docmost/db/repos/attachment/attachment.repo';
 import { v7 as uuid7 } from 'uuid';
 import {
@@ -61,6 +61,24 @@ import {
   AuthProvenanceData,
   agentSourceFields,
 } from '../../../common/decorators/auth-provenance.decorator';
+import { DEFAULT_TEMPORARY_NOTE_HOURS } from '../constants/temporary-note.constants';
+
+// Hard upper bound on how deep the recursive page-tree CTEs (ancestor /
+// descendant traversals) may walk. Real page trees are only a handful of levels
+// deep, so this cap never truncates a legitimate result; it purely defends the
+// recursive CTEs against runaway iteration if a parent/child cycle ever exists
+// in the data (e.g. one slipped in before the move guard, #207 #8). Without it a
+// cycle makes `withRecursive` loop forever (hang / statement timeout), and the
+// move guard itself calls one of these CTEs — so a cycle would disable the very
+// guard meant to prevent it. Each CTE carries a depth counter and stops here.
+const MAX_PAGE_TREE_DEPTH = 10_000;
+
+// Advisory-lock namespace (the first key of pg_advisory_xact_lock) used to
+// serialize concurrent page moves within a single space so the cycle check and
+// the move UPDATE stay atomic (see movePage, #207 #7). A dedicated namespace
+// constant keeps these locks from colliding with any other advisory lock; the
+// second key is hashtext(spaceId). Fits a signed int4 ('page' in ASCII).
+const PAGE_MOVE_LOCK_NAMESPACE = 0x70616765;
 
 @Injectable()
 export class PageService {
@@ -123,6 +141,20 @@ export class PageService {
       parentPageId = parentPage.id;
     }
 
+    // Freeze the death timer here so later changes to the workspace setting
+    // never reschedule existing temporary notes. NULL => permanent page.
+    let temporaryExpiresAt: Date | undefined;
+    if (createPageDto.temporary) {
+      const workspace = await this.db
+        .selectFrom('workspaces')
+        .select(['temporaryNoteHours'])
+        .where('id', '=', workspaceId)
+        .executeTakeFirst();
+      const hours =
+        workspace?.temporaryNoteHours ?? DEFAULT_TEMPORARY_NOTE_HOURS;
+      temporaryExpiresAt = new Date(Date.now() + hours * 60 * 60 * 1000);
+    }
+
     let content = undefined;
     let textContent = undefined;
     let ydoc = undefined;
@@ -155,6 +187,7 @@ export class PageService {
       // (creatorId/lastUpdatedById); these only annotate the source. A normal
       // user request leaves the column default ('user').
       ...agentSourceFields(provenance, 'lastUpdatedSource', 'lastUpdatedAiChatId'),
+      temporaryExpiresAt,
       content,
       textContent,
       ydoc,
@@ -339,6 +372,7 @@ export class PageService {
         'spaceId',
         'creatorId',
         'isTemplate',
+        'temporaryExpiresAt',
         'deletedAt',
       ])
       .select((eb) => this.pageRepo.withHasChildren(eb))
@@ -601,7 +635,13 @@ export class PageService {
       slugIdMap.set(entry.oldSlugId, entry);
     }
 
-    const attachmentMap = new Map<string, ICopyPageAttachment>();
+    // Keyed by old attachmentId. A single attachment can be referenced by more
+    // than one page in the copied subtree (e.g. a block copy-pasted into a child
+    // page keeps the same attachmentId). Each referencing page needs its own
+    // fresh attachment id / row / blob copy, so the value is a LIST of copy
+    // entries rather than a single one — otherwise the last page's entry would
+    // clobber the others and their images would 404 in the copies (#206 attach-1).
+    const attachmentMap = new Map<string, ICopyPageAttachment[]>();
 
     const insertablePages: InsertablePage[] = await Promise.all(
       pages.map(async (page) => {
@@ -617,12 +657,14 @@ export class PageService {
           attachmentIds.forEach((attachmentId: string) => {
             const newPageId = pageFromMap.newPageId;
             const newAttachmentId = uuid7();
-            attachmentMap.set(attachmentId, {
+            const existingEntries = attachmentMap.get(attachmentId) ?? [];
+            existingEntries.push({
               newPageId: newPageId,
               oldPageId: page.id,
               oldAttachmentId: attachmentId,
               newAttachmentId: newAttachmentId,
             });
+            attachmentMap.set(attachmentId, existingEntries);
 
             prosemirrorDoc.descendants((node: PMNode) => {
               if (isAttachmentNode(node.type.name)) {
@@ -819,51 +861,53 @@ export class PageService {
         .execute();
 
       for (const attachment of attachments) {
-        try {
-          const pageAttachment = attachmentMap.get(attachment.id);
-
-          // make sure the copied attachment belongs to the page it was copied from
-          if (attachment.pageId !== pageAttachment.oldPageId) {
-            continue;
-          }
-
-          const newAttachmentId = pageAttachment.newAttachmentId;
-
-          const newPageId = pageAttachment.newPageId;
-
-          const newPathFile = attachment.filePath.replace(
-            attachment.id,
-            newAttachmentId,
-          );
-
+        // One source attachment may need to be copied for several destination
+        // pages (it is referenced by more than one page in the subtree). Copy a
+        // distinct blob + row for every referencing page so each copy resolves
+        // (#206 attach-1). The old per-page ownership guard is gone: when the
+        // same attachmentId is shared, only one page would ever match the row's
+        // pageId, silently dropping the other copies.
+        const pageAttachments = attachmentMap.get(attachment.id) ?? [];
+        for (const pageAttachment of pageAttachments) {
           try {
-            await this.storageService.copy(attachment.filePath, newPathFile);
+            const newAttachmentId = pageAttachment.newAttachmentId;
 
-            await this.db
-              .insertInto('attachments')
-              .values({
-                id: newAttachmentId,
-                type: attachment.type,
-                filePath: newPathFile,
-                fileName: attachment.fileName,
-                fileSize: attachment.fileSize,
-                mimeType: attachment.mimeType,
-                fileExt: attachment.fileExt,
-                creatorId: attachment.creatorId,
-                workspaceId: attachment.workspaceId,
-                pageId: newPageId,
-                spaceId: spaceId,
-              })
-              .execute();
-          } catch (err) {
-            this.logger.error(
-              `Duplicate page: failed to copy attachment ${attachment.id}`,
-              err,
+            const newPageId = pageAttachment.newPageId;
+
+            const newPathFile = attachment.filePath.replace(
+              attachment.id,
+              newAttachmentId,
             );
-            // Continue with other attachments even if one fails
+
+            try {
+              await this.storageService.copy(attachment.filePath, newPathFile);
+
+              await this.db
+                .insertInto('attachments')
+                .values({
+                  id: newAttachmentId,
+                  type: attachment.type,
+                  filePath: newPathFile,
+                  fileName: attachment.fileName,
+                  fileSize: attachment.fileSize,
+                  mimeType: attachment.mimeType,
+                  fileExt: attachment.fileExt,
+                  creatorId: attachment.creatorId,
+                  workspaceId: attachment.workspaceId,
+                  pageId: newPageId,
+                  spaceId: spaceId,
+                })
+                .execute();
+            } catch (err) {
+              this.logger.error(
+                `Duplicate page: failed to copy attachment ${attachment.id}`,
+                err,
+              );
+              // Continue with other attachments even if one fails
+            }
+          } catch (err) {
+            this.logger.error(err);
           }
-        } catch (err) {
-          this.logger.error(err);
         }
       }
     }
@@ -915,34 +959,61 @@ export class PageService {
       }
     }
 
-    // Server-side cycle guard: a page may not be moved into itself or into any
-    // page within its own subtree. Without this, an MCP/REST/agent caller (or a
-    // fast drag racing the client check) could persist a cycle and broadcast it.
-    // Only relevant when re-parenting under a concrete parent; moving to root
-    // (parentPageId null/undefined) can never create a cycle.
-    if (dto.parentPageId) {
-      if (dto.parentPageId === dto.pageId) {
-        throw new BadRequestException('Cannot move a page into its own subtree');
-      }
-      // Walk the destination parent's ancestor chain (reusing the breadcrumb
-      // ancestor CTE). If the page being moved appears among those ancestors,
-      // the destination lives inside the moved page's subtree -> cycle.
-      const destAncestors = await this.getPageBreadCrumbs(dto.parentPageId);
-      if (destAncestors.some((ancestor) => ancestor.id === dto.pageId)) {
-        throw new BadRequestException('Cannot move a page into its own subtree');
-      }
-    }
+    // Server-side cycle guard + the move UPDATE run in ONE transaction. A page
+    // may not be moved into itself or into any page within its own subtree;
+    // without this an MCP/REST/agent caller (or a fast drag racing the client
+    // check) could persist a cycle and broadcast it. Crucially, doing the guard
+    // and the write as two separate, unlocked statements is a TOCTOU race: two
+    // concurrent moves ("A under B" and "B under A") can each read the same
+    // pre-write acyclic snapshot, both pass the guard, then persist
+    // A.parentPageId=B AND B.parentPageId=A — a parent/child cycle (#207 #7). A
+    // per-space advisory lock (held until COMMIT) serializes all moves within a
+    // space: the second mover blocks until the first commits and then sees the
+    // freshly written parent, so its guard rejects the cycle.
+    const updateResult = await executeTx(this.db, async (trx) => {
+      await sql`select pg_advisory_xact_lock(${sql.lit(
+        PAGE_MOVE_LOCK_NAMESPACE,
+      )}, hashtext(${movedPage.spaceId}))`.execute(trx);
 
-    const updateResult = await this.pageRepo.updatePage(
-      {
-        position: dto.position,
-        parentPageId: parentPageId,
-        // Agent-edit provenance: annotate the source on an agent move. A normal
-        // user request leaves the existing source value unchanged.
-        ...agentSourceFields(provenance, 'lastUpdatedSource', 'lastUpdatedAiChatId'),
-      },
-      dto.pageId,
-    );
+      // Only relevant when re-parenting under a concrete parent; moving to root
+      // (parentPageId null/undefined) can never create a cycle.
+      if (dto.parentPageId) {
+        if (dto.parentPageId === dto.pageId) {
+          throw new BadRequestException(
+            'Cannot move a page into its own subtree',
+          );
+        }
+        // Walk the destination parent's ancestor chain (reusing the breadcrumb
+        // ancestor CTE) inside the lock. If the page being moved appears among
+        // those ancestors, the destination lives inside the moved page's
+        // subtree -> cycle.
+        const destAncestors = await this.getPageBreadCrumbs(
+          dto.parentPageId,
+          trx,
+        );
+        if (destAncestors.some((ancestor) => ancestor.id === dto.pageId)) {
+          throw new BadRequestException(
+            'Cannot move a page into its own subtree',
+          );
+        }
+      }
+
+      return this.pageRepo.updatePage(
+        {
+          position: dto.position,
+          parentPageId: parentPageId,
+          // Agent-edit provenance: annotate the source on an agent move. A
+          // normal user request leaves the existing source value unchanged.
+          ...agentSourceFields(
+            provenance,
+            'lastUpdatedSource',
+            'lastUpdatedAiChatId',
+          ),
+        },
+        dto.pageId,
+        trx,
+      );
+    });
 
     // Guard against a phantom broadcast: if the row was concurrently deleted or
     // otherwise not updated, skip the PAGE_MOVED event so we don't replay a move
@@ -981,8 +1052,8 @@ export class PageService {
     });
   }
 
-  async getPageBreadCrumbs(childPageId: string) {
-    const ancestors = await this.db
+  async getPageBreadCrumbs(childPageId: string, trx?: KyselyTransaction) {
+    const ancestors = await dbOrTx(this.db, trx)
       .withRecursive('page_ancestors', (db) =>
         db
           .selectFrom('pages')
@@ -996,6 +1067,9 @@ export class PageService {
             'spaceId',
             'deletedAt',
           ])
+          // Depth counter: bounds the walk so a parent/child cycle in the data
+          // can't make this recursive CTE loop forever (#207 #8).
+          .select(sql<number>`0`.as('depth'))
           .where('id', '=', childPageId)
           .where('deletedAt', 'is', null)
           .unionAll((exp) =>
@@ -1011,12 +1085,25 @@ export class PageService {
                 'p.spaceId',
                 'p.deletedAt',
               ])
+              .select(sql<number>`pa.depth + 1`.as('depth'))
               .innerJoin('page_ancestors as pa', 'pa.parentPageId', 'p.id')
-              .where('p.deletedAt', 'is', null),
+              .where('p.deletedAt', 'is', null)
+              .where(sql<number>`pa.depth`, '<', MAX_PAGE_TREE_DEPTH),
           ),
       )
       .selectFrom('page_ancestors')
-      .selectAll('page_ancestors')
+      // Explicit column list (not selectAll) so the internal `depth` counter
+      // never leaks into the breadcrumb result shape.
+      .select([
+        'id',
+        'slugId',
+        'title',
+        'icon',
+        'position',
+        'parentPageId',
+        'spaceId',
+        'deletedAt',
+      ])
       .select((eb) =>
         eb
           .exists(
@@ -1137,16 +1224,21 @@ export class PageService {
         db
           .selectFrom('pages')
           .select(['id'])
+          // Depth counter: bounds the walk so a parent/child cycle in the data
+          // can't make this recursive CTE loop forever (#207 #8).
+          .select(sql<number>`0`.as('depth'))
           .where('id', '=', pageId)
           .unionAll((exp) =>
             exp
               .selectFrom('pages as p')
               .select(['p.id'])
-              .innerJoin('page_descendants as pd', 'pd.id', 'p.parentPageId'),
+              .select(sql<number>`pd.depth + 1`.as('depth'))
+              .innerJoin('page_descendants as pd', 'pd.id', 'p.parentPageId')
+              .where(sql<number>`pd.depth`, '<', MAX_PAGE_TREE_DEPTH),
           ),
       )
       .selectFrom('page_descendants')
-      .selectAll()
+      .select(['id'])
       .execute();
 
     const pageIds = descendants.map((d) => d.id);

apps/server/src/core/page/services/spec/temporary-note-cleanup.service.spec.ts

@@ -0,0 +1,154 @@
+import { TemporaryNoteCleanupService } from '../temporary-note-cleanup.service';
+
+/**
+ * Chainable Kysely stub that records every `.where(...)` call so the test can
+ * assert the sweep only selects armed, expired, not-yet-trashed notes. The
+ * terminal `.execute()` resolves the configured expired rows (the batch SELECT);
+ * `.executeTakeFirst()` resolves the per-row deadline re-read done just before
+ * each `removePage`. By default the re-read reports the note as still armed and
+ * still expired (epoch deadline < now), so the sweep proceeds to delete it;
+ * tests override `reReadFirst` to simulate a concurrent "Make permanent".
+ */
+function makeDbStub(expiredRows: any[]) {
+  const whereCalls: any[][] = [];
+  const reReadFirst = jest
+    .fn()
+    .mockResolvedValue({ temporaryExpiresAt: new Date(0), deletedAt: null });
+  const builder: any = {
+    selectFrom: jest.fn(() => builder),
+    select: jest.fn(() => builder),
+    where: jest.fn((...args: any[]) => {
+      whereCalls.push(args);
+      return builder;
+    }),
+    limit: jest.fn(() => builder),
+    execute: jest.fn().mockResolvedValue(expiredRows),
+    executeTakeFirst: reReadFirst,
+  };
+  return { builder, whereCalls, reReadFirst };
+}
+
+describe('TemporaryNoteCleanupService.sweepExpiredTemporaryNotes', () => {
+  it('selects only armed, expired, not-yet-trashed notes', async () => {
+    const { builder, whereCalls } = makeDbStub([]);
+    const pageRepo = { removePage: jest.fn() } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+
+    // temporaryExpiresAt IS NOT NULL, temporaryExpiresAt < now, deletedAt IS NULL
+    const cols = whereCalls.map((c) => c[0]);
+    const ops = whereCalls.map((c) => c[1]);
+    expect(cols).toEqual([
+      'temporaryExpiresAt',
+      'temporaryExpiresAt',
+      'deletedAt',
+    ]);
+    expect(ops).toEqual(['is not', '<', 'is']);
+    // last operand is the trash filter -> null
+    expect(whereCalls[2][2]).toBeNull();
+    // The batch SELECT is capped so a large backlog is not pulled at once.
+    expect(builder.limit).toHaveBeenCalledTimes(1);
+    expect(builder.limit.mock.calls[0][0]).toBeGreaterThan(0);
+  });
+
+  it('soft-deletes each expired note via removePage, attributed to its creator', async () => {
+    const expired = [
+      { id: 'p1', creatorId: 'u1', workspaceId: 'w1' },
+      { id: 'p2', creatorId: 'u2', workspaceId: 'w1' },
+    ];
+    const { builder } = makeDbStub(expired);
+    const pageRepo = { removePage: jest.fn().mockResolvedValue(undefined) } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+
+    expect(pageRepo.removePage).toHaveBeenCalledTimes(2);
+    expect(pageRepo.removePage).toHaveBeenNthCalledWith(1, 'p1', 'u1', 'w1');
+    expect(pageRepo.removePage).toHaveBeenNthCalledWith(2, 'p2', 'u2', 'w1');
+  });
+
+  it('continues past a failing note (one bad removePage does not abort the sweep)', async () => {
+    const expired = [
+      { id: 'bad', creatorId: 'u1', workspaceId: 'w1' },
+      { id: 'good', creatorId: 'u2', workspaceId: 'w1' },
+    ];
+    const { builder } = makeDbStub(expired);
+    const pageRepo = {
+      removePage: jest
+        .fn()
+        .mockRejectedValueOnce(new Error('boom'))
+        .mockResolvedValueOnce(undefined),
+    } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await expect(
+      service.sweepExpiredTemporaryNotes(),
+    ).resolves.toBeUndefined();
+    expect(pageRepo.removePage).toHaveBeenCalledTimes(2);
+    expect(pageRepo.removePage).toHaveBeenNthCalledWith(2, 'good', 'u2', 'w1');
+  });
+
+  it('does NOT trash a note made permanent in the race window', async () => {
+    // The batch SELECT saw the note as expired, but before its turn in the loop
+    // the user clicked "Make permanent" (temporary_expires_at -> null). The
+    // deadline re-read must catch this and skip the delete so the keep wins.
+    const expired = [{ id: 'p1', creatorId: 'u1', workspaceId: 'w1' }];
+    const { builder, reReadFirst } = makeDbStub(expired);
+    reReadFirst.mockResolvedValueOnce({
+      temporaryExpiresAt: null,
+      deletedAt: null,
+    });
+    const pageRepo = { removePage: jest.fn() } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+
+    expect(reReadFirst).toHaveBeenCalledTimes(1);
+    expect(pageRepo.removePage).not.toHaveBeenCalled();
+  });
+
+  it('skips a note already trashed since the batch SELECT', async () => {
+    const expired = [{ id: 'p1', creatorId: 'u1', workspaceId: 'w1' }];
+    const { builder, reReadFirst } = makeDbStub(expired);
+    reReadFirst.mockResolvedValueOnce({
+      temporaryExpiresAt: new Date(0),
+      deletedAt: new Date(),
+    });
+    const pageRepo = { removePage: jest.fn() } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+
+    expect(pageRepo.removePage).not.toHaveBeenCalled();
+  });
+
+  it('does NOT trash a note re-armed to a future deadline in the race window', async () => {
+    // The batch SELECT saw the note as expired, but before its turn in the loop
+    // the user disarmed it and re-armed it to a fresh, still-future deadline
+    // (temporary_expires_at -> now + 1h). The deadline re-read must catch that
+    // the note is no longer expired and skip the delete so the keep wins.
+    const expired = [{ id: 'p1', creatorId: 'u1', workspaceId: 'w1' }];
+    const { builder, reReadFirst } = makeDbStub(expired);
+    reReadFirst.mockResolvedValueOnce({
+      temporaryExpiresAt: new Date(Date.now() + 60 * 60 * 1000),
+      deletedAt: null,
+    });
+    const pageRepo = { removePage: jest.fn() } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+
+    expect(reReadFirst).toHaveBeenCalledTimes(1);
+    expect(pageRepo.removePage).not.toHaveBeenCalled();
+  });
+
+  it('does nothing when no notes are expired', async () => {
+    const { builder } = makeDbStub([]);
+    const pageRepo = { removePage: jest.fn() } as any;
+    const service = new TemporaryNoteCleanupService(builder, pageRepo);
+
+    await service.sweepExpiredTemporaryNotes();
+    expect(pageRepo.removePage).not.toHaveBeenCalled();
+  });
+});

apps/server/src/core/page/services/temporary-note-cleanup.service.ts

@@ -0,0 +1,105 @@
+import { Injectable, Logger } from '@nestjs/common';
+import { Interval } from '@nestjs/schedule';
+import { InjectKysely } from 'nestjs-kysely';
+import { KyselyDB } from '@docmost/db/types/kysely.types';
+import { PageRepo } from '@docmost/db/repos/page/page.repo';
+
+/**
+ * Background sweeper for temporary notes ("structure or die"). A note whose
+ * frozen deadline (`pages.temporary_expires_at`) has passed is auto-moved to
+ * trash via the exact same soft-delete path as a manual delete. Modelled on
+ * TrashCleanupService; `@nestjs/schedule` is already enabled globally.
+ */
+@Injectable()
+export class TemporaryNoteCleanupService {
+  private readonly logger = new Logger(TemporaryNoteCleanupService.name);
+
+  // Cap a single sweep so a large backlog (e.g. many notes created during
+  // downtime under a short lifetime) is not loaded into memory at once. The
+  // remainder is drained on the next hourly run; sub-hour overshoot is fine.
+  private static readonly SWEEP_BATCH_LIMIT = 500;
+
+  constructor(
+    @InjectKysely() private readonly db: KyselyDB,
+    private readonly pageRepo: PageRepo,
+  ) {}
+
+  // Hourly granularity: lifetimes are configured in hours, so a sub-hour
+  // overshoot past the deadline is acceptable.
+  @Interval('temporary-note-cleanup', 60 * 60 * 1000)
+  async sweepExpiredTemporaryNotes() {
+    try {
+      const now = new Date();
+
+      const expired = await this.db
+        .selectFrom('pages')
+        .select(['id', 'creatorId', 'workspaceId'])
+        .where('temporaryExpiresAt', 'is not', null)
+        .where('temporaryExpiresAt', '<', now)
+        .where('deletedAt', 'is', null) // not already in trash
+        .limit(TemporaryNoteCleanupService.SWEEP_BATCH_LIMIT)
+        .execute();
+
+      let trashed = 0;
+      for (const page of expired) {
+        try {
+          // Re-check the deadline at deletion time. The SELECT above is not
+          // transactional, so a user may click "Make permanent"
+          // (toggleTemporary sets temporary_expires_at = null) in the window
+          // between the SELECT and this per-row removePage. removePage deletes
+          // by id with only a `deletedAt IS NULL` filter and never re-reads the
+          // deadline, so without this guard a concurrently-kept note would
+          // still be trashed. Re-read the row and skip it unless it is still
+          // armed AND still expired, so a concurrent make-permanent wins.
+          const current = await this.db
+            .selectFrom('pages')
+            .select(['temporaryExpiresAt', 'deletedAt'])
+            .where('id', '=', page.id)
+            .executeTakeFirst();
+
+          if (
+            !current ||
+            current.deletedAt !== null ||
+            current.temporaryExpiresAt === null ||
+            new Date(current.temporaryExpiresAt) >= now
+          ) {
+            // Made permanent, already trashed, or no longer expired since the
+            // SELECT — leave it alone.
+            continue;
+          }
+
+          // Reuse the exact soft-delete path: recursive over children, removes
+          // shares in a transaction, and emits PAGE_SOFT_DELETED (tree
+          // invalidation + watcher notifications). Attribute the automatic
+          // deletion to the note's creator (no schema change). Both the SELECT
+          // above and removePage filter `deletedAt IS NULL`, so a double sweep
+          // is idempotent.
+          await this.pageRepo.removePage(
+            page.id,
+            // creatorId is set on every created page; a temporary note always
+            // has one. Cast to satisfy the non-null deletedById parameter.
+            page.creatorId as string,
+            page.workspaceId,
+          );
+          trashed++;
+        } catch (error) {
+          this.logger.error(
+            `Failed to trash expired temporary note ${page.id}`,
+            error instanceof Error ? error.stack : undefined,
+          );
+        }
+      }
+
+      if (trashed > 0) {
+        this.logger.debug(
+          `Temporary-note cleanup completed: ${trashed} notes trashed`,
+        );
+      }
+    } catch (error) {
+      this.logger.error(
+        'Temporary-note cleanup job failed',
+        error instanceof Error ? error.stack : undefined,
+      );
+    }
+  }
+}

apps/server/src/core/page/transclusion/dto/toggle-temporary.dto.ts

@@ -0,0 +1,15 @@
+import { IsBoolean, IsOptional, IsUUID } from 'class-validator';
+
+export class ToggleTemporaryDto {
+  @IsUUID()
+  pageId!: string;
+
+  /**
+   * When omitted, the temporary state is toggled relative to its current value.
+   * true  -> arm the timer (now + workspace temporaryNoteHours);
+   * false -> clear it (make permanent — "structure and survive").
+   */
+  @IsOptional()
+  @IsBoolean()
+  temporary?: boolean;
+}

apps/server/src/core/page/transclusion/page-template.controller.ts

@@ -16,8 +16,12 @@ import { TemplateLookupDto } from './dto/template-lookup.dto';
 import { PageRepo } from '@docmost/db/repos/page/page.repo';
 import { PageAccessService } from '../page-access/page-access.service';
 import { ToggleTemplateDto } from './dto/toggle-template.dto';
+import { ToggleTemporaryDto } from './dto/toggle-temporary.dto';
 import { UserThrottlerGuard } from '../../../integrations/throttle/user-throttler.guard';
 import { PAGE_TEMPLATE_THROTTLER } from '../../../integrations/throttle/throttler-names';
+import { InjectKysely } from 'nestjs-kysely';
+import { KyselyDB } from '@docmost/db/types/kysely.types';
+import { DEFAULT_TEMPORARY_NOTE_HOURS } from '../constants/temporary-note.constants';
 
 @UseGuards(JwtAuthGuard)
 @Controller('pages')
@@ -26,6 +30,7 @@ export class PageTemplateController {
     private readonly transclusionService: TransclusionService,
     private readonly pageRepo: PageRepo,
     private readonly pageAccessService: PageAccessService,
+    @InjectKysely() private readonly db: KyselyDB,
   ) {}
 
   /**
@@ -82,4 +87,54 @@ export class PageTemplateController {
 
     return { pageId: page.id, isTemplate };
   }
+
+  /**
+   * Arm or disarm the "death timer" on a page (`pages.temporary_expires_at`).
+   * Mirror of toggle-template: requires Edit on the page/space (CASL enforced in
+   * `validateCanEdit`). Arming freezes the deadline at now + the workspace's
+   * temporaryNoteHours; disarming ("Make permanent") clears it. Same workspace
+   * defense-in-depth as toggle-template (NotFound, never Forbidden, on mismatch).
+   */
+  @UseGuards(JwtAuthGuard, UserThrottlerGuard)
+  @Throttle({ [PAGE_TEMPLATE_THROTTLER]: { limit: 30, ttl: 60000 } })
+  @HttpCode(HttpStatus.OK)
+  @Post('toggle-temporary')
+  async toggleTemporary(
+    @Body() dto: ToggleTemporaryDto,
+    @AuthUser() user: User,
+  ) {
+    const page = await this.pageRepo.findById(dto.pageId);
+    if (!page || page.deletedAt) {
+      throw new NotFoundException('Page not found');
+    }
+
+    if (page.workspaceId !== user.workspaceId) {
+      // Defense-in-depth: never act on a page outside the caller's workspace.
+      // Use NotFound (not Forbidden) to avoid leaking cross-workspace existence.
+      throw new NotFoundException('Page not found');
+    }
+
+    await this.pageAccessService.validateCanEdit(page, user);
+
+    const makeTemporary =
+      typeof dto.temporary === 'boolean'
+        ? dto.temporary
+        : page.temporaryExpiresAt == null;
+
+    let temporaryExpiresAt: Date | null = null;
+    if (makeTemporary) {
+      const workspace = await this.db
+        .selectFrom('workspaces')
+        .select(['temporaryNoteHours'])
+        .where('id', '=', user.workspaceId)
+        .executeTakeFirst();
+      const hours =
+        workspace?.temporaryNoteHours ?? DEFAULT_TEMPORARY_NOTE_HOURS;
+      temporaryExpiresAt = new Date(Date.now() + hours * 60 * 60 * 1000);
+    }
+
+    await this.pageRepo.updatePage({ temporaryExpiresAt }, page.id);
+
+    return { pageId: page.id, temporaryExpiresAt };
+  }
 }

apps/server/src/core/page/transclusion/spec/page-template.controller.spec.ts

@@ -9,6 +9,7 @@ import { PageRepo } from '@docmost/db/repos/page/page.repo';
 import { PageAccessService } from '../../page-access/page-access.service';
 import { JwtAuthGuard } from '../../../../common/guards/jwt-auth.guard';
 import { UserThrottlerGuard } from '../../../../integrations/throttle/user-throttler.guard';
+import { KYSELY_MODULE_CONNECTION_TOKEN } from 'nestjs-kysely';
 
 describe('PageTemplateController.toggleTemplate', () => {
   let controller: PageTemplateController;
@@ -40,6 +41,8 @@ describe('PageTemplateController.toggleTemplate', () => {
         { provide: TransclusionService, useValue: transclusionService },
         { provide: PageRepo, useValue: pageRepo },
         { provide: PageAccessService, useValue: pageAccessService },
+        // toggleTemporary reads the workspace lifetime; toggleTemplate ignores it.
+        { provide: KYSELY_MODULE_CONNECTION_TOKEN(), useValue: {} },
       ],
     })
       .overrideGuard(JwtAuthGuard)

apps/server/src/core/page/transclusion/spec/page-temporary.controller.spec.ts

@@ -0,0 +1,220 @@
+import { Test } from '@nestjs/testing';
+import { ForbiddenException, NotFoundException } from '@nestjs/common';
+import { plainToInstance } from 'class-transformer';
+import { validate } from 'class-validator';
+import { KYSELY_MODULE_CONNECTION_TOKEN } from 'nestjs-kysely';
+import { PageTemplateController } from '../page-template.controller';
+import { TransclusionService } from '../transclusion.service';
+import { ToggleTemporaryDto } from '../dto/toggle-temporary.dto';
+import { PageRepo } from '@docmost/db/repos/page/page.repo';
+import { PageAccessService } from '../../page-access/page-access.service';
+import { JwtAuthGuard } from '../../../../common/guards/jwt-auth.guard';
+import { UserThrottlerGuard } from '../../../../integrations/throttle/user-throttler.guard';
+import { DEFAULT_TEMPORARY_NOTE_HOURS } from '../../constants/temporary-note.constants';
+
+/**
+ * Minimal chainable Kysely stub: every builder method returns `this`, and the
+ * terminal `executeTakeFirst` resolves the configured workspace row.
+ */
+function makeDbStub(workspaceRow: { temporaryNoteHours: number | null } | undefined) {
+  const builder: any = {
+    selectFrom: () => builder,
+    select: () => builder,
+    where: () => builder,
+    executeTakeFirst: jest.fn().mockResolvedValue(workspaceRow),
+  };
+  return builder;
+}
+
+describe('PageTemplateController.toggleTemporary', () => {
+  let controller: PageTemplateController;
+  let pageRepo: { findById: jest.Mock; updatePage: jest.Mock };
+  let pageAccessService: { validateCanEdit: jest.Mock };
+
+  const user = { id: 'u1', workspaceId: 'w1' } as any;
+
+  async function buildController(
+    page: any,
+    workspaceRow: { temporaryNoteHours: number | null } | undefined = {
+      temporaryNoteHours: null,
+    },
+  ) {
+    pageRepo = {
+      findById: jest.fn().mockResolvedValue(page),
+      updatePage: jest.fn().mockResolvedValue(undefined),
+    };
+    pageAccessService = {
+      validateCanEdit: jest.fn().mockResolvedValue(undefined),
+    };
+
+    const module = await Test.createTestingModule({
+      controllers: [PageTemplateController],
+      providers: [
+        { provide: TransclusionService, useValue: { lookupTemplate: jest.fn() } },
+        { provide: PageRepo, useValue: pageRepo },
+        { provide: PageAccessService, useValue: pageAccessService },
+        {
+          provide: KYSELY_MODULE_CONNECTION_TOKEN(),
+          useValue: makeDbStub(workspaceRow),
+        },
+      ],
+    })
+      .overrideGuard(JwtAuthGuard)
+      .useValue({ canActivate: () => true })
+      .overrideGuard(UserThrottlerGuard)
+      .useValue({ canActivate: () => true })
+      .compile();
+
+    controller = module.get(PageTemplateController);
+  }
+
+  beforeEach(() => {
+    jest.useFakeTimers().setSystemTime(new Date('2026-06-26T00:00:00.000Z'));
+  });
+
+  afterEach(() => {
+    jest.useRealTimers();
+  });
+
+  it('throws NotFound and does not touch the page when missing', async () => {
+    await buildController(null);
+    await expect(
+      controller.toggleTemporary({ pageId: 'p1' } as any, user),
+    ).rejects.toBeInstanceOf(NotFoundException);
+    expect(pageAccessService.validateCanEdit).not.toHaveBeenCalled();
+    expect(pageRepo.updatePage).not.toHaveBeenCalled();
+  });
+
+  it('throws NotFound (not Forbidden) for a cross-workspace page', async () => {
+    await buildController({
+      id: 'p1',
+      workspaceId: 'OTHER',
+      deletedAt: null,
+      temporaryExpiresAt: null,
+    });
+    await expect(
+      controller.toggleTemporary({ pageId: 'p1' } as any, user),
+    ).rejects.toBeInstanceOf(NotFoundException);
+    expect(pageRepo.updatePage).not.toHaveBeenCalled();
+  });
+
+  it('enforces CASL edit: when validateCanEdit throws, the timer is NOT changed', async () => {
+    await buildController({
+      id: 'p1',
+      workspaceId: 'w1',
+      deletedAt: null,
+      temporaryExpiresAt: null,
+    });
+    pageAccessService.validateCanEdit.mockRejectedValue(new ForbiddenException());
+    await expect(
+      controller.toggleTemporary({ pageId: 'p1' } as any, user),
+    ).rejects.toBeInstanceOf(ForbiddenException);
+    expect(pageRepo.updatePage).not.toHaveBeenCalled();
+  });
+
+  it('arms the timer (toggle) using the default hours when the page is permanent', async () => {
+    await buildController({
+      id: 'p1',
+      workspaceId: 'w1',
+      deletedAt: null,
+      temporaryExpiresAt: null,
+    });
+    const out = await controller.toggleTemporary({ pageId: 'p1' } as any, user);
+
+    const expected = new Date(
+      Date.now() + DEFAULT_TEMPORARY_NOTE_HOURS * 60 * 60 * 1000,
+    );
+    expect(pageAccessService.validateCanEdit).toHaveBeenCalled();
+    expect(pageRepo.updatePage).toHaveBeenCalledWith(
+      { temporaryExpiresAt: expected },
+      'p1',
+    );
+    expect(out).toEqual({ pageId: 'p1', temporaryExpiresAt: expected });
+  });
+
+  it('uses the workspace temporaryNoteHours override when set', async () => {
+    await buildController(
+      {
+        id: 'p1',
+        workspaceId: 'w1',
+        deletedAt: null,
+        temporaryExpiresAt: null,
+      },
+      { temporaryNoteHours: 3 },
+    );
+    const out = await controller.toggleTemporary({ pageId: 'p1' } as any, user);
+    const expected = new Date(Date.now() + 3 * 60 * 60 * 1000);
+    expect(pageRepo.updatePage).toHaveBeenCalledWith(
+      { temporaryExpiresAt: expected },
+      'p1',
+    );
+    expect(out.temporaryExpiresAt).toEqual(expected);
+  });
+
+  it('clears the timer (make permanent) when toggling an armed note', async () => {
+    await buildController({
+      id: 'p1',
+      workspaceId: 'w1',
+      deletedAt: null,
+      temporaryExpiresAt: new Date('2026-06-27T00:00:00.000Z'),
+    });
+    const out = await controller.toggleTemporary({ pageId: 'p1' } as any, user);
+    expect(pageRepo.updatePage).toHaveBeenCalledWith(
+      { temporaryExpiresAt: null },
+      'p1',
+    );
+    expect(out).toEqual({ pageId: 'p1', temporaryExpiresAt: null });
+  });
+
+  it('respects an explicit temporary:false instead of toggling', async () => {
+    await buildController({
+      id: 'p1',
+      workspaceId: 'w1',
+      deletedAt: null,
+      temporaryExpiresAt: null, // already permanent, but explicit false
+    });
+    const out = await controller.toggleTemporary(
+      { pageId: 'p1', temporary: false } as any,
+      user,
+    );
+    expect(pageRepo.updatePage).toHaveBeenCalledWith(
+      { temporaryExpiresAt: null },
+      'p1',
+    );
+    expect(out.temporaryExpiresAt).toBeNull();
+  });
+});
+
+describe('ToggleTemporaryDto validation (class-validator)', () => {
+  const uuid = '00000000-0000-4000-8000-000000000001';
+
+  it('accepts a valid UUID with no flag (toggle)', async () => {
+    const dto = plainToInstance(ToggleTemporaryDto, { pageId: uuid });
+    expect(await validate(dto)).toHaveLength(0);
+  });
+
+  it('accepts an explicit boolean temporary', async () => {
+    const dto = plainToInstance(ToggleTemporaryDto, {
+      pageId: uuid,
+      temporary: true,
+    });
+    expect(await validate(dto)).toHaveLength(0);
+  });
+
+  it('rejects a non-UUID pageId', async () => {
+    const dto = plainToInstance(ToggleTemporaryDto, { pageId: 'nope' });
+    const errors = await validate(dto);
+    expect(errors).toHaveLength(1);
+    expect(errors[0].constraints).toHaveProperty('isUuid');
+  });
+
+  it('rejects a non-boolean temporary', async () => {
+    const dto = plainToInstance(ToggleTemporaryDto, {
+      pageId: uuid,
+      temporary: 'yes',
+    });
+    const errors = await validate(dto);
+    expect(errors).toHaveLength(1);
+    expect(errors[0].constraints).toHaveProperty('isBoolean');
+  });
+});

apps/server/src/core/share/dto/share-alias.dto.ts

@@ -0,0 +1,44 @@
+import {
+  IsBoolean,
+  IsNotEmpty,
+  IsOptional,
+  IsString,
+} from 'class-validator';
+
+/**
+ * Create/retarget a vanity alias for a page. `confirmReassign` is the
+ * two-step guard for the "address already points at another page" case: the
+ * first call without it gets a 409 carrying the current target, the client
+ * confirms, and retries with `confirmReassign: true`.
+ */
+export class SetShareAliasDto {
+  @IsString()
+  @IsNotEmpty()
+  pageId: string;
+
+  @IsString()
+  @IsNotEmpty()
+  alias: string;
+
+  @IsBoolean()
+  @IsOptional()
+  confirmReassign?: boolean;
+}
+
+export class RemoveShareAliasDto {
+  @IsString()
+  @IsNotEmpty()
+  aliasId: string;
+}
+
+export class ShareAliasAvailabilityDto {
+  @IsString()
+  @IsNotEmpty()
+  alias: string;
+}
+
+export class ShareAliasForPageDto {
+  @IsString()
+  @IsNotEmpty()
+  pageId: string;
+}

apps/server/src/core/share/share-alias-redirect.controller.spec.ts

@@ -0,0 +1,252 @@
+import * as fs from 'node:fs';
+
+// `@sindresorhus/slugify` is ESM-only and not in jest's transformIgnorePatterns,
+// so the real module fails to parse under ts-jest. Stub it with a minimal,
+// deterministic slugifier — this spec asserts the controller's slug *assembly*
+// (`<title-slug>-<slugId>`, 70-char clamp, `untitled` fallback), not the upstream
+// slug algorithm. The factory keeps the real ESM module from ever being loaded.
+jest.mock('@sindresorhus/slugify', () => ({
+  __esModule: true,
+  default: (input: string) =>
+    String(input)
+      .toLowerCase()
+      .trim()
+      .replace(/[^a-z0-9]+/g, '-')
+      .replace(/^-+|-+$/g, ''),
+}));
+
+import { ShareAliasRedirectController } from './share-alias-redirect.controller';
+
+/**
+ * Routing/leak guard for the PUBLIC `GET /l/:alias` resolver.
+ *
+ * This is the most security-sensitive surface of the alias feature: an
+ * unauthenticated route that MUST serve the plain SPA index (exactly like any
+ * unknown path) for an unknown / dangling / no-longer-readable alias so that the
+ * existence of a name never leaks. Only a resolvable, still-readable alias may
+ * 302 to the canonical `/share/<key>/p/<title-slug>-<slugId>` page (302 — never
+ * 301 — because the target is retargetable). These tests pin that routing and
+ * the defensive percent-decoding, mirroring `share-seo.controller.routing.spec`.
+ */
+
+const STREAM_SENTINEL = { __isStream: true } as unknown as fs.ReadStream;
+
+// Stub fs at CALL time (jest.spyOn), NOT module load (jest.mock): the controller
+// transitively pulls bcrypt, whose native module is located by node-gyp-build
+// reading the filesystem at import time — a module-level fs mock breaks that.
+beforeEach(() => {
+  jest.spyOn(fs, 'existsSync').mockReturnValue(true);
+  jest.spyOn(fs, 'createReadStream').mockReturnValue(STREAM_SENTINEL);
+});
+afterEach(() => jest.restoreAllMocks());
+
+function makeRes() {
+  const res: any = {
+    sent: undefined as unknown,
+    statusCode: undefined as number | undefined,
+    redirectUrl: undefined as string | undefined,
+    type: jest.fn(() => res),
+    status: jest.fn((code: number) => {
+      res.statusCode = code;
+      return res;
+    }),
+    send: jest.fn((v: unknown) => {
+      res.sent = v;
+      return res;
+    }),
+    redirect: jest.fn((url: string, code: number) => {
+      res.redirectUrl = url;
+      res.statusCode = code;
+      return res;
+    }),
+  };
+  return res;
+}
+
+function makeController(opts: {
+  resolved?: { share: any; page: any } | null;
+  selfHosted?: boolean;
+}) {
+  const shareAliasService = {
+    resolveReadableTarget: jest.fn(async () => opts.resolved ?? null),
+  };
+  const workspaceRepo = {
+    findFirst: jest.fn(async () => ({ id: 'ws-self' })),
+    findByHostname: jest.fn(async (sub: string) =>
+      sub === 'acme' ? { id: 'ws-acme' } : null,
+    ),
+  };
+  const environmentService = {
+    isSelfHosted: jest.fn(() => opts.selfHosted ?? true),
+  };
+  const controller = new ShareAliasRedirectController(
+    shareAliasService as any,
+    workspaceRepo as any,
+    environmentService as any,
+  );
+  return { controller, shareAliasService, workspaceRepo, environmentService };
+}
+
+const selfReq: any = { raw: { headers: { host: 'self' } } };
+
+describe('ShareAliasRedirectController.resolve', () => {
+  it('302-redirects a resolvable alias to the canonical share page', async () => {
+    const { controller, shareAliasService } = makeController({
+      resolved: {
+        share: { key: 'SHAREKEY' },
+        page: { slugId: 'abc123', title: 'Quarterly Report' },
+      },
+    });
+    const res = makeRes();
+
+    await controller.resolve('promo', selfReq, res);
+
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalledWith(
+      'promo',
+      'ws-self',
+    );
+    expect(res.redirect).toHaveBeenCalledWith(
+      '/share/SHAREKEY/p/quarterly-report-abc123',
+      302,
+    );
+    // No index stream was served on a hit.
+    expect(res.sent).toBeUndefined();
+  });
+
+  it('falls back to "untitled" in the slug when the target has no title', async () => {
+    const { controller } = makeController({
+      resolved: { share: { key: 'K' }, page: { slugId: 'sid', title: '' } },
+    });
+    const res = makeRes();
+
+    await controller.resolve('promo', selfReq, res);
+
+    expect(res.redirect).toHaveBeenCalledWith('/share/K/p/untitled-sid', 302);
+  });
+
+  it('clamps the title-slug to the first 70 characters of the page title', async () => {
+    // 119-char title; only the first 70 chars must reach the slug. The 70-char
+    // boundary deliberately falls mid-word ("Entire" -> "entir") so the clamp is
+    // unambiguous: anything past char 70 ("...e Fiscal Year...") must be dropped.
+    const longTitle =
+      'The Comprehensive Quarterly Financial Performance Report For The Entire Fiscal Year Two Thousand Twenty Five And Beyond';
+    const { controller } = makeController({
+      resolved: {
+        share: { key: 'K' },
+        page: { slugId: 'sid', title: longTitle },
+      },
+    });
+    const res = makeRes();
+
+    await controller.resolve('promo', selfReq, res);
+
+    expect(res.redirect).toHaveBeenCalledWith(
+      '/share/K/p/the-comprehensive-quarterly-financial-performance-report-for-the-entir-sid',
+      302,
+    );
+  });
+
+  it('streams the SPA index WITHOUT a 302 for an unknown/dangling/unreadable alias (no leak)', async () => {
+    const { controller, shareAliasService } = makeController({ resolved: null });
+    const res = makeRes();
+
+    await controller.resolve('does-not-exist', selfReq, res);
+
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalled();
+    // The plain index stream was served and no redirect leaked alias existence.
+    expect(res.redirect).not.toHaveBeenCalled();
+    expect(res.sent).toBe(STREAM_SENTINEL);
+    expect(res.type).toHaveBeenCalledWith('text/html');
+  });
+
+  it('streams the SPA index without even resolving when the workspace is null', async () => {
+    // Subdomain host that maps to no workspace => workspace === null.
+    const { controller, shareAliasService, workspaceRepo } = makeController({
+      selfHosted: false,
+    });
+    const res = makeRes();
+    const req: any = { raw: { headers: { host: 'unknown.example.com' } } };
+
+    await controller.resolve('promo', req, res);
+
+    expect(workspaceRepo.findByHostname).toHaveBeenCalledWith('unknown');
+    // Never even attempts to resolve (alias existence cannot leak per-host).
+    expect(shareAliasService.resolveReadableTarget).not.toHaveBeenCalled();
+    expect(res.redirect).not.toHaveBeenCalled();
+    expect(res.sent).toBe(STREAM_SENTINEL);
+  });
+
+  it('defensively decodes broken percent-encoding and treats it as unknown', async () => {
+    const { controller, shareAliasService } = makeController({ resolved: null });
+    const res = makeRes();
+
+    // '%E0%A4%A' is invalid -> decodeURIComponent throws -> raw value is used,
+    // and the alias resolves to nothing (no crash, served as index).
+    await controller.resolve('%E0%A4%A', selfReq, res);
+
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalledWith(
+      '%E0%A4%A',
+      'ws-self',
+    );
+    expect(res.redirect).not.toHaveBeenCalled();
+    expect(res.sent).toBe(STREAM_SENTINEL);
+  });
+
+  it('decodes a valid percent-encoded alias before resolving', async () => {
+    const { controller, shareAliasService } = makeController({ resolved: null });
+    const res = makeRes();
+
+    await controller.resolve('my%2Dlink', selfReq, res);
+
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalledWith(
+      'my-link',
+      'ws-self',
+    );
+  });
+
+  it('resolves the workspace via findFirst on the self-hosted path', async () => {
+    const { controller, workspaceRepo, shareAliasService } = makeController({
+      selfHosted: true,
+      resolved: null,
+    });
+    const res = makeRes();
+
+    await controller.resolve('promo', selfReq, res);
+
+    expect(workspaceRepo.findFirst).toHaveBeenCalled();
+    expect(workspaceRepo.findByHostname).not.toHaveBeenCalled();
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalledWith(
+      'promo',
+      'ws-self',
+    );
+  });
+
+  it('resolves the workspace via findByHostname (subdomain) on the cloud path', async () => {
+    const { controller, workspaceRepo, shareAliasService } = makeController({
+      selfHosted: false,
+      resolved: null,
+    });
+    const res = makeRes();
+    const req: any = { raw: { headers: { host: 'acme.example.com' } } };
+
+    await controller.resolve('promo', req, res);
+
+    expect(workspaceRepo.findByHostname).toHaveBeenCalledWith('acme');
+    expect(workspaceRepo.findFirst).not.toHaveBeenCalled();
+    expect(shareAliasService.resolveReadableTarget).toHaveBeenCalledWith(
+      'promo',
+      'ws-acme',
+    );
+  });
+
+  it('serves a 404 when no built client index exists', async () => {
+    jest.spyOn(fs, 'existsSync').mockReturnValue(false);
+    const { controller } = makeController({ resolved: null });
+    const res = makeRes();
+
+    await controller.resolve('promo', selfReq, res);
+
+    expect(res.status).toHaveBeenCalledWith(404);
+    expect(res.redirect).not.toHaveBeenCalled();
+  });
+});

apps/server/src/core/share/share-alias-redirect.controller.ts

@@ -0,0 +1,95 @@
+import { Controller, Get, Param, Req, Res } from '@nestjs/common';
+import { FastifyReply, FastifyRequest } from 'fastify';
+import { join } from 'path';
+import * as fs from 'node:fs';
+import slugify from '@sindresorhus/slugify';
+import { WorkspaceRepo } from '@docmost/db/repos/workspace/workspace.repo';
+import { EnvironmentService } from '../../integrations/environment/environment.service';
+import { Workspace } from '@docmost/db/types/entity.types';
+import { ShareAliasService } from './share-alias.service';
+
+/**
+ * Public resolver for vanity links `GET /l/:alias`. Excluded from the global
+ * `/api` prefix (see main.ts) and parallel to ShareSeoController.
+ *
+ * On a hit it issues a 302 (NEVER 301) to the canonical
+ * `/share/:key/p/:slug` page, so:
+ *   - the existing share render + SSR meta is reused verbatim (crawlers follow
+ *     the 302 and get the correct preview);
+ *   - because the alias target is mutable, a temporary redirect is always
+ *     re-resolved — a cached 301 would pin clients to the pre-swap page.
+ *
+ * Any unknown / dangling / no-longer-readable alias serves the plain SPA index
+ * (same as any unknown path) so the existence of a name never leaks.
+ */
+@Controller('l')
+export class ShareAliasRedirectController {
+  constructor(
+    private readonly shareAliasService: ShareAliasService,
+    private readonly workspaceRepo: WorkspaceRepo,
+    private readonly environmentService: EnvironmentService,
+  ) {}
+
+  @Get(':alias')
+  async resolve(
+    @Param('alias') rawAlias: string,
+    @Req() req: FastifyRequest,
+    @Res({ passthrough: false }) res: FastifyReply,
+  ) {
+    // NestJS does not apply middlewares to paths excluded from the global /api
+    // prefix, so the DomainMiddleware workspace resolution is duplicated here
+    // (same workaround as ShareSeoController).
+    let workspace: Workspace = null;
+    if (this.environmentService.isSelfHosted()) {
+      workspace = await this.workspaceRepo.findFirst();
+    } else {
+      const header = req.raw.headers.host;
+      const subdomain = header?.split('.')[0];
+      workspace = subdomain
+        ? await this.workspaceRepo.findByHostname(subdomain)
+        : null;
+    }
+
+    const clientDistPath = join(__dirname, '..', '..', '..', '..', 'client/dist');
+    const indexFilePath = join(clientDistPath, 'index.html');
+
+    let decoded = rawAlias;
+    try {
+      decoded = decodeURIComponent(rawAlias);
+    } catch {
+      // Malformed percent-encoding -> treat as unknown alias.
+    }
+
+    const resolved = workspace
+      ? await this.shareAliasService.resolveReadableTarget(
+          decoded,
+          workspace.id,
+        )
+      : null;
+
+    if (!resolved) {
+      return this.sendIndex(indexFilePath, res);
+    }
+
+    const slug = buildPageSlug(resolved.page.slugId, resolved.page.title);
+    // 302, NOT 301: the alias is retargetable, so the redirect must always be
+    // re-resolved by clients/crawlers.
+    return res.redirect(`/share/${resolved.share.key}/p/${slug}`, 302);
+  }
+
+  private sendIndex(indexFilePath: string, res: FastifyReply) {
+    if (!fs.existsSync(indexFilePath)) {
+      // No built client (e.g. API-only dev): nothing to serve.
+      res.status(404).send('Not found');
+      return;
+    }
+    const stream = fs.createReadStream(indexFilePath);
+    res.type('text/html').send(stream);
+  }
+}
+
+/** Canonical share page slug: `<title-slug>-<slugId>` (mirrors the client). */
+function buildPageSlug(slugId: string, title?: string): string {
+  const titleSlug = slugify(title?.substring(0, 70) || 'untitled');
+  return `${titleSlug}-${slugId}`;
+}

apps/server/src/core/share/share-alias.controller.spec.ts

@@ -0,0 +1,260 @@
+import {
+  BadRequestException,
+  ForbiddenException,
+  NotFoundException,
+} from '@nestjs/common';
+import { ShareAliasController } from './share-alias.controller';
+
+/**
+ * Authz-gate tests for the authenticated alias management controller. The access
+ * decisions for creating/retargeting/removing an alias live in THIS controller
+ * (the service spec delegates authorization to the caller), so each gate is
+ * pinned here against mocked PageRepo / ShareService / ShareAliasService /
+ * PageAccessService. A regression that drops any gate must fail here.
+ */
+describe('ShareAliasController authz gates', () => {
+  function makeController() {
+    const shareAliasService = {
+      setAlias: jest.fn(async () => ({ id: 'alias-1' })),
+      removeAlias: jest.fn(async () => undefined),
+      getAliasById: jest.fn(),
+      getAliasForPage: jest.fn(),
+      checkAvailability: jest.fn(),
+    };
+    const shareService = {
+      resolveReadableSharePage: jest.fn(),
+      isSharingAllowed: jest.fn(),
+    };
+    const pageRepo = { findById: jest.fn() };
+    const pageAccessService = {
+      validateCanEdit: jest.fn(async () => undefined),
+      validateCanView: jest.fn(async () => undefined),
+    };
+    const controller = new ShareAliasController(
+      shareAliasService as any,
+      shareService as any,
+      pageRepo as any,
+      pageAccessService as any,
+    );
+    return {
+      controller,
+      shareAliasService,
+      shareService,
+      pageRepo,
+      pageAccessService,
+    };
+  }
+
+  const user: any = { id: 'u-1' };
+  const workspace: any = { id: 'ws-1' };
+
+  describe('set', () => {
+    it('throws NotFoundException for a nonexistent page', async () => {
+      const { controller, pageRepo, pageAccessService } = makeController();
+      pageRepo.findById.mockResolvedValue(null);
+
+      await expect(
+        controller.set({ pageId: 'p-x', alias: 'promo' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(NotFoundException);
+      expect(pageAccessService.validateCanEdit).not.toHaveBeenCalled();
+    });
+
+    it('throws NotFoundException for a page in another workspace', async () => {
+      const { controller, pageRepo } = makeController();
+      pageRepo.findById.mockResolvedValue({
+        id: 'p-1',
+        workspaceId: 'ws-OTHER',
+      });
+
+      await expect(
+        controller.set({ pageId: 'p-1', alias: 'promo' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(NotFoundException);
+    });
+
+    it('enforces validateCanEdit before setting the alias', async () => {
+      const { controller, pageRepo, pageAccessService, shareService } =
+        makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      pageAccessService.validateCanEdit.mockRejectedValue(
+        new ForbiddenException('no edit'),
+      );
+
+      await expect(
+        controller.set({ pageId: 'p-1', alias: 'promo' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(ForbiddenException);
+      // Gate short-circuits before any share resolution.
+      expect(shareService.resolveReadableSharePage).not.toHaveBeenCalled();
+    });
+
+    it('throws BadRequestException when the page is not publicly shared', async () => {
+      const { controller, pageRepo, shareService } = makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      shareService.resolveReadableSharePage.mockResolvedValue(null);
+
+      await expect(
+        controller.set({ pageId: 'p-1', alias: 'promo' } as any, user, workspace),
+      ).rejects.toThrow('Page is not publicly shared');
+      await expect(
+        controller.set({ pageId: 'p-1', alias: 'promo' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(BadRequestException);
+    });
+
+    it('throws ForbiddenException when public sharing is disabled', async () => {
+      const { controller, pageRepo, shareService } = makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      shareService.resolveReadableSharePage.mockResolvedValue({
+        share: { spaceId: 'sp-1' },
+      });
+      shareService.isSharingAllowed.mockResolvedValue(false);
+
+      await expect(
+        controller.set({ pageId: 'p-1', alias: 'promo' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(ForbiddenException);
+    });
+
+    it('delegates to setAlias on the happy path with all gates passed', async () => {
+      const { controller, pageRepo, shareService, shareAliasService } =
+        makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      shareService.resolveReadableSharePage.mockResolvedValue({
+        share: { spaceId: 'sp-1' },
+      });
+      shareService.isSharingAllowed.mockResolvedValue(true);
+
+      const result = await controller.set(
+        { pageId: 'p-1', alias: 'promo', confirmReassign: true } as any,
+        user,
+        workspace,
+      );
+
+      expect(shareAliasService.setAlias).toHaveBeenCalledWith({
+        workspaceId: 'ws-1',
+        pageId: 'p-1',
+        creatorId: 'u-1',
+        alias: 'promo',
+        confirmReassign: true,
+      });
+      expect(result).toEqual({ id: 'alias-1' });
+    });
+  });
+
+  describe('remove', () => {
+    it('throws NotFoundException for an unknown alias', async () => {
+      const { controller, shareAliasService } = makeController();
+      shareAliasService.getAliasById.mockResolvedValue(null);
+
+      await expect(
+        controller.remove({ aliasId: 'a-x' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(NotFoundException);
+      expect(shareAliasService.removeAlias).not.toHaveBeenCalled();
+    });
+
+    it('requires validateCanEdit on the current target before removing', async () => {
+      const { controller, shareAliasService, pageRepo, pageAccessService } =
+        makeController();
+      shareAliasService.getAliasById.mockResolvedValue({
+        id: 'a-1',
+        pageId: 'p-1',
+      });
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      pageAccessService.validateCanEdit.mockRejectedValue(
+        new ForbiddenException('no edit'),
+      );
+
+      await expect(
+        controller.remove({ aliasId: 'a-1' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(ForbiddenException);
+      expect(shareAliasService.removeAlias).not.toHaveBeenCalled();
+    });
+
+    it('removes a dangling alias (pageId null) WITHOUT an edit check', async () => {
+      const { controller, shareAliasService, pageRepo, pageAccessService } =
+        makeController();
+      shareAliasService.getAliasById.mockResolvedValue({
+        id: 'a-1',
+        pageId: null,
+      });
+
+      await controller.remove({ aliasId: 'a-1' } as any, user, workspace);
+
+      expect(pageRepo.findById).not.toHaveBeenCalled();
+      expect(pageAccessService.validateCanEdit).not.toHaveBeenCalled();
+      expect(shareAliasService.removeAlias).toHaveBeenCalledWith('a-1', 'ws-1');
+    });
+
+    it('removes when the editor can edit the current target', async () => {
+      const { controller, shareAliasService, pageRepo, pageAccessService } =
+        makeController();
+      shareAliasService.getAliasById.mockResolvedValue({
+        id: 'a-1',
+        pageId: 'p-1',
+      });
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+
+      await controller.remove({ aliasId: 'a-1' } as any, user, workspace);
+
+      expect(pageAccessService.validateCanEdit).toHaveBeenCalled();
+      expect(shareAliasService.removeAlias).toHaveBeenCalledWith('a-1', 'ws-1');
+    });
+
+    it('removes even if the recorded target page no longer exists', async () => {
+      const { controller, shareAliasService, pageRepo, pageAccessService } =
+        makeController();
+      shareAliasService.getAliasById.mockResolvedValue({
+        id: 'a-1',
+        pageId: 'p-gone',
+      });
+      pageRepo.findById.mockResolvedValue(null);
+
+      await controller.remove({ aliasId: 'a-1' } as any, user, workspace);
+
+      expect(pageAccessService.validateCanEdit).not.toHaveBeenCalled();
+      expect(shareAliasService.removeAlias).toHaveBeenCalledWith('a-1', 'ws-1');
+    });
+  });
+
+  describe('forPage', () => {
+    it('throws NotFoundException for a cross-workspace/nonexistent page', async () => {
+      const { controller, pageRepo, pageAccessService } = makeController();
+      pageRepo.findById.mockResolvedValue({
+        id: 'p-1',
+        workspaceId: 'ws-OTHER',
+      });
+
+      await expect(
+        controller.forPage({ pageId: 'p-1' } as any, user, workspace),
+      ).rejects.toBeInstanceOf(NotFoundException);
+      expect(pageAccessService.validateCanView).not.toHaveBeenCalled();
+    });
+
+    it('requires validateCanView and returns the alias (or null)', async () => {
+      const { controller, pageRepo, pageAccessService, shareAliasService } =
+        makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      shareAliasService.getAliasForPage.mockResolvedValue({ id: 'a-1' });
+
+      const result = await controller.forPage(
+        { pageId: 'p-1' } as any,
+        user,
+        workspace,
+      );
+
+      expect(pageAccessService.validateCanView).toHaveBeenCalled();
+      expect(result).toEqual({ id: 'a-1' });
+    });
+
+    it('returns null when the page has no alias', async () => {
+      const { controller, pageRepo, shareAliasService } = makeController();
+      pageRepo.findById.mockResolvedValue({ id: 'p-1', workspaceId: 'ws-1' });
+      shareAliasService.getAliasForPage.mockResolvedValue(undefined);
+
+      const result = await controller.forPage(
+        { pageId: 'p-1' } as any,
+        user,
+        workspace,
+      );
+
+      expect(result).toBeNull();
+    });
+  });
+});