fix(footnotes): strip bare definitions on rebuild; MCP full-doc + zip-import canonicalize tests (#228)

Review #6 (approve-with-comments) follow-ups:
1. canonicalize step 7 now strips bare footnoteDefinitions at ANY depth
   (stripFootnoteDefinitionsDeep), not just footnotesList, in BOTH copies. A
   definition hand-authored outside a list (e.g. nested in a callout via a
   raw-JSON write path) was left in place while a copy was also added to the
   rebuilt list -> duplicate, idempotent, self-perpetuating. Runs only in the
   rebuild path (after the lists are stripped); the fast-path / placement-keep
   branch is untouched. Added a shared-corpus case (bare def nested in a callout)
   to pin it in both mirrors.
2. markdown-clipboard: removed the dead top-level footnoteReference check in
   canonicalizePastedFootnotes (an inline atom is never a top-level slice child;
   only the descendants scan can find it).

Test coverage:
4. New MCP binding tests (full-doc-write-canonicalize.test.mjs): update_page_json
   and copy_page_content canonicalize the persisted full doc, asserted via a new
   `replacePage` seam (symmetric to the existing `mutatePage` seam) so no live
   collab socket is needed. Routed both writers through the seam.
5. New server spec (file-import-task.service.footnote-canonicalize.spec.ts): the
   zip-import path (processGenericImport) canonicalizes footnotes — real
   markdown->HTML->JSON via a real ImportService over a temp-dir .md file, DB trx
   stubbed to capture the persisted page content. FileImportTaskService had no
   spec before.

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

.env.example

@@ -132,6 +132,14 @@ MCP_DOCMOST_PASSWORD=
 # NEVER set is_agent on a human or shared account — every action by that account
 # (including normal human edits) would then be mis-attributed as AI.
 
+# Agent-roles catalog source: an http(s):// base URL to the catalog's raw files
+# (the server appends /index.json and /bundles/<id>/<lang>.json). This value is
+# baked into the Docker image at build time per branch (see the Dockerfile ARG
+# AI_AGENT_ROLES_CATALOG_URL and the CI build-args). Set it here only to point a
+# local/non-Docker run at a catalog; if unset, the "import role from catalog"
+# admin feature is unavailable. Local-filesystem sources are no longer supported.
+# AI_AGENT_ROLES_CATALOG_URL=
+
 # Per-embedding-call timeout in milliseconds for the RAG indexer.
 # A slow/hung embeddings endpoint fails after this and the batch continues.
 # AI_EMBEDDING_TIMEOUT_MS=120000

.github/workflows/develop.yml

@@ -52,6 +52,7 @@ jobs:
           platforms: linux/amd64
           build-args: |
             APP_VERSION=${{ steps.version.outputs.value }}
+            AI_AGENT_ROLES_CATALOG_URL=https://raw.githubusercontent.com/vvzvlad/gitmost/develop/agent-roles-catalog
           push: true
           tags: ${{ env.IMAGE }}:develop
           cache-from: type=gha,scope=develop-amd64

.github/workflows/release.yml

@@ -17,6 +17,7 @@ permissions:
 env:
   VERSION: ${{ inputs.version || github.ref_name }}
   IMAGE: ghcr.io/vvzvlad/gitmost
+  AI_AGENT_ROLES_CATALOG_URL: https://raw.githubusercontent.com/vvzvlad/gitmost/main/agent-roles-catalog
 
 jobs:
   # Run the reusable test suite first so a failing test blocks the image build.
@@ -57,6 +58,7 @@ jobs:
           platforms: ${{ matrix.platform }}
           build-args: |
             APP_VERSION=${{ env.VERSION }}
+            AI_AGENT_ROLES_CATALOG_URL=${{ env.AI_AGENT_ROLES_CATALOG_URL }}
           outputs: type=image,name=${{ env.IMAGE }},push-by-digest=true,name-canonical=true,push=true
           cache-from: type=gha,scope=${{ matrix.suffix }}
           cache-to: type=gha,scope=${{ matrix.suffix }},mode=max,ignore-error=true
@@ -85,6 +87,7 @@ jobs:
           platforms: ${{ matrix.platform }}
           build-args: |
             APP_VERSION=${{ env.VERSION }}
+            AI_AGENT_ROLES_CATALOG_URL=${{ env.AI_AGENT_ROLES_CATALOG_URL }}
           push: false
           tags: |
             ${{ env.IMAGE }}:latest

AGENTS.md

@@ -254,7 +254,7 @@ The API server is a Fastify app with a global `/api` prefix (`main.ts` excludes
 - **Redis** backs caching, the BullMQ queues, the WebSocket Socket.IO adapter, and collaboration sync.
 
 ### The two AI subsystems (the main fork additions)
-1. **Embedded MCP server** (`integrations/mcp/` + `packages/mcp`). The standalone `@docmost/mcp` server (38 agent-native tools: per-block patch/insert/delete by id, scripted `(doc)=>doc` transforms with dry-run diff, table editing, version diff/restore, comments, images, shares) is bundled and served over HTTP at `/mcp`. It writes through Docmost's real-time-collaboration layer so concurrent human edits aren't clobbered. Each request authenticates **per-user** via the `Authorization` header — either HTTP Basic (`base64(email:password)`, the user's own Docmost login, validated through `AuthService`) or a Bearer access JWT (the user's `authToken`) — and the session acts under that user's permissions. `MCP_DOCMOST_EMAIL` / `MCP_DOCMOST_PASSWORD` are an **optional service-account fallback**, used only when a request carries neither Basic nor Bearer credentials (back-compat for CI/scripts). An admin enables MCP with a workspace toggle (Workspace settings → AI). Optionally protected by a shared `MCP_TOKEN`: when set, every `/mcp` request must carry a matching `X-MCP-Token` header (its own header, separate from `Authorization`, which now carries the per-user Basic/Bearer credentials). Note: this changed from the older `Authorization: Bearer <MCP_TOKEN>` scheme — see `.env.example` and the CHANGELOG Breaking Changes entry.
+1. **Embedded MCP server** (`integrations/mcp/` + `packages/mcp`). The standalone `@docmost/mcp` server (39 agent-native tools: per-block patch/insert/delete by id, scripted `(doc)=>doc` transforms with dry-run diff, table editing, version diff/restore, comments, images, shares) is bundled and served over HTTP at `/mcp`. It writes through Docmost's real-time-collaboration layer so concurrent human edits aren't clobbered. Each request authenticates **per-user** via the `Authorization` header — either HTTP Basic (`base64(email:password)`, the user's own Docmost login, validated through `AuthService`) or a Bearer access JWT (the user's `authToken`) — and the session acts under that user's permissions. `MCP_DOCMOST_EMAIL` / `MCP_DOCMOST_PASSWORD` are an **optional service-account fallback**, used only when a request carries neither Basic nor Bearer credentials (back-compat for CI/scripts). An admin enables MCP with a workspace toggle (Workspace settings → AI). Optionally protected by a shared `MCP_TOKEN`: when set, every `/mcp` request must carry a matching `X-MCP-Token` header (its own header, separate from `Authorization`, which now carries the per-user Basic/Bearer credentials). Note: this changed from the older `Authorization: Bearer <MCP_TOKEN>` scheme — see `.env.example` and the CHANGELOG Breaking Changes entry.
 2. **AI agent chat** (`core/ai-chat/` server + `apps/client/src/features/ai-chat/` client). A built-in agent over the wiki using the Vercel **AI SDK** (`ai`, `@ai-sdk/*`) against any OpenAI-compatible provider configured per workspace (`integrations/ai/` — credentials encrypted at rest via `integrations/crypto`, stored in `ai_provider_credentials`). Key pieces:
    - `core/ai-chat/tools/` — the agent's ~40 read+write tools. Every tool runs under the **calling user's** CASL permissions via a per-user loopback access token (`docmost-client.loader.ts`), so the agent can never exceed what the user could do. Only **reversible** operations are exposed (page history + trash; no permanent delete). Agent edits get an "AI agent" provenance badge in page history (`20260616T130000-agent-provenance` migration).
    - `core/ai-chat/embedding/` — RAG indexer + a BullMQ consumer on `AI_QUEUE` that embeds pages into `page_embeddings` (vector search), complementing Postgres full-text search. Pages are (re)indexed on edit; `AI_EMBEDDING_TIMEOUT_MS` bounds a hung embeddings endpoint.

CHANGELOG.md

@@ -28,6 +28,37 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   answer was cut off and builds on it instead of restarting; the rest of the
   queue still flushes normally afterward. (#198)
 
+- **Importable multilingual agent-roles catalog.** Admins can browse a curated
+  catalog of agent roles, grouped into bundles and offered in several languages,
+  and import the ones they want into the workspace (with skip-or-rename handling
+  for name collisions); the same role in a different language imports as a
+  separate install. An imported role remembers its catalog origin and offers a
+  one-click update when the catalog ships a newer revision. Backed by four new
+  admin endpoints — `POST /ai-chat/roles/catalog` (browse bundles),
+  `/catalog/bundle` (read one bundle's roles), `/import`, and
+  `/update-from-catalog` — and a new `source` column linking a role to its
+  catalog slug/language/version. The catalog source is configured via the
+  `AI_AGENT_ROLES_CATALOG_URL` env var — an `http(s)://` base URL to the
+  catalog's raw files; the image ships a per-branch default baked in CI, and it
+  can be overridden at runtime via the env var (see `.env.example`). (#222)
+- **Author footnotes inline from an agent, and deterministic server-side footnote
+  canonicalization on every non-editor write path.** A new MCP `insert_footnote`
+  tool places a footnote at a body anchor by content only — the agent supplies
+  WHERE (anchor text) and WHAT (markdown); the number and the bottom
+  `footnotesList` are derived server-side, so an agent can never assign a number,
+  edit the list, or desync, and a same-content note reuses one definition. Under
+  the hood, the editor's footnote-integrity invariant (one trailing list,
+  numbering by first reference, no orphans/duplicates, no raw `[^id]`) is now
+  enforced as a pure `canonicalizeFootnotes(doc)` on the FULL-document write paths
+  that bypass the editor's plugins: server markdown/HTML import, `PageService`
+  create and full-document (`replace`) updates, the client markdown paste, and the
+  MCP markdown page-import / `update_page` (markdown) / `update_page_json` /
+  `docmost_transform` / `insert_footnote` / `copy_page_content` paths. It is
+  idempotent (a no-op once canonical) and is deliberately NOT applied to
+  append/prepend fragments, nor to COMMENT bodies — a comment may legitimately
+  contain a standalone footnote definition, which canonicalization would drop.
+  (#228)
+
 ### Fixed
 
 - **A shared page now keeps EXACTLY ONE custom address (`/l/:alias`).** Editing a
@@ -41,6 +72,13 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   page), so any previously-live duplicate `/l/<old>` link begins returning the
   generic 404 after upgrade — intended, but not undoable by `down()`. (#226,
   #227)
+- **Typing a custom address already used by another page no longer looks like a
+  dead end.** The share modal previously flagged such a name with a red "This
+  address is already in use" error, hiding the fact that saving offers to MOVE
+  the address to the current page. The field now shows an informational hint —
+  "This address is in use. Saving will move it to this page." — and keeps Save
+  enabled, so the existing reassign-confirm flow (`409 ALIAS_REASSIGN_REQUIRED` →
+  "Move custom address?") is discoverable instead of reading as terminal. (#227)
 
 ## [0.94.0] - 2026-06-26
 
@@ -120,6 +158,13 @@ per-workspace rolling-day token budget.
   applies it through the existing `/pages/update` route — reflecting it in the
   title field and broadcasting to other clients. Gated by the `settings.ai.generative`
   flag and throttled per user. (#199)
+- **AI chat: header button auto-opens the chat bound to the current document.**
+  Clicking the AI-chat button in the header while viewing a page now reopens the
+  latest chat tied to that document instead of whatever chat was last active,
+  reusing the existing `ai_chats.page_id` provenance (no migration). The newest
+  chat you created on the page wins; with no bound chat — or off a page, or if
+  the lookup fails — it falls soft to a fresh chat and keeps the current
+  selection otherwise. (#191)
 
 ### Changed
 

Dockerfile

@@ -23,6 +23,11 @@ RUN apt-get update \
 
 WORKDIR /app
 
+# Agent-roles catalog base URL: per-branch default set at build time (CI);
+# overridable at runtime via the AI_AGENT_ROLES_CATALOG_URL env var.
+ARG AI_AGENT_ROLES_CATALOG_URL=""
+ENV AI_AGENT_ROLES_CATALOG_URL=$AI_AGENT_ROLES_CATALOG_URL
+
 # Copy apps
 COPY --from=builder /app/apps/server/dist /app/apps/server/dist
 COPY --from=builder /app/apps/client/dist /app/apps/client/dist

README.md

@@ -34,7 +34,7 @@ The goal of the fork is a **100% open, AGPL-only build with no Enterprise-Editio
 | --- | --- |
 | **EE code removed** | Stripped all client and server Enterprise-Edition code; ships as a clean community/AGPL build with no license checks. |
 | **Comment resolution** | Re-implemented from scratch as a community feature (resolve / re-open with Open/Resolved tabs). No EE code reused, available to anyone who can comment. |
-| **Embedded MCP server** | A community MCP server (`@docmost/mcp`, 38 tools) is served over HTTP at `/mcp` — no enterprise license required. Replaces the removed license-gated EE MCP. |
+| **Embedded MCP server** | A community MCP server (`@docmost/mcp`, 39 tools) is served over HTTP at `/mcp` — no enterprise license required. Replaces the removed license-gated EE MCP. |
 | **AI agent chat** | Built-in AI agent chat over your wiki, written from scratch as a community feature — no enterprise license. The agent reads and edits pages on your behalf (scoped to your permissions), with full-text + vector (RAG) search and optional web access via external MCP servers. |
 | **Rebranding** | App logo / name changed from *Docmost* to *Gitmost*. |
 | **Compact page tree** | Default page-tree indentation reduced from 16px to 8px per nesting level. |
@@ -44,7 +44,7 @@ The goal of the fork is a **100% open, AGPL-only build with no Enterprise-Editio
 ### Embedded MCP server
 
 Gitmost has **our own MCP server** — [docmost-mcp](https://github.com/vvzvlad/docmost-mcp),
-which we wrote — **built directly into the app** and served at `/mcp`. It exposes **38
+which we wrote — **built directly into the app** and served at `/mcp`. It exposes **39
 agent-native tools**: surgical per-block edits (patch / insert / delete by id),
 structure-preserving find/replace, scripted `(doc) => doc` transforms with a dry-run diff,
 structured table editing, version history with diff / restore, comments, images and share
@@ -60,7 +60,7 @@ every little fix. And it needs no enterprise license.
 | | **Gitmost `/mcp` (our docmost-mcp)** | Docmost's built-in MCP |
 | --- | :---: | :---: |
 | **Enterprise license** | Not required | Required |
-| **Tools** | 38, agent-native | Coarse (read Markdown, page CRUD, replace whole page) |
+| **Tools** | 39, agent-native | Coarse (read Markdown, page CRUD, replace whole page) |
 | **Per-block edits / find-replace / scripted transforms** | ✅ | — |
 | **Structured table editing, version diff / restore** | ✅ | — |
 | **Comments, images, share links** | ✅ | — |

README.ru.md

@@ -33,7 +33,7 @@
 | --- | --- |
 | **Удалён EE-код** | Вырезан весь код Enterprise-редакции на клиенте и сервере; это чистая community/AGPL-сборка без лицензионных проверок. |
 | **Резолв комментариев** | Переписан с нуля как community-функция (резолв / переоткрытие с вкладками «Открытые» / «Решённые»). EE-код не используется, доступно любому, кто может комментировать. |
-| **Встроенный MCP-сервер** | Community MCP-сервер (`@docmost/mcp`, 38 инструментов) отдаётся по HTTP на `/mcp` — без enterprise-лицензии. Заменяет удалённый лицензируемый EE MCP. |
+| **Встроенный MCP-сервер** | Community MCP-сервер (`@docmost/mcp`, 39 инструментов) отдаётся по HTTP на `/mcp` — без enterprise-лицензии. Заменяет удалённый лицензируемый EE MCP. |
 | **Чат с AI-агентом** | Встроенный чат с AI-агентом по содержимому вики, написанный с нуля как community-функция — без enterprise-лицензии. Агент читает и редактирует страницы от вашего имени (в рамках ваших прав), с полнотекстовым + векторным (RAG) поиском и опциональным доступом в интернет через внешние MCP-серверы. |
 | **Ребрендинг** | Логотип / название приложения изменены с *Docmost* на *Gitmost*. |
 | **Компактное дерево страниц** | Отступ дерева страниц по умолчанию уменьшен с 16px до 8px на уровень вложенности. |
@@ -44,7 +44,7 @@
 
 В Gitmost есть **наш собственный MCP-сервер** — [docmost-mcp](https://github.com/vvzvlad/docmost-mcp),
 который мы написали сами, — **встроенный прямо в приложение** и доступный на `/mcp`. Он даёт
-**38 agent-native инструментов**: точечное редактирование по блокам (patch / insert / delete
+**39 agent-native инструментов**: точечное редактирование по блокам (patch / insert / delete
 по id), find/replace с сохранением структуры, скриптовые трансформации `(doc) => doc` с
 предпросмотром диффа, структурное редактирование таблиц, история версий с диффом /
 восстановлением, комментарии, изображения и ссылки на шаринг — всё применяется через слой
@@ -60,7 +60,7 @@ real-time-коллаборации Docmost, поэтому запись нико
 | | **`/mcp` в Gitmost (наш docmost-mcp)** | Родной MCP у Docmost |
 | --- | :---: | :---: |
 | **Enterprise-лицензия** | Не нужна | Нужна |
-| **Инструменты** | 38, agent-native | Примитивные (Markdown, CRUD страниц, замена целиком) |
+| **Инструменты** | 39, agent-native | Примитивные (Markdown, CRUD страниц, замена целиком) |
 | **Правки по блокам / find-replace / скриптовые трансформации** | ✅ | — |
 | **Структурное редактирование таблиц, дифф / восстановление версий** | ✅ | — |
 | **Комментарии, изображения, ссылки на шаринг** | ✅ | — |

agent-roles-catalog/README.md

@@ -0,0 +1,193 @@
+# Agent roles catalog
+
+This directory is **data, not application code**. It holds the content of an
+"agent roles catalog": reusable agent role definitions (system prompts plus a
+little metadata), grouped into bundles and translated into one or more
+languages. A separate server reads these files and serves them; nothing here is
+executable application logic except the validation script.
+
+## File layout
+
+```
+agent-roles-catalog/
+  index.json                  # the catalog manifest: bundles, languages, role versions
+  bundles/
+    <bundle-id>/
+      <lang>.json             # one file per declared language (e.g. ru.json, en.json)
+  scripts/
+    check.mjs                 # validates the catalog (no dependencies)
+    content-hashes.json       # check artifact: per-role content-hash lock (NOT served)
+  package.json                # defines the `check` script
+  README.md
+```
+
+Currently shipped bundles:
+
+- `editorial` — the editorial suite (structural-editor, line-editor,
+  fact-checker, proofreader, narrator), languages `ru`, `en`.
+- `research` — a single `researcher` role, languages `ru`, `en`.
+
+## How it's served
+
+The server does not bundle this data; it reads it at request time from a single
+configured location, the `AI_AGENT_ROLES_CATALOG_URL` env var
+(`EnvironmentService.getAiAgentRolesCatalogSource()`), an `http(s)://` base URL
+to the catalog's raw files. The server fetches `<base>/index.json` for the
+manifest and `<base>/bundles/<bundle-id>/<lang>.json` for each opened bundle
+file (REMOTE only).
+
+That base URL is provided as a per-branch default in the Docker image (set in
+CI: a `develop` build points at the `develop` raw URL, a release build at the
+`main` raw URL) and can be overridden at runtime via the
+`AI_AGENT_ROLES_CATALOG_URL` env var. Local-filesystem sources are no longer
+supported; if the value is unset the catalog is unavailable.
+
+The fetched JSON is re-validated server-side (the catalog is treated as
+untrusted input). See `.env.example` for the variable and the CHANGELOG for the
+rollout.
+
+## `index.json` schema
+
+```jsonc
+{
+  "schemaVersion": 1,
+  "bundles": [
+    {
+      "id": "editorial",                       // unique bundle id; matches bundles/<id>/
+      "name": { "ru": "...", "en": "..." },    // localized display name
+      "description": { "ru": "...", "en": "..." },
+      "languages": ["ru", "en"],               // which <lang>.json files must exist
+      "roles": [
+        { "slug": "structural-editor", "version": 1 }
+        // ...
+      ]
+    }
+  ]
+}
+```
+
+`version` lives **here, in index.json**, per role. Bump it whenever a role's
+content (instructions, name, description, etc.) changes, so consumers can detect
+updates.
+
+## Bundle (`<lang>.json`) schema
+
+```jsonc
+{
+  "schemaVersion": 1,
+  "language": "ru",
+  "roles": [
+    {
+      "slug": "structural-editor",   // REQUIRED, unique across the whole catalog
+      "emoji": "🧱",
+      "name": "...",                 // REQUIRED, localized
+      "description": "...",          // localized
+      "instructions": "...",         // REQUIRED, the system prompt, localized
+      "autoStart": true,             // whether the role starts working immediately
+      "launchMessage": "..."         // first message sent on launch (or null)
+    }
+  ]
+}
+```
+
+Notes:
+
+- `modelConfig` is intentionally absent; the server treats an absent
+  `modelConfig` as `null`.
+- A role's `slug`, `emoji`, and `autoStart` are identical across all language
+  files of the same bundle. Only `name`, `description`, `instructions`, and
+  `launchMessage` are translated.
+
+## Slug uniqueness
+
+**Every `slug` must be UNIQUE ACROSS THE WHOLE CATALOG**, not just within a
+bundle. A slug appears once per language file of its bundle (same slug in
+`ru.json` and `en.json`), but no two different bundles may share a slug.
+`scripts/check.mjs` enforces this.
+
+## How to add things
+
+### Add a role to an existing bundle
+
+1. Add an entry to that bundle's `roles[]` in `index.json` with a new unique
+   `slug` and `version: 1`.
+2. Add a role object with the same `slug` to **every** `<lang>.json` of the
+   bundle, translating `name`, `description`, `instructions`, and
+   `launchMessage`.
+3. Run the check (see below).
+
+### Add a bundle
+
+1. Add a bundle object to `index.json` (`id`, `name`, `description`,
+   `languages`, `roles`).
+2. Create `bundles/<id>/<lang>.json` for each declared language, with one role
+   object per `roles[]` entry.
+3. Run the check.
+
+### Add a language to a bundle
+
+1. Add the language code to that bundle's `languages[]` in `index.json`.
+2. Create `bundles/<id>/<lang>.json` containing every role of the bundle,
+   translated.
+3. Run the check.
+
+### Change a role's content
+
+Edit the role in the relevant `<lang>.json` file(s) and **bump that role's
+`version`** in `index.json`. Then run `node scripts/check.mjs --update-hashes`
+to refresh the content-hash lock (`scripts/content-hashes.json`). `check.mjs`
+now **fails if a role's content changed but its `version` was not bumped**, so
+this step is mandatory — the lock can only be refreshed after the bump.
+
+## Validating
+
+From this directory:
+
+```sh
+node scripts/check.mjs   # or: npm run check
+```
+
+It fails (exit code 1) if any slug is duplicated across the catalog, if a
+bundle's index `roles[]` don't match the slugs present in each language file, if
+a declared language file is missing, or if any role is missing a required field
+(`slug`, `name`, `instructions`). It prints `OK` on success.
+
+### Content-hash guard
+
+`check.mjs` also guards against changing a role's content without bumping its
+`version`. It keeps a lockfile, `scripts/content-hashes.json`, mapping each role
+`slug` to `{ version, hash }`, where `hash` is a SHA-256 over the role's
+content fields (`emoji`, `autoStart`, `name`, `description`, `instructions`,
+`launchMessage`) across all of its language files, in a deterministic canonical
+form. This lockfile is a **check artifact only** — the server fetches only
+`index.json` and the bundle `<lang>.json` files, never this file, so it has no
+effect on the served catalog or its schema.
+
+On a normal run, for every role the check recomputes the hash and compares it
+against the lock:
+
+- content unchanged and versions agree → OK;
+- content changed but `version` not bumped above the lock → **error** asking you
+  to bump and refresh;
+- content changed and `version` bumped → **error** asking you to record it by
+  refreshing the lock;
+- role missing from the lock, or a lock entry for a role that no longer exists →
+  **error** asking you to refresh.
+
+Refresh the lock with:
+
+```sh
+node scripts/check.mjs --update-hashes   # alias: --fix
+```
+
+This recomputes the lock from the current catalog, prunes entries for removed
+roles, and prints what changed — but it **refuses to write** (exit 1) if any
+role's content changed while its `index.json` version was not bumped, so the
+version bump is always enforced first. The check also requires every
+`index.json` role to carry a finite numeric `version` (the server requires the
+same).
+
+Known, accepted limitation: a deliberate prune-then-readd of a slug (remove the
+role and run `--update-hashes`, then re-add it with changed content at the same
+version) is **not** caught, because a brand-new slug has no lock baseline to
+enforce a bump against.

agent-roles-catalog/index.json

@@ -0,0 +1,31 @@
+{
+  "schemaVersion": 1,
+  "bundles": [
+    {
+      "id": "editorial",
+      "name": { "ru": "Редакторский набор", "en": "Editorial suite" },
+      "description": {
+        "ru": "Полный цикл редактуры статьи: структура, стиль, корректура, факты и нарратив.",
+        "en": "The full article-editing cycle: structure, style, copyediting, facts, and narrative."
+      },
+      "languages": ["ru", "en"],
+      "roles": [
+        { "slug": "structural-editor", "version": 2 },
+        { "slug": "line-editor", "version": 2 },
+        { "slug": "fact-checker", "version": 2 },
+        { "slug": "proofreader", "version": 3 },
+        { "slug": "narrator", "version": 1 }
+      ]
+    },
+    {
+      "id": "research",
+      "name": { "ru": "Исследование", "en": "Research" },
+      "description": {
+        "ru": "Глубокое исследование темы с подготовкой отчёта.",
+        "en": "Deep research on a topic with a prepared report."
+      },
+      "languages": ["ru", "en"],
+      "roles": [ { "slug": "researcher", "version": 1 } ]
+    }
+  ]
+}

agent-roles-catalog/package.json

@@ -0,0 +1,8 @@
+{
+  "name": "agent-roles-catalog",
+  "private": true,
+  "type": "module",
+  "scripts": {
+    "check": "node scripts/check.mjs"
+  }
+}

agent-roles-catalog/scripts/check.mjs

@@ -0,0 +1,353 @@
+#!/usr/bin/env node
+// Validates the agent roles catalog.
+// Fails (exit 1) on: duplicate slugs across the whole catalog, mismatches
+// between a bundle's index roles[] and the slugs present in each language
+// file, a missing declared language file, or a role missing required fields.
+
+import { readFileSync, writeFileSync, existsSync } from "node:fs";
+import { createHash } from "node:crypto";
+import { fileURLToPath } from "node:url";
+import { dirname, join } from "node:path";
+
+const __dirname = dirname(fileURLToPath(import.meta.url));
+const catalogDir = join(__dirname, "..");
+
+// `--update-hashes` (alias `--fix`) recomputes the content-hash lockfile from
+// the current catalog instead of just validating against it.
+const updateHashes =
+  process.argv.includes("--update-hashes") || process.argv.includes("--fix");
+
+// The content-hash lockfile lives under scripts/ and is a CHECK ARTIFACT only:
+// the server never fetches it, so it has zero impact on the served schema.
+const lockPath = join(__dirname, "content-hashes.json");
+
+const errors = [];
+
+function readJson(path) {
+  try {
+    return JSON.parse(readFileSync(path, "utf8"));
+  } catch (err) {
+    errors.push(`Cannot read/parse ${path}: ${err.message}`);
+    return null;
+  }
+}
+
+const indexPath = join(catalogDir, "index.json");
+if (!existsSync(indexPath)) {
+  console.error(`Missing index.json at ${indexPath}`);
+  process.exit(1);
+}
+
+const index = readJson(indexPath);
+if (!index) {
+  for (const e of errors) console.error(e);
+  process.exit(1);
+}
+
+const bundles = Array.isArray(index.bundles) ? index.bundles : [];
+if (bundles.length === 0) {
+  errors.push("index.json has no bundles[]");
+}
+
+// Track every slug seen across the whole catalog to detect duplicates.
+const slugSeen = new Map(); // slug -> "bundleId/lang"
+
+for (const bundle of bundles) {
+  const bundleId = bundle.id;
+  if (!bundleId) {
+    errors.push("A bundle in index.json is missing an id");
+    continue;
+  }
+
+  const indexSlugs = (bundle.roles || []).map((r) => r.slug);
+  // Duplicate slugs inside the bundle index roles[].
+  const indexSlugSet = new Set(indexSlugs);
+  if (indexSlugSet.size !== indexSlugs.length) {
+    errors.push(`Bundle "${bundleId}" index.json roles[] contains duplicate slugs`);
+  }
+
+  // Each index role must carry a finite numeric "version". The server requires
+  // this (see ai-agent-roles-catalog.provider.ts), and the content-hash guard
+  // below relies on it for the bump comparison, so enforce it here too.
+  for (const r of bundle.roles || []) {
+    if (typeof r.version !== "number" || !Number.isFinite(r.version)) {
+      errors.push(
+        `Bundle "${bundleId}" index.json role "${r.slug}" is missing a numeric "version"`
+      );
+    }
+  }
+
+  const languages = Array.isArray(bundle.languages) ? bundle.languages : [];
+  if (languages.length === 0) {
+    errors.push(`Bundle "${bundleId}" declares no languages`);
+  }
+
+  for (const lang of languages) {
+    const langPath = join(catalogDir, "bundles", bundleId, `${lang}.json`);
+    if (!existsSync(langPath)) {
+      errors.push(`Bundle "${bundleId}" declares language "${lang}" but ${langPath} is missing`);
+      continue;
+    }
+
+    const langFile = readJson(langPath);
+    if (!langFile) continue;
+
+    const roles = Array.isArray(langFile.roles) ? langFile.roles : [];
+    const fileSlugs = roles.map((r) => r && r.slug);
+
+    // (d) Required fields per role.
+    for (const role of roles) {
+      for (const field of ["slug", "name", "instructions"]) {
+        if (role == null || role[field] == null || role[field] === "") {
+          errors.push(
+            `Bundle "${bundleId}/${lang}" has a role missing required field "${field}" (slug=${role && role.slug})`
+          );
+        }
+      }
+    }
+
+    // (b) index roles[] must match the slugs present in each language file.
+    const fileSlugSet = new Set(fileSlugs);
+    const missingInFile = indexSlugs.filter((s) => !fileSlugSet.has(s));
+    const extraInFile = fileSlugs.filter((s) => !indexSlugSet.has(s));
+    if (missingInFile.length > 0) {
+      errors.push(
+        `Bundle "${bundleId}/${lang}" is missing roles declared in index.json: ${missingInFile.join(", ")}`
+      );
+    }
+    if (extraInFile.length > 0) {
+      errors.push(
+        `Bundle "${bundleId}/${lang}" has roles not declared in index.json: ${extraInFile.join(", ")}`
+      );
+    }
+
+    // (a) Duplicate slugs across the whole catalog.
+    for (const slug of fileSlugs) {
+      if (!slug) continue;
+      const where = `${bundleId}/${lang}`;
+      // Only flag duplicates across DIFFERENT bundles or files; the same slug
+      // is expected to appear once per language file of the same bundle.
+      if (slugSeen.has(slug)) {
+        const prev = slugSeen.get(slug);
+        const prevBundle = prev.split("/")[0];
+        if (prevBundle !== bundleId) {
+          errors.push(
+            `Slug "${slug}" is duplicated across the catalog: ${prev} and ${where}`
+          );
+        }
+      } else {
+        slugSeen.set(slug, where);
+      }
+    }
+  }
+}
+
+// ---------------------------------------------------------------------------
+// Content-hash guard: detect "content changed without a version bump".
+//
+// check.mjs cannot use git history, so we maintain a lockfile
+// (scripts/content-hashes.json) mapping each role slug to its recorded
+// { version, hash }. On every run we recompute each role's content hash and
+// compare it against the lock; a content change is only allowed once the role's
+// version in index.json has been bumped and the lock refreshed.
+//
+// Known, accepted limitation: a deliberate prune-then-readd of a slug (remove
+// the role and run --update-hashes, then re-add it with changed content at the
+// same version) is NOT caught, because a brand-new slug has no lock baseline to
+// enforce a bump against. We document this rather than building tombstones.
+// ---------------------------------------------------------------------------
+
+// Content fields hashed for each role, in a fixed canonical order. `slug` is
+// identity (not content) and `version` lives in index.json, so neither is here.
+// `modelConfig` (an OPTIONAL role field the server also serves) is intentionally
+// EXCLUDED: no shipped role uses it today, and being an object it would need a
+// deterministic deep canonicalization (recursive key sort) before hashing —
+// otherwise JSON.stringify key-order would make the hash non-deterministic. If a
+// role ever gains a `modelConfig`, add it here WITH such canonicalization so a
+// change to it is still caught by the bump guard.
+const CONTENT_FIELDS = [
+  "emoji",
+  "autoStart",
+  "name",
+  "description",
+  "instructions",
+  "launchMessage",
+];
+
+// Build a map of slug -> { version, langRoles: { lang: roleObject } } from the
+// current catalog so we can compute hashes and read index versions.
+function collectCatalogRoles() {
+  const out = new Map(); // slug -> { version, langRoles: Map<lang, role> }
+  for (const bundle of bundles) {
+    const bundleId = bundle.id;
+    if (!bundleId) continue;
+    const languages = Array.isArray(bundle.languages) ? bundle.languages : [];
+    for (const r of bundle.roles || []) {
+      if (!r || !r.slug) continue;
+      if (!out.has(r.slug)) {
+        out.set(r.slug, { version: r.version, langRoles: new Map() });
+      } else {
+        // Same slug declared twice in index.json roles[]; already flagged above.
+        out.get(r.slug).version = r.version;
+      }
+    }
+    for (const lang of languages) {
+      const langPath = join(catalogDir, "bundles", bundleId, `${lang}.json`);
+      if (!existsSync(langPath)) continue;
+      const langFile = readJson(langPath);
+      if (!langFile) continue;
+      const roles = Array.isArray(langFile.roles) ? langFile.roles : [];
+      for (const role of roles) {
+        if (!role || !role.slug) continue;
+        const entry = out.get(role.slug);
+        if (!entry) continue; // role not declared in index.json; flagged above.
+        entry.langRoles.set(lang, role);
+      }
+    }
+  }
+  return out;
+}
+
+// Deterministic content hash for a role: languages sorted ascending, each
+// language's content fields taken in CONTENT_FIELDS order (null when absent).
+function contentHash(langRoles) {
+  const langs = [...langRoles.keys()].sort();
+  const canonical = langs.map((lang) => {
+    const role = langRoles.get(lang);
+    const fields = {};
+    for (const field of CONTENT_FIELDS) {
+      fields[field] = role && role[field] != null ? role[field] : null;
+    }
+    return [lang, fields];
+  });
+  return createHash("sha256").update(JSON.stringify(canonical)).digest("hex");
+}
+
+// Compute current { version, hash } for every catalog role.
+const catalogRoles = collectCatalogRoles();
+const current = new Map(); // slug -> { version, hash }
+for (const [slug, entry] of catalogRoles) {
+  current.set(slug, {
+    version: entry.version,
+    hash: contentHash(entry.langRoles),
+  });
+}
+
+// Load the existing lock (may be absent on first run).
+let lock = {};
+if (existsSync(lockPath)) {
+  const parsed = readJson(lockPath);
+  if (parsed && typeof parsed === "object") lock = parsed;
+}
+
+if (updateHashes) {
+  // Refresh the lock from the current catalog, but refuse to write if any role's
+  // content changed without its version being bumped above the existing lock.
+  const blockers = [];
+  for (const [slug, cur] of current) {
+    const prev = lock[slug];
+    if (!prev) continue; // new role; nothing to enforce a bump against.
+    if (cur.hash === prev.hash) continue; // content unchanged.
+    // Defense-in-depth: a non-numeric version must never pass the bump check via
+    // `undefined <= N` (which is false). The standard checks already flag a
+    // missing numeric version, but guard here too before comparing.
+    if (typeof cur.version !== "number" || !Number.isFinite(cur.version)) {
+      blockers.push(
+        `role "${slug}" content changed but its index.json "version" is missing or not numeric; set a numeric "version" before refreshing the lock`
+      );
+    } else if (cur.version <= prev.version) {
+      blockers.push(
+        `role "${slug}" content changed but its version was not bumped (still ${prev.version}); bump "version" in index.json before refreshing the lock`
+      );
+    }
+  }
+  // Still honor the standard checks before allowing a write.
+  if (errors.length > 0) {
+    console.error("Catalog check FAILED:");
+    for (const e of errors) console.error(`  - ${e}`);
+    process.exit(1);
+  }
+  if (blockers.length > 0) {
+    console.error("Refusing to update content-hash lock:");
+    for (const b of blockers) console.error(`  - ${b}`);
+    process.exit(1);
+  }
+
+  // Compute the change summary relative to the old lock, pruning removed slugs.
+  const newLock = {};
+  const added = [];
+  const changed = [];
+  const removed = [];
+  for (const [slug, cur] of [...current].sort((a, b) => a[0].localeCompare(b[0]))) {
+    newLock[slug] = { version: cur.version, hash: cur.hash };
+    const prev = lock[slug];
+    if (!prev) added.push(slug);
+    else if (prev.hash !== cur.hash || prev.version !== cur.version) changed.push(slug);
+  }
+  for (const slug of Object.keys(lock)) {
+    if (!current.has(slug)) removed.push(slug);
+  }
+  writeFileSync(lockPath, JSON.stringify(newLock, null, 2) + "\n");
+  console.log(`Wrote ${lockPath}`);
+  if (added.length) console.log(`  added:   ${added.join(", ")}`);
+  if (changed.length) console.log(`  updated: ${changed.join(", ")}`);
+  if (removed.length) console.log(`  pruned:  ${removed.join(", ")}`);
+  if (!added.length && !changed.length && !removed.length) {
+    console.log("  (no changes; lock already up to date)");
+  }
+  console.log("OK");
+  process.exit(0);
+}
+
+// Normal run: validate current content against the lock.
+for (const [slug, cur] of current) {
+  const prev = lock[slug];
+  if (!prev) {
+    errors.push(
+      `role "${slug}" is not recorded in the content-hash lock; run: node scripts/check.mjs --update-hashes`
+    );
+    continue;
+  }
+  if (cur.hash === prev.hash) {
+    // Content unchanged; the lock version must still agree with index.json.
+    if (cur.version !== prev.version) {
+      errors.push(
+        `role "${slug}" content is unchanged but its index.json version (${cur.version}) differs from the lock (${prev.version}); run: node scripts/check.mjs --update-hashes`
+      );
+    }
+    continue;
+  }
+  // Content changed.
+  // Defense-in-depth: treat a non-numeric version as an error before the `<=`
+  // comparison, so a missing version can never silently pass the bump check
+  // (and we avoid a misleading "version bumped to undefined" message).
+  if (typeof cur.version !== "number" || !Number.isFinite(cur.version)) {
+    errors.push(
+      `role "${slug}" content changed but its index.json "version" is missing or not numeric; set a numeric "version", then run: node scripts/check.mjs --update-hashes`
+    );
+  } else if (cur.version <= prev.version) {
+    errors.push(
+      `role "${slug}" content changed but its version was not bumped (still ${prev.version}); bump "version" in index.json, then run: node scripts/check.mjs --update-hashes`
+    );
+  } else {
+    errors.push(
+      `role "${slug}" content changed and version bumped to ${cur.version}; record it by running: node scripts/check.mjs --update-hashes`
+    );
+  }
+}
+// Lock entries for slugs that no longer exist in the catalog.
+for (const slug of Object.keys(lock)) {
+  if (!current.has(slug)) {
+    errors.push(
+      `content-hash lock has entry for unknown role "${slug}" (no longer in the catalog); run: node scripts/check.mjs --update-hashes`
+    );
+  }
+}
+
+if (errors.length > 0) {
+  console.error("Catalog check FAILED:");
+  for (const e of errors) console.error(`  - ${e}`);
+  process.exit(1);
+}
+
+console.log("OK");

agent-roles-catalog/scripts/content-hashes.json

@@ -0,0 +1,26 @@
+{
+  "fact-checker": {
+    "version": 2,
+    "hash": "d7ad1dae07d6f4321e7d40c5b36259dbf930264d748834809c4fb77294bf72e3"
+  },
+  "line-editor": {
+    "version": 2,
+    "hash": "cca324110dc6f96d2a8a239a2fb95b0ba09fad5806c9b6090a3c210ea7883ceb"
+  },
+  "narrator": {
+    "version": 1,
+    "hash": "36b38785fea6ae1c70bf6fb6b29ae5278bb86e389e61f7b9736675a589fa434c"
+  },
+  "proofreader": {
+    "version": 3,
+    "hash": "a36047c5cab837b2a727f63d4ddafc269b1fc44b90b365e770ecdb8f77e13952"
+  },
+  "researcher": {
+    "version": 1,
+    "hash": "853658fda43ddbe0a4d08f2c6e50b5116d29a2e9ccd7f46e173e65920d8f6ace"
+  },
+  "structural-editor": {
+    "version": 2,
+    "hash": "83093baa7262aef8193871a1afcf2b43b11a56fe2d00cade41355cf66d972b74"
+  }
+}

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

@@ -1333,6 +1333,7 @@
   "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",
+  "This address is in use. Saving will move it to this page.": "This address is in use. Saving will move it to this page.",
   "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?",
@@ -1346,5 +1347,22 @@
   "Could not generate a title": "Could not generate a title",
   "AI title generation is disabled": "AI title generation is disabled",
   "AI is not configured": "AI is not configured",
-  "Too many requests, please try again later": "Too many requests, please try again later"
+  "Too many requests, please try again later": "Too many requests, please try again later",
+  "Import from catalog": "Import from catalog",
+  "Browse the catalog": "Browse the catalog",
+  "Role catalog": "Role catalog",
+  "On name conflict": "On name conflict",
+  "Skip": "Skip",
+  "Import": "Import",
+  "Installed": "Installed",
+  "v{{from}} → v{{to}}": "v{{from}} → v{{to}}",
+  "Imported {{created}}, renamed {{renamed}}, skipped {{skipped}}": "Imported {{created}}, renamed {{renamed}}, skipped {{skipped}}",
+  "Failed to import {{count}} role(s)": "Failed to import {{count}} role(s)",
+  "The role catalog is unavailable": "The role catalog is unavailable",
+  "Please try again later.": "Please try again later.",
+  "No bundles available": "No bundles available",
+  "Already up to date": "Already up to date",
+  "Updated to the latest version": "Updated to the latest version",
+  "This role is no longer in the catalog": "This role is no longer in the catalog",
+  "This language is no longer available in the catalog": "This language is no longer available in the catalog"
 }

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

@@ -1190,6 +1190,7 @@
   "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",
+  "This address is in use. Saving will move it to this page.": "Этот адрес уже используется. При сохранении он будет перемещён на эту страницу.",
   "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?",
@@ -1203,5 +1204,23 @@
   "Could not generate a title": "Не удалось придумать название",
   "AI title generation is disabled": "Генерация названий через AI отключена",
   "AI is not configured": "AI не настроен",
-  "Too many requests, please try again later": "Слишком много запросов, попробуйте позже"
+  "Too many requests, please try again later": "Слишком много запросов, попробуйте позже",
+  "Import from catalog": "Импорт из каталога",
+  "Browse the catalog": "Открыть каталог",
+  "Role catalog": "Каталог ролей",
+  "On name conflict": "При конфликте имён",
+  "Skip": "Пропустить",
+  "Import": "Импортировать",
+  "Installed": "Установлено",
+  "v{{from}} → v{{to}}": "v{{from}} → v{{to}}",
+  "Imported {{created}}, renamed {{renamed}}, skipped {{skipped}}": "Импортировано: {{created}}, переименовано: {{renamed}}, пропущено: {{skipped}}",
+  "Failed to import {{count}} role(s)": "Не удалось импортировать ролей: {{count}}",
+  "The role catalog is unavailable": "Каталог ролей недоступен",
+  "Please try again later.": "Попробуйте позже.",
+  "No bundles available": "Наборы недоступны",
+  "No roles configured": "Роли не настроены",
+  "Already up to date": "Уже актуальна",
+  "Updated to the latest version": "Обновлено до последней версии",
+  "This role is no longer in the catalog": "Эта роль больше не представлена в каталоге",
+  "This language is no longer available in the catalog": "Этот язык больше не доступен в каталоге"
 }

apps/client/src/components/layouts/global/app-header.tsx

@@ -10,12 +10,12 @@ import classes from "./app-header.module.css";
 import { BrandLogo } from "@/components/ui/brand-logo";
 import TopMenu from "@/components/layouts/global/top-menu.tsx";
 import { Link } from "react-router-dom";
-import { useAtom, useSetAtom } from "jotai";
+import { useAtom } from "jotai";
 import {
   desktopSidebarAtom,
   mobileSidebarAtom,
 } from "@/components/layouts/global/hooks/atoms/sidebar-atom.ts";
-import { aiChatWindowOpenAtom } from "@/features/ai-chat/atoms/ai-chat-atom.ts";
+import { useOpenAiChatForCurrentPage } from "@/features/ai-chat/hooks/use-open-ai-chat.ts";
 import { workspaceAtom } from "@/features/user/atoms/current-user-atom.ts";
 import { useToggleSidebar } from "@/components/layouts/global/hooks/hooks/use-toggle-sidebar.ts";
 import SidebarToggle from "@/components/ui/sidebar-toggle-button.tsx";
@@ -38,7 +38,9 @@ export function AppHeader() {
   const toggleDesktop = useToggleSidebar(desktopSidebarAtom);
 
   const [workspace] = useAtom(workspaceAtom);
-  const setAiChatWindowOpen = useSetAtom(aiChatWindowOpenAtom);
+  // Opening from the header auto-opens the document's bound chat (last chat
+  // created on the current page); off a page it keeps the current selection.
+  const openAiChat = useOpenAiChatForCurrentPage();
   // AI chat entry point: only shown when the workspace enables it (A7 gate).
   const aiChatEnabled = workspace?.settings?.ai?.chat === true;
 
@@ -105,7 +107,7 @@ export function AppHeader() {
                 color="dark"
                 size="sm"
                 aria-label={t("AI chat")}
-                onClick={() => setAiChatWindowOpen((v) => !v)}
+                onClick={openAiChat}
               >
                 <IconMessage size={20} />
               </ActionIcon>

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

@@ -0,0 +1,135 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import { renderHook, act } from "@testing-library/react";
+import { Provider, createStore } from "jotai";
+import type { ReactNode } from "react";
+import { useOpenAiChatForCurrentPage } from "./use-open-ai-chat";
+import {
+  activeAiChatIdAtom,
+  aiChatWindowOpenAtom,
+  aiChatDraftAtom,
+  selectedAiRoleIdAtom,
+} from "@/features/ai-chat/atoms/ai-chat-atom.ts";
+
+// useMatch is the only react-router-dom export the hook uses; drive its return
+// per test to simulate "on a page" vs "off a page".
+const useMatchMock = vi.fn();
+vi.mock("react-router-dom", () => ({
+  useMatch: () => useMatchMock(),
+}));
+
+// The bound-chat resolver is the network boundary; stub it per test.
+const getBoundChatMock = vi.fn();
+vi.mock("@/features/ai-chat/services/ai-chat-service.ts", () => ({
+  getBoundChat: (pageId: string) => getBoundChatMock(pageId),
+}));
+
+// Put the hook on a page route by default ("doc-p1" -> page id "p1"); individual
+// tests override useMatch to go off-page.
+function onPage(pageSlug = "doc-p1") {
+  useMatchMock.mockReturnValue({ params: { pageSlug } });
+}
+function offPage() {
+  useMatchMock.mockReturnValue(null);
+}
+
+// Render the hook inside an explicit jotai store so atom side effects are
+// assertable; the store is returned for setup + assertions.
+function setup(seed?: (store: ReturnType<typeof createStore>) => void) {
+  const store = createStore();
+  seed?.(store);
+  const wrapper = ({ children }: { children: ReactNode }) => (
+    <Provider store={store}>{children}</Provider>
+  );
+  const { result } = renderHook(() => useOpenAiChatForCurrentPage(), { wrapper });
+  return { store, open: () => act(() => result.current()) };
+}
+
+describe("useOpenAiChatForCurrentPage", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+    onPage();
+  });
+
+  it("on a page: resolves the bound chat, selects it, and opens the window", async () => {
+    getBoundChatMock.mockResolvedValue("bound-chat-1");
+    const { store, open } = setup((s) => s.set(aiChatDraftAtom, "stale draft"));
+
+    await open();
+
+    expect(getBoundChatMock).toHaveBeenCalledWith("p1");
+    expect(store.get(activeAiChatIdAtom)).toBe("bound-chat-1");
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+    expect(store.get(aiChatDraftAtom)).toBe(""); // cleared on a real switch
+  });
+
+  it("on a page with no bound chat: opens a fresh chat (null)", async () => {
+    getBoundChatMock.mockResolvedValue(null);
+    const { store, open } = setup((s) => s.set(activeAiChatIdAtom, "previous"));
+
+    await open();
+
+    expect(store.get(activeAiChatIdAtom)).toBeNull();
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+  });
+
+  it("off a page: keeps the current selection and does NOT resolve", async () => {
+    offPage();
+    const { store, open } = setup((s) => {
+      s.set(activeAiChatIdAtom, "keep-me");
+      s.set(aiChatDraftAtom, "untouched");
+    });
+
+    await open();
+
+    expect(getBoundChatMock).not.toHaveBeenCalled();
+    expect(store.get(activeAiChatIdAtom)).toBe("keep-me");
+    expect(store.get(aiChatDraftAtom)).toBe("untouched"); // no switch -> kept
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+  });
+
+  it("window already open: re-click does NOT re-resolve or switch chats", async () => {
+    getBoundChatMock.mockResolvedValue("would-switch");
+    const { store, open } = setup((s) => {
+      s.set(aiChatWindowOpenAtom, true);
+      s.set(activeAiChatIdAtom, "current");
+    });
+
+    await open();
+
+    expect(getBoundChatMock).not.toHaveBeenCalled();
+    expect(store.get(activeAiChatIdAtom)).toBe("current");
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+  });
+
+  it("does NOT clear the draft when the resolved chat equals the current one", async () => {
+    getBoundChatMock.mockResolvedValue("same");
+    const { store, open } = setup((s) => {
+      s.set(activeAiChatIdAtom, "same");
+      s.set(aiChatDraftAtom, "in-progress");
+    });
+
+    await open();
+
+    expect(store.get(aiChatDraftAtom)).toBe("in-progress"); // no switch
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+  });
+
+  it("fail-soft: a resolve error opens a fresh chat (null)", async () => {
+    getBoundChatMock.mockRejectedValue(new Error("network"));
+    const { store, open } = setup((s) => s.set(activeAiChatIdAtom, "previous"));
+
+    await open();
+
+    expect(store.get(activeAiChatIdAtom)).toBeNull();
+    expect(store.get(aiChatWindowOpenAtom)).toBe(true);
+  });
+
+  it("clears the picked role on a real switch", async () => {
+    getBoundChatMock.mockResolvedValue("bound");
+    const { store, open } = setup((s) => s.set(selectedAiRoleIdAtom, "role-1"));
+
+    await open();
+
+    expect(store.get(selectedAiRoleIdAtom)).toBeNull();
+  });
+});

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

@@ -0,0 +1,67 @@
+import { useCallback } from "react";
+import { useAtom, useSetAtom } from "jotai";
+import { useMatch } from "react-router-dom";
+import {
+  aiChatWindowOpenAtom,
+  activeAiChatIdAtom,
+  aiChatDraftAtom,
+  selectedAiRoleIdAtom,
+} from "@/features/ai-chat/atoms/ai-chat-atom.ts";
+import { getBoundChat } from "@/features/ai-chat/services/ai-chat-service.ts";
+import { extractPageSlugId } from "@/lib";
+
+/**
+ * The generic "open the AI chat" action, WITH document binding: when invoked
+ * while viewing a page, it resolves that page's bound chat and selects it before
+ * opening — so the last chat for this document re-opens by itself. With no bound
+ * chat (or off a page) it keeps the current selection / opens a fresh chat. Used
+ * by the app-header entry point; NOT by the provenance badge (which deep-links).
+ */
+export function useOpenAiChatForCurrentPage() {
+  const [windowOpen, setWindowOpen] = useAtom(aiChatWindowOpenAtom);
+  const [activeChatId, setActiveChatId] = useAtom(activeAiChatIdAtom);
+  const setDraft = useSetAtom(aiChatDraftAtom);
+  const setSelectedRoleId = useSetAtom(selectedAiRoleIdAtom);
+
+  // Same route-match trick the window uses: read :pageSlug from the pathname.
+  // AiChatWindow lives in a pathless parent layout route, so useParams() can't
+  // see :pageSlug — match the full path against the authenticated page route.
+  const match = useMatch("/s/:spaceSlug/p/:pageSlug");
+  const pageId = extractPageSlugId(match?.params?.pageSlug);
+
+  return useCallback(async () => {
+    // Re-clicks while the window is already open (incl. minimized) must NOT
+    // re-resolve and yank the user to another chat: resolve only on a genuine
+    // closed -> open transition. (`windowOpen` is already true here, so there
+    // is nothing to set — just bail.)
+    if (windowOpen) return;
+    // Open the window FIRST so the control feels instant: the bound-chat
+    // round-trip below must never gate the window appearing, or on a slow
+    // connection the first click reads as a hung control until the POST returns.
+    setWindowOpen(true);
+    let resolved: string | null = activeChatId; // off-a-page: keep current
+    if (pageId) {
+      try {
+        resolved = await getBoundChat(pageId); // null => fresh chat
+      } catch {
+        resolved = null; // fail-soft: a fresh chat is always a safe fallback
+      }
+    }
+    // Clear the composer draft / picked role ONLY on an actual switch, so
+    // reopening the same chat does not wipe an in-progress draft. Applied after
+    // the resolve so the window is already visible while the switch settles.
+    if (resolved !== activeChatId) {
+      setActiveChatId(resolved);
+      setDraft("");
+      setSelectedRoleId(null);
+    }
+  }, [
+    windowOpen,
+    activeChatId,
+    pageId,
+    setWindowOpen,
+    setActiveChatId,
+    setDraft,
+    setSelectedRoleId,
+  ]);
+}

apps/client/src/features/ai-chat/queries/ai-chat-query.ts

@@ -13,21 +13,40 @@ import {
   deleteAiRole,
   getAiChatMessages,
   getAiChats,
+  getAiRoleCatalog,
+  getAiRoleCatalogBundle,
   getAiRoles,
+  importAiRolesFromCatalog,
   renameAiChat,
   updateAiRole,
+  updateAiRoleFromCatalog,
 } from "@/features/ai-chat/services/ai-chat-service.ts";
 import {
   IAiChat,
   IAiChatMessageRow,
   IAiRole,
+  IAiRoleCatalog,
+  IAiRoleCatalogBundle,
   IAiRoleCreate,
+  IAiRoleImportPayload,
+  IAiRoleImportResult,
   IAiRoleUpdate,
+  IAiRoleUpdateFromCatalogResult,
 } from "@/features/ai-chat/types/ai-chat.types.ts";
 import { IPagination } from "@/lib/types.ts";
 
 export const AI_CHATS_RQ_KEY = ["ai-chats"];
 export const AI_ROLES_RQ_KEY = ["ai-roles"];
+// Catalog reads resolve bundle names per language, so the language is part of
+// the cache key (a language switch refetches rather than reusing stale names).
+export const AI_ROLE_CATALOG_RQ_KEY = (language: string) => [
+  "ai-role-catalog",
+  language,
+];
+export const AI_ROLE_CATALOG_BUNDLE_RQ_KEY = (
+  bundleId: string,
+  language: string,
+) => ["ai-role-catalog-bundle", bundleId, language];
 export const AI_CHAT_MESSAGES_RQ_KEY = (chatId: string) => [
   "ai-chat-messages",
   chatId,
@@ -223,3 +242,109 @@ export function useDeleteAiRoleMutation() {
     },
   });
 }
+
+/**
+ * Browse the role catalog for a language. Gated by `enabled` so the (admin-only)
+ * fetch runs only when the catalog modal is open. The catalog can 502 when the
+ * curated source is unreachable; callers handle the error state in the UI.
+ */
+export function useAiRoleCatalogQuery(language: string, enabled: boolean) {
+  return useQuery<IAiRoleCatalog, Error>({
+    queryKey: AI_ROLE_CATALOG_RQ_KEY(language),
+    queryFn: () => getAiRoleCatalog(language),
+    enabled,
+  });
+}
+
+/**
+ * Open one catalog bundle (role content + versions). Gated by `enabled` so the
+ * fetch only runs when a bundle is actually expanded.
+ */
+export function useAiRoleCatalogBundleQuery(
+  bundleId: string,
+  language: string,
+  enabled: boolean,
+) {
+  return useQuery<IAiRoleCatalogBundle, Error>({
+    queryKey: AI_ROLE_CATALOG_BUNDLE_RQ_KEY(bundleId, language),
+    queryFn: () => getAiRoleCatalogBundle(bundleId, language),
+    enabled,
+  });
+}
+
+export function useImportAiRolesFromCatalogMutation() {
+  const queryClient = useQueryClient();
+  const { t } = useTranslation();
+
+  return useMutation<IAiRoleImportResult, Error, IAiRoleImportPayload>({
+    mutationFn: (payload) => importAiRolesFromCatalog(payload),
+    onSuccess: (result) => {
+      notifications.show({
+        message: t("Imported {{created}}, renamed {{renamed}}, skipped {{skipped}}", {
+          created: result.created,
+          renamed: result.renamed,
+          skipped: result.skipped,
+        }),
+      });
+      // Surface partial failures (e.g. unique-name races) as a red warning.
+      if (result.errors.length > 0) {
+        notifications.show({
+          color: "red",
+          message: t("Failed to import {{count}} role(s)", {
+            count: result.errors.length,
+          }),
+        });
+      }
+      queryClient.invalidateQueries({ queryKey: AI_ROLES_RQ_KEY });
+      // Imported roles can appear in the chat picker / badges.
+      queryClient.invalidateQueries({ queryKey: AI_CHATS_RQ_KEY });
+    },
+    onError: (error) => {
+      const message = error["response"]?.data?.message;
+      notifications.show({
+        message: message ?? t("Failed to update data"),
+        color: "red",
+      });
+    },
+  });
+}
+
+export function useUpdateAiRoleFromCatalogMutation() {
+  const queryClient = useQueryClient();
+  const { t } = useTranslation();
+
+  return useMutation<IAiRoleUpdateFromCatalogResult, Error, string>({
+    mutationFn: (id) => updateAiRoleFromCatalog(id),
+    onSuccess: (result) => {
+      // The server returns updated:false with a reason for a no-op (already
+      // up to date / removed from catalog / language no longer offered). Map
+      // each reason to a specific message instead of a generic "up to date".
+      // Narrow the discriminated union via `"reason" in result` (the `updated`
+      // boolean discriminant does not narrow under this project's
+      // strictNullChecks:false). Inside the branch, `reason` is the typed literal
+      // union, so the comparisons below are compiler-checked.
+      let message: string;
+      if (!("reason" in result)) {
+        message = t("Updated to the latest version");
+      } else if (result.reason === "not-in-catalog") {
+        message = t("This role is no longer in the catalog");
+      } else if (result.reason === "language-unavailable") {
+        message = t("This language is no longer available in the catalog");
+      } else {
+        // "up-to-date" (the only remaining reason).
+        message = t("Already up to date");
+      }
+      notifications.show({ message });
+      queryClient.invalidateQueries({ queryKey: AI_ROLES_RQ_KEY });
+      // The role badge denormalized onto the chat list may have changed.
+      queryClient.invalidateQueries({ queryKey: AI_CHATS_RQ_KEY });
+    },
+    onError: (error) => {
+      const message = error["response"]?.data?.message;
+      notifications.show({
+        message: message ?? t("Failed to update data"),
+        color: "red",
+      });
+    },
+  });
+}

apps/client/src/features/ai-chat/queries/import-from-catalog-message.test.tsx

@@ -0,0 +1,106 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import React from "react";
+import { renderHook, waitFor } from "@testing-library/react";
+import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
+import type { IAiRoleImportResult } from "@/features/ai-chat/types/ai-chat.types.ts";
+
+// `useImportAiRolesFromCatalogMutation` always shows an Imported/renamed/skipped
+// summary, and ADDITIONALLY a red "Failed to import N role(s)" notification when
+// the result carries partial errors. These tests pin both branches via
+// renderHook with a mocked service (twin precedent:
+// update-from-catalog-message.test.tsx).
+
+const notificationsShowMock = vi.fn();
+vi.mock("@mantine/notifications", () => ({
+  notifications: { show: (opts: unknown) => notificationsShowMock(opts) },
+}));
+
+// `t` echoes the key with interpolated values so we assert against the exact
+// English message strings (mirrors react-i18next's default interpolation).
+vi.mock("react-i18next", () => ({
+  useTranslation: () => ({
+    t: (key: string, vars?: Record<string, unknown>) =>
+      vars
+        ? key.replace(/\{\{(\w+)\}\}/g, (_m, name) => String(vars[name]))
+        : key,
+  }),
+}));
+
+vi.mock("@/features/ai-chat/services/ai-chat-service.ts", () => ({
+  importAiRolesFromCatalog: vi.fn(),
+  // Other named exports referenced by ai-chat-query.ts must exist on the mock so
+  // the module import resolves; they are unused by these tests.
+  createAiRole: vi.fn(),
+  deleteAiChat: vi.fn(),
+  deleteAiRole: vi.fn(),
+  getAiChatMessages: vi.fn(),
+  getAiChats: vi.fn(),
+  getAiRoleCatalog: vi.fn(),
+  getAiRoleCatalogBundle: vi.fn(),
+  getAiRoles: vi.fn(),
+  renameAiChat: vi.fn(),
+  updateAiRole: vi.fn(),
+  updateAiRoleFromCatalog: vi.fn(),
+}));
+
+import { importAiRolesFromCatalog } from "@/features/ai-chat/services/ai-chat-service.ts";
+import { useImportAiRolesFromCatalogMutation } from "@/features/ai-chat/queries/ai-chat-query.ts";
+
+function createWrapper() {
+  const queryClient = new QueryClient({
+    defaultOptions: { queries: { retry: false }, mutations: { retry: false } },
+  });
+  return function Wrapper({ children }: { children: React.ReactNode }) {
+    return (
+      <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+    );
+  };
+}
+
+async function runMutation(result: IAiRoleImportResult) {
+  vi.mocked(importAiRolesFromCatalog).mockResolvedValue(result);
+  const { result: hook } = renderHook(
+    () => useImportAiRolesFromCatalogMutation(),
+    { wrapper: createWrapper() },
+  );
+  hook.current.mutate({
+    bundleId: "general",
+    language: "en",
+    conflict: "rename",
+  });
+  await waitFor(() => expect(hook.current.isSuccess).toBe(true));
+}
+
+describe("useImportAiRolesFromCatalogMutation — success notifications", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+  });
+
+  it("errors:[] -> only the summary notification (counts interpolated)", async () => {
+    await runMutation({ created: 3, renamed: 1, skipped: 2, errors: [] });
+    expect(notificationsShowMock).toHaveBeenCalledTimes(1);
+    expect(notificationsShowMock).toHaveBeenCalledWith({
+      message: "Imported 3, renamed 1, skipped 2",
+    });
+  });
+
+  it("errors.length > 0 -> summary PLUS the red failure notification", async () => {
+    await runMutation({
+      created: 1,
+      renamed: 0,
+      skipped: 0,
+      errors: [
+        { slug: "a", message: "name taken" },
+        { slug: "b", message: "name taken" },
+      ],
+    });
+    expect(notificationsShowMock).toHaveBeenCalledTimes(2);
+    expect(notificationsShowMock).toHaveBeenNthCalledWith(1, {
+      message: "Imported 1, renamed 0, skipped 0",
+    });
+    expect(notificationsShowMock).toHaveBeenNthCalledWith(2, {
+      color: "red",
+      message: "Failed to import 2 role(s)",
+    });
+  });
+});

apps/client/src/features/ai-chat/queries/update-from-catalog-message.test.tsx

@@ -0,0 +1,100 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import React from "react";
+import { renderHook, waitFor } from "@testing-library/react";
+import { QueryClient, QueryClientProvider } from "@tanstack/react-query";
+import type { IAiRoleUpdateFromCatalogResult } from "@/features/ai-chat/types/ai-chat.types.ts";
+
+// `useUpdateAiRoleFromCatalogMutation` maps the server's discriminated result to
+// a user-facing notification message. These tests pin each of the four branches
+// (updated / not-in-catalog / language-unavailable / up-to-date) via renderHook
+// with a mocked service (precedent: share-query.null-normalization.test.tsx).
+
+const notificationsShowMock = vi.fn();
+vi.mock("@mantine/notifications", () => ({
+  notifications: { show: (opts: unknown) => notificationsShowMock(opts) },
+}));
+
+// `t` echoes the key so we assert against the exact English message strings.
+vi.mock("react-i18next", () => ({
+  useTranslation: () => ({ t: (key: string) => key }),
+}));
+
+vi.mock("@/features/ai-chat/services/ai-chat-service.ts", () => ({
+  updateAiRoleFromCatalog: vi.fn(),
+  // Other named exports referenced by ai-chat-query.ts must exist on the mock so
+  // the module import resolves; they are unused by these tests.
+  createAiRole: vi.fn(),
+  deleteAiChat: vi.fn(),
+  deleteAiRole: vi.fn(),
+  getAiChatMessages: vi.fn(),
+  getAiChats: vi.fn(),
+  getAiRoleCatalog: vi.fn(),
+  getAiRoleCatalogBundle: vi.fn(),
+  getAiRoles: vi.fn(),
+  importAiRolesFromCatalog: vi.fn(),
+  renameAiChat: vi.fn(),
+  updateAiRole: vi.fn(),
+}));
+
+import { updateAiRoleFromCatalog } from "@/features/ai-chat/services/ai-chat-service.ts";
+import { useUpdateAiRoleFromCatalogMutation } from "@/features/ai-chat/queries/ai-chat-query.ts";
+
+function createWrapper() {
+  const queryClient = new QueryClient({
+    defaultOptions: { queries: { retry: false }, mutations: { retry: false } },
+  });
+  return function Wrapper({ children }: { children: React.ReactNode }) {
+    return (
+      <QueryClientProvider client={queryClient}>{children}</QueryClientProvider>
+    );
+  };
+}
+
+async function runMutation(result: IAiRoleUpdateFromCatalogResult) {
+  vi.mocked(updateAiRoleFromCatalog).mockResolvedValue(result);
+  const { result: hook } = renderHook(
+    () => useUpdateAiRoleFromCatalogMutation(),
+    { wrapper: createWrapper() },
+  );
+  hook.current.mutate("role-1");
+  await waitFor(() => expect(hook.current.isSuccess).toBe(true));
+}
+
+describe("useUpdateAiRoleFromCatalogMutation — reason → message", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+  });
+
+  it("updated:true -> 'Updated to the latest version'", async () => {
+    await runMutation({
+      updated: true,
+      fromVersion: 1,
+      toVersion: 2,
+      role: { id: "role-1" } as never,
+    });
+    expect(notificationsShowMock).toHaveBeenCalledWith({
+      message: "Updated to the latest version",
+    });
+  });
+
+  it("not-in-catalog -> 'This role is no longer in the catalog'", async () => {
+    await runMutation({ updated: false, reason: "not-in-catalog" });
+    expect(notificationsShowMock).toHaveBeenCalledWith({
+      message: "This role is no longer in the catalog",
+    });
+  });
+
+  it("language-unavailable -> 'This language is no longer available in the catalog'", async () => {
+    await runMutation({ updated: false, reason: "language-unavailable" });
+    expect(notificationsShowMock).toHaveBeenCalledWith({
+      message: "This language is no longer available in the catalog",
+    });
+  });
+
+  it("up-to-date -> 'Already up to date'", async () => {
+    await runMutation({ updated: false, reason: "up-to-date" });
+    expect(notificationsShowMock).toHaveBeenCalledWith({
+      message: "Already up to date",
+    });
+  });
+});

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

@@ -6,8 +6,13 @@ import {
   IAiChatMessageRow,
   IAiChatMessagesParams,
   IAiRole,
+  IAiRoleCatalog,
+  IAiRoleCatalogBundle,
   IAiRoleCreate,
+  IAiRoleImportPayload,
+  IAiRoleImportResult,
   IAiRoleUpdate,
+  IAiRoleUpdateFromCatalogResult,
 } from "@/features/ai-chat/types/ai-chat.types.ts";
 
 /**
@@ -37,6 +42,17 @@ export async function getAiChatMessages(
   return req.data;
 }
 
+/**
+ * Resolve the chat bound to a document (the current user's most-recent chat
+ * created on that page), or null when there is none. Drives auto-open-on-page.
+ */
+export async function getBoundChat(pageId: string): Promise<string | null> {
+  const req = await api.post<{ chatId: string | null }>("/ai-chat/bound-chat", {
+    pageId,
+  });
+  return req.data.chatId;
+}
+
 /** Rename a chat. */
 export async function renameAiChat(data: {
   chatId: string;
@@ -112,3 +128,54 @@ export async function deleteAiRole(id: string): Promise<{ success: true }> {
   });
   return req.data;
 }
+
+/**
+ * Role catalog API (`/ai-chat/roles/*`, admin-only — the server enforces this).
+ * Browse a curated catalog, import roles/bundles into the workspace, and update
+ * an imported role when the catalog ships a newer version. Same `{ data }`
+ * unwrap convention as above.
+ */
+
+/** Browse the catalog, optionally localized to `language`. */
+export async function getAiRoleCatalog(
+  language?: string,
+): Promise<IAiRoleCatalog> {
+  const req = await api.post<IAiRoleCatalog>("/ai-chat/roles/catalog", {
+    language,
+  });
+  return req.data;
+}
+
+/** Open one catalog bundle in a language (role content + versions). */
+export async function getAiRoleCatalogBundle(
+  bundleId: string,
+  language: string,
+): Promise<IAiRoleCatalogBundle> {
+  const req = await api.post<IAiRoleCatalogBundle>(
+    "/ai-chat/roles/catalog/bundle",
+    { bundleId, language },
+  );
+  return req.data;
+}
+
+/** Import roles from a catalog bundle into the workspace (admin). */
+export async function importAiRolesFromCatalog(
+  payload: IAiRoleImportPayload,
+): Promise<IAiRoleImportResult> {
+  const req = await api.post<IAiRoleImportResult>(
+    "/ai-chat/roles/import",
+    payload,
+  );
+  return req.data;
+}
+
+/** Update an already-imported role from its catalog source (admin). */
+export async function updateAiRoleFromCatalog(
+  id: string,
+): Promise<IAiRoleUpdateFromCatalogResult> {
+  const req = await api.post<IAiRoleUpdateFromCatalogResult>(
+    "/ai-chat/roles/update-from-catalog",
+    { id },
+  );
+  return req.data;
+}

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

@@ -57,10 +57,79 @@ export interface IAiRole {
   autoStart: boolean;
   // Custom auto-start text; null/empty => the default launch message is sent.
   launchMessage: string | null;
+  // Catalog origin of an imported role, or null for a manually-created one.
+  // Admin-only (present only in the admin list view); the picker view omits it.
+  // The admin UI compares `version` against the catalog to offer an update.
+  source?: { slug: string; language: string; version: number } | null;
   createdAt?: string;
   updatedAt?: string;
 }
 
+/** One bundle's summary in the catalog index (mirrors `getCatalog().bundles[]`). */
+export interface IAiRoleCatalogBundleSummary {
+  id: string;
+  name: string;
+  description: string | null;
+  languages: string[];
+  roles: { slug: string; version: number }[];
+}
+
+/** The browsable catalog index (mirrors `getCatalog()`). */
+export interface IAiRoleCatalog {
+  languages: string[];
+  bundles: IAiRoleCatalogBundleSummary[];
+}
+
+/** A single role inside an opened catalog bundle (localized content + version). */
+export interface IAiRoleCatalogRole {
+  slug: string;
+  emoji: string | null;
+  name: string;
+  description: string | null;
+  instructions: string;
+  autoStart: boolean;
+  launchMessage: string | null;
+  version: number;
+}
+
+/** An opened catalog bundle (mirrors `getCatalogBundle()`). */
+export interface IAiRoleCatalogBundle {
+  bundleId: string;
+  language: string;
+  roles: IAiRoleCatalogRole[];
+}
+
+/** Import payload (mirrors the server `ImportFromCatalogDto`). */
+export interface IAiRoleImportPayload {
+  bundleId: string;
+  language: string;
+  // Omitted => import the whole bundle; otherwise only these slugs.
+  slugs?: string[];
+  conflict: "skip" | "rename";
+}
+
+/** Import result counts (mirrors `importFromCatalog()`). */
+export interface IAiRoleImportResult {
+  created: number;
+  skipped: number;
+  renamed: number;
+  errors: { slug: string; message: string }[];
+}
+
+/**
+ * Update-from-catalog result (mirrors the server `updateFromCatalog()`). A
+ * discriminated union on `updated`: a no-op carries a typed `reason` the UI maps
+ * to a specific message; a successful update carries the version bump + new role.
+ * Keeping the union (not a widened `reason?: string`) lets the consumer's literal
+ * comparisons be compiler-checked.
+ */
+export type IAiRoleUpdateFromCatalogResult =
+  | {
+      updated: false;
+      reason: "not-in-catalog" | "up-to-date" | "language-unavailable";
+    }
+  | { updated: true; fromVersion: number; toVersion: number; role: IAiRole };
+
 /** Admin create payload for a role. */
 export interface IAiRoleCreate {
   name: string;

apps/client/src/features/ai-chat/utils/catalog-role-install-state.test.ts

@@ -0,0 +1,107 @@
+import { describe, it, expect } from "vitest";
+import { catalogRoleInstallState } from "./catalog-role-install-state.ts";
+import type { IAiRole } from "@/features/ai-chat/types/ai-chat.types.ts";
+
+// Build a workspace role with a catalog source. Fields irrelevant to the
+// install-state decision are filled with harmless defaults.
+function installedRole(
+  source: { slug: string; language: string; version: number },
+  overrides: Partial<IAiRole> = {},
+): IAiRole {
+  return {
+    id: `role-${source.slug}-${source.language}`,
+    name: source.slug,
+    emoji: null,
+    description: null,
+    enabled: true,
+    autoStart: true,
+    launchMessage: null,
+    source,
+    ...overrides,
+  };
+}
+
+const catalogRole = { slug: "writer", version: 3 };
+
+// Mirrors the role-launch.ts precedent: the modal's role-state computation is a
+// pure function so the import/installed/update decision is testable directly.
+describe("catalogRoleInstallState", () => {
+  it("no matching installed role -> import", () => {
+    const result = catalogRoleInstallState(catalogRole, [], "en");
+    expect(result).toEqual({ state: "import" });
+  });
+
+  it("same slug + language, installed version > catalog -> installed", () => {
+    const installed = installedRole({
+      slug: "writer",
+      language: "en",
+      version: 5,
+    });
+    const result = catalogRoleInstallState(catalogRole, [installed], "en");
+    expect(result).toEqual({ state: "installed", installed });
+  });
+
+  it("same slug + language, installed version == catalog -> installed", () => {
+    const installed = installedRole({
+      slug: "writer",
+      language: "en",
+      version: 3,
+    });
+    const result = catalogRoleInstallState(catalogRole, [installed], "en");
+    expect(result).toEqual({ state: "installed", installed });
+  });
+
+  it("same slug + language, installed version < catalog -> update (from/to)", () => {
+    const installed = installedRole({
+      slug: "writer",
+      language: "en",
+      version: 1,
+    });
+    const result = catalogRoleInstallState(catalogRole, [installed], "en");
+    expect(result).toEqual({
+      state: "update",
+      installed,
+      fromVersion: 1,
+      toVersion: 3,
+    });
+  });
+
+  it("same slug but DIFFERENT language -> import (a separate install)", () => {
+    // 'writer' is installed in 'ru'; browsing the 'en' catalog must offer it as a
+    // fresh import, not treat the ru copy as already installed.
+    const installed = installedRole({
+      slug: "writer",
+      language: "ru",
+      version: 5,
+    });
+    const result = catalogRoleInstallState(catalogRole, [installed], "en");
+    expect(result).toEqual({ state: "import" });
+  });
+
+  it("matches the right language when the same slug is installed in several", () => {
+    const ru = installedRole(
+      { slug: "writer", language: "ru", version: 5 },
+      { id: "ru-role" },
+    );
+    const en = installedRole(
+      { slug: "writer", language: "en", version: 1 },
+      { id: "en-role" },
+    );
+    const result = catalogRoleInstallState(catalogRole, [ru, en], "en");
+    expect(result).toEqual({
+      state: "update",
+      installed: en,
+      fromVersion: 1,
+      toVersion: 3,
+    });
+  });
+
+  it("ignores manually-created roles (no source) sharing the name", () => {
+    const manual = installedRole(
+      { slug: "writer", language: "en", version: 9 },
+      { source: null },
+    );
+    const result = catalogRoleInstallState(catalogRole, [manual], "en");
+    expect(result).toEqual({ state: "import" });
+  });
+});

apps/client/src/features/ai-chat/utils/catalog-role-install-state.ts

@@ -0,0 +1,49 @@
+import type {
+  IAiRole,
+  IAiRoleCatalogRole,
+} from "@/features/ai-chat/types/ai-chat.types.ts";
+
+/**
+ * The install state of a single catalog role relative to the workspace's
+ * existing roles. Extracted as a pure function so the catalog modal's role-state
+ * computation is unit-testable without mounting the component (mirrors the
+ * `roleLaunchMessage` precedent in role-launch.ts).
+ *
+ * A catalog role is matched to an installed role by BOTH `source.slug` and
+ * `source.language`: the same slug in a different language is a separate install
+ * (so it shows as "import", not "installed"). When matched, the installed source
+ * version decides the state:
+ *   - no match                          -> "import"
+ *   - matched & installed version >= catalog version -> "installed"
+ *   - matched & installed version <  catalog version -> "update" (from -> to)
+ */
+export type CatalogRoleInstallState =
+  | { state: "import" }
+  | { state: "installed"; installed: IAiRole }
+  | {
+      state: "update";
+      installed: IAiRole;
+      fromVersion: number;
+      toVersion: number;
+    };
+
+export function catalogRoleInstallState(
+  role: Pick<IAiRoleCatalogRole, "slug" | "version">,
+  workspaceRoles: IAiRole[],
+  language: string,
+): CatalogRoleInstallState {
+  const installed = workspaceRoles.find(
+    (r) => r.source?.slug === role.slug && r.source?.language === language,
+  );
+  if (!installed) return { state: "import" };
+  const fromVersion = installed.source?.version ?? 0;
+  if (fromVersion >= role.version) {
+    return { state: "installed", installed };
+  }
+  return {
+    state: "update",
+    installed,
+    fromVersion,
+    toVersion: role.version,
+  };
+}

apps/client/src/features/editor/extensions/markdown-clipboard.canonicalize.test.ts

@@ -0,0 +1,168 @@
+import { describe, it, expect } from "vitest";
+import { Editor } from "@tiptap/core";
+import { Document } from "@tiptap/extension-document";
+import { Paragraph } from "@tiptap/extension-paragraph";
+import { Text } from "@tiptap/extension-text";
+import { Node as PMNode, Fragment, Slice } from "@tiptap/pm/model";
+import {
+  FootnoteReference,
+  FootnotesList,
+  FootnoteDefinition,
+  FOOTNOTE_REFERENCE_NAME,
+  FOOTNOTE_DEFINITION_NAME,
+  FOOTNOTES_LIST_NAME,
+} from "@docmost/editor-ext";
+import { canonicalizePastedFootnotes } from "./markdown-clipboard";
+
+/**
+ * A markdown paste builds its ProseMirror fragment via DOM -> parseSlice and is
+ * applied with a manual transaction (handlePaste returns true), so it bypasses
+ * the editor's footnoteSyncPlugin — which never reorders an existing list. These
+ * tests pin canonicalizePastedFootnotes, the focused hook that makes a pasted
+ * out-of-order markdown footnote block come out canonical (issue #228).
+ */
+
+const extensions = [
+  Document,
+  Paragraph,
+  Text,
+  FootnoteReference,
+  FootnotesList,
+  FootnoteDefinition,
+];
+
+function makeSchema() {
+  const editor = new Editor({ extensions, content: { type: "doc", content: [] } });
+  const { schema } = editor;
+  return { editor, schema };
+}
+
+/** List footnote def ids of the (single) footnotesList in a slice, in order. */
+function listIds(slice: Slice): string[] {
+  const out: string[] = [];
+  slice.content.forEach((node: PMNode) => {
+    if (node.type.name === FOOTNOTES_LIST_NAME) {
+      node.content.forEach((def: PMNode) => {
+        if (def.type.name === FOOTNOTE_DEFINITION_NAME) out.push(def.attrs.id);
+      });
+    }
+  });
+  return out;
+}
+
+function hasList(slice: Slice): boolean {
+  let found = false;
+  slice.content.forEach((n: PMNode) => {
+    if (n.type.name === FOOTNOTES_LIST_NAME) found = true;
+  });
+  return found;
+}
+
+describe("canonicalizePastedFootnotes", () => {
+  it("reorders a pasted block to reference order, dedups reuse, drops orphans", () => {
+    const { editor, schema } = makeSchema();
+    // Body references c, a, b (and again a => reuse); definitions a, b, c, z
+    // (z is an orphan) — the exact shape a markdown paste produces.
+    const slice = new Slice(
+      Fragment.fromArray([
+        schema.nodes.paragraph.create(null, [
+          schema.text("body "),
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "c" }),
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "a" }),
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "b" }),
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "a" }),
+        ]),
+        schema.nodes[FOOTNOTES_LIST_NAME].create(null, [
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "a" }, [
+            schema.nodes.paragraph.create(null, [schema.text("note A")]),
+          ]),
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "b" }, [
+            schema.nodes.paragraph.create(null, [schema.text("note B")]),
+          ]),
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "c" }, [
+            schema.nodes.paragraph.create(null, [schema.text("note C")]),
+          ]),
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "z" }, [
+            schema.nodes.paragraph.create(null, [schema.text("orphan")]),
+          ]),
+        ]),
+      ]),
+      0,
+      0,
+    );
+
+    const out = canonicalizePastedFootnotes(slice, schema);
+    // Reference order, orphan z dropped, reused a appears once.
+    expect(listIds(out)).toEqual(["c", "a", "b"]);
+    editor.destroy();
+  });
+
+  it("leaves a reference-ONLY paste untouched (no synthesized definitions)", () => {
+    // A paste that reuses an id defined in the TARGET doc must NOT gain a
+    // synthesized empty definition here — it carries no footnotesList of its own.
+    const { editor, schema } = makeSchema();
+    const slice = new Slice(
+      Fragment.from(
+        schema.nodes.paragraph.create(null, [
+          schema.text("see "),
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "a" }),
+        ]),
+      ),
+      0,
+      0,
+    );
+    const out = canonicalizePastedFootnotes(slice, schema);
+    expect(hasList(out)).toBe(false);
+    expect(out).toBe(slice); // returned unchanged (same reference)
+    editor.destroy();
+  });
+
+  it("leaves a definitions-ONLY paste untouched (no references -> no empty paste)", () => {
+    // A whole-block paste of ONLY definitions (a footnotesList with no matching
+    // footnoteReference anywhere in the selection). Canonicalizing it would strip
+    // the reference-less list -> an EMPTY paste, losing the pasted text. The hook
+    // must leave such a block untouched.
+    const { editor, schema } = makeSchema();
+    const slice = new Slice(
+      Fragment.fromArray([
+        schema.nodes[FOOTNOTES_LIST_NAME].create(null, [
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "a" }, [
+            schema.nodes.paragraph.create(null, [schema.text("note A")]),
+          ]),
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "b" }, [
+            schema.nodes.paragraph.create(null, [schema.text("note B")]),
+          ]),
+        ]),
+      ]),
+      0,
+      0,
+    );
+    const out = canonicalizePastedFootnotes(slice, schema);
+    expect(out).toBe(slice); // returned unchanged (same reference, content kept)
+    expect(listIds(out)).toEqual(["a", "b"]);
+    editor.destroy();
+  });
+
+  it("leaves an open (partial) slice untouched even if it carries a list", () => {
+    // An open slice (openStart/openEnd > 0) is a partial selection, not a
+    // standalone block, so it is returned as-is BEFORE any footnote handling.
+    const { editor, schema } = makeSchema();
+    const slice = new Slice(
+      Fragment.fromArray([
+        schema.nodes.paragraph.create(null, [
+          schema.nodes[FOOTNOTE_REFERENCE_NAME].create({ id: "a" }),
+        ]),
+        schema.nodes[FOOTNOTES_LIST_NAME].create(null, [
+          schema.nodes[FOOTNOTE_DEFINITION_NAME].create({ id: "a" }, [
+            schema.nodes.paragraph.create(null, [schema.text("A")]),
+          ]),
+        ]),
+      ]),
+      1,
+      1,
+    );
+    const out = canonicalizePastedFootnotes(slice, schema);
+    expect(out).toBe(slice);
+    editor.destroy();
+  });
+});

apps/client/src/features/editor/extensions/markdown-clipboard.ts

@@ -3,7 +3,14 @@ import { Extension } from "@tiptap/core";
 import { Plugin, PluginKey, TextSelection } from "@tiptap/pm/state";
 import { DOMParser, DOMSerializer, Fragment, Slice } from "@tiptap/pm/model";
 import { find } from "linkifyjs";
-import { markdownToHtml, htmlToMarkdown } from "@docmost/editor-ext";
+import {
+  markdownToHtml,
+  htmlToMarkdown,
+  canonicalizeFootnotes,
+  FOOTNOTES_LIST_NAME,
+  FOOTNOTE_REFERENCE_NAME,
+} from "@docmost/editor-ext";
+import type { Schema } from "@tiptap/pm/model";
 
 export const MarkdownClipboard = Extension.create({
   name: "markdownClipboard",
@@ -83,12 +90,25 @@ export const MarkdownClipboard = Extension.create({
             const body = elementFromString(parsed);
             normalizeTableColumnWidths(body);
 
-            const contentNodes = DOMParser.fromSchema(
+            const parsedSlice = DOMParser.fromSchema(
               this.editor.schema,
             ).parseSlice(body, {
               preserveWhitespace: true,
             });
 
+            // A markdown paste builds its ProseMirror fragment directly (DOM ->
+            // parseSlice), bypassing the editor's footnoteSyncPlugin, which never
+            // reorders an existing list. So a pasted markdown block whose footnote
+            // definitions are out of order (or contains orphan defs) would be
+            // stored out of order. Canonicalize the self-contained pasted block so
+            // its footnotes come out reference-ordered, deduped and orphan-free
+            // (issue #228). See canonicalizePastedFootnotes for why this is scoped
+            // to whole-block pastes that carry their own footnotesList.
+            const contentNodes = canonicalizePastedFootnotes(
+              parsedSlice,
+              this.editor.schema,
+            );
+
             tr.replaceRange(from, to, contentNodes);
             const insertEnd = tr.mapping.map(from, 1);
             tr.setSelection(TextSelection.near(tr.doc.resolve(Math.max(from, insertEnd - 2)), -1));
@@ -133,6 +153,54 @@ export const MarkdownClipboard = Extension.create({
   },
 });
 
+/**
+ * Reorder/dedup the footnotes of a SELF-CONTAINED pasted markdown block to the
+ * canonical invariant (the live footnoteSyncPlugin never reorders an existing
+ * list, so an out-of-order pasted block would otherwise persist out of order).
+ *
+ * Scoped deliberately to whole-block pastes (openStart/openEnd === 0) that carry
+ * their OWN footnotesList: canonicalizeFootnotes would synthesize empty
+ * definitions for any reference lacking a definition, which is correct for a
+ * standalone block but would be wrong for a reference-only paste that REUSES a
+ * footnote already defined in the target document — so those are left untouched
+ * for the paste/sync plugins to merge. Residual: when the pasted block is merged
+ * into a doc that already has footnotes, ordering RELATIVE to the pre-existing
+ * footnotes is still governed by the sync plugin (which does not reorder).
+ *
+ * Also requires at least one footnoteReference in the selection: a definitions-ONLY
+ * paste (`[^a]: …` with no `[^a]` reference in the same block) has no references,
+ * so canonicalizeFootnotes would drop the whole list and the paste would come out
+ * EMPTY — losing the pasted text. Such a block is left as-is for the sync plugin.
+ */
+export function canonicalizePastedFootnotes(slice: Slice, schema: Schema): Slice {
+  if (slice.openStart !== 0 || slice.openEnd !== 0) return slice;
+
+  let hasFootnotesList = false;
+  let hasReference = false;
+  slice.content.forEach((node) => {
+    if (node.type.name === FOOTNOTES_LIST_NAME) hasFootnotesList = true;
+    // footnoteReference is an inline atom, never a top-level slice child here
+    // (this function early-returns for open slices, so children are whole
+    // blocks), so it is only reachable by descending.
+    node.descendants((child) => {
+      if (child.type.name === FOOTNOTE_REFERENCE_NAME) hasReference = true;
+    });
+  });
+  if (!hasFootnotesList) return slice;
+  // No reference anywhere -> a definitions-only paste; canonicalizing would strip
+  // the reference-less list (empty paste). Leave it untouched.
+  if (!hasReference) return slice;
+
+  const content = slice.content.toJSON();
+  if (!Array.isArray(content)) return slice;
+
+  const canonical = canonicalizeFootnotes({ type: "doc", content }) as {
+    content?: unknown[];
+  };
+  const fragment = Fragment.fromJSON(schema, canonical.content ?? []);
+  return new Slice(fragment, 0, 0);
+}
+
 function elementFromString(value) {
   // add a wrapper to preserve leading and trailing whitespace
   const wrappedValue = `<body>${value}</body>`;

apps/client/src/features/home/components/new-note-button.tsx

@@ -1,4 +1,4 @@
-import { Button, Group, Menu, Text } from "@mantine/core";
+import { Button, Menu, Stack, Text } from "@mantine/core";
 import { IconHourglass, IconPlus } from "@tabler/icons-react";
 import { ReactNode } from "react";
 import { useNavigate } from "react-router-dom";
@@ -21,11 +21,15 @@ function CreateNoteButton({
   temporary,
   label,
   icon,
+  color,
 }: {
   writableSpaces: ISpace[];
   temporary: boolean;
   label: string;
   icon: ReactNode;
+  // Mantine color token; lets the temporary action tint toward the warm
+  // orange/amber used by the clock marker + banner while "New note" stays neutral.
+  color: string;
 }) {
   const { t } = useTranslation();
   const navigate = useNavigate();
@@ -54,7 +58,8 @@ function CreateNoteButton({
       <Button
         size="md"
         variant="light"
-        color="gray"
+        color={color}
+        fullWidth
         leftSection={icon}
         loading={isPending}
         onClick={() => createNote(writableSpaces[0])}
@@ -71,7 +76,8 @@ function CreateNoteButton({
         <Button
           size="md"
           variant="light"
-          color="gray"
+          color={color}
+          fullWidth
           leftSection={icon}
           loading={isPending}
         >
@@ -109,8 +115,10 @@ function CreateNoteButton({
 // Prominent home-screen actions to create a new note (page). Because the home
 // screen has no active space, the target space is resolved from the user's
 // writable spaces: created directly when there is one, picked from a dropdown
-// when there are several. Renders two equal-width buttons: a regular note and a
-// temporary note (which auto-moves to Trash after the workspace lifetime).
+// when there are several. Renders two full-width, vertically stacked buttons: a
+// neutral regular note and an orange-tinted temporary note (which auto-moves to
+// Trash after the workspace lifetime). Stacking full-width keeps the longer
+// "New temporary note" label from clipping on narrow mobile widths.
 export default function NewNoteButton() {
   const { t } = useTranslation();
   const { data } = useGetSpacesQuery({ limit: 100 });
@@ -121,19 +129,21 @@ export default function NewNoteButton() {
   if (writableSpaces.length === 0) return null;
 
   return (
-    <Group grow gap="sm">
+    <Stack gap="sm">
       <CreateNoteButton
         writableSpaces={writableSpaces}
         temporary={false}
         label={t("New note")}
         icon={<IconPlus size={18} />}
+        color="gray"
       />
       <CreateNoteButton
         writableSpaces={writableSpaces}
         temporary={true}
         label={t("New temporary note")}
         icon={<IconHourglass size={18} />}
+        color="orange"
       />
-    </Group>
+    </Stack>
   );
 }

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

@@ -0,0 +1,69 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import { getDefaultStore } from "jotai";
+
+// Mock the app entry so importing the query module doesn't boot the whole app
+// (it only needs queryClient's cache methods, which we stub here). The spies are
+// declared via vi.hoisted so they exist before the hoisted vi.mock factory runs.
+const { setQueryData, getQueryData, invalidateQueries } = vi.hoisted(() => ({
+  setQueryData: vi.fn(),
+  getQueryData: vi.fn(() => undefined as unknown),
+  invalidateQueries: vi.fn(),
+}));
+vi.mock("@/main.tsx", () => ({
+  queryClient: { setQueryData, getQueryData, invalidateQueries },
+}));
+
+import { syncTemporaryExpiresInCache } from "./page-embed-query";
+import { treeDataAtom } from "@/features/page/tree/atoms/tree-data-atom.ts";
+import { SpaceTreeNode } from "@/features/page/tree/types.ts";
+
+const mkNode = (id: string, slugId: string): SpaceTreeNode =>
+  ({
+    id,
+    slugId,
+    name: id,
+    position: "a0",
+    spaceId: "space-1",
+    parentPageId: null,
+    hasChildren: false,
+    children: [],
+  }) as unknown as SpaceTreeNode;
+
+describe("syncTemporaryExpiresInCache — treeDataAtom patch", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+    getQueryData.mockReturnValue(undefined);
+  });
+
+  it("patches the in-tree node's temporaryExpiresAt (sidebar marker updates without reload)", () => {
+    const store = getDefaultStore();
+    const tree = [mkNode("p1", "slug-1"), mkNode("p2", "slug-2")];
+    store.set(treeDataAtom, tree);
+
+    const deadline = "2026-07-01T00:00:00.000Z";
+    syncTemporaryExpiresInCache({ id: "p1", slugId: "slug-1" }, deadline);
+
+    const next = store.get(treeDataAtom);
+    // A new atom value was written...
+    expect(next).not.toBe(tree);
+    // ...the matching node gained the deadline...
+    expect(next.find((n) => n.id === "p1")?.temporaryExpiresAt).toBe(deadline);
+    // ...and the untouched sibling is unchanged.
+    expect(next.find((n) => n.id === "p2")?.temporaryExpiresAt).toBeUndefined();
+  });
+
+  it("leaves the atom value at the SAME reference when the id is absent from the tree (no write)", () => {
+    const store = getDefaultStore();
+    const tree = [mkNode("p1", "slug-1")];
+    store.set(treeDataAtom, tree);
+
+    syncTemporaryExpiresInCache(
+      { id: "not-in-tree", slugId: "missing" },
+      "2026-07-01T00:00:00.000Z",
+    );
+
+    // treeModel.update is a no-op (same reference) for an unknown id, so the
+    // guard skips the store write entirely — same reference back.
+    expect(store.get(treeDataAtom)).toBe(tree);
+  });
+});

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

@@ -1,5 +1,6 @@
 import { useMutation } from "@tanstack/react-query";
 import { notifications } from "@mantine/notifications";
+import { getDefaultStore } from "jotai";
 import {
   toggleTemplate,
   toggleTemporary,
@@ -9,6 +10,9 @@ import type {
   ToggleTemporaryResponse,
 } from "@/features/page-embed/types/page-embed.types";
 import { queryClient } from "@/main.tsx";
+import { treeDataAtom } from "@/features/page/tree/atoms/tree-data-atom.ts";
+import { treeModel } from "@/features/page/tree/model/tree-model";
+import { SpaceTreeNode } from "@/features/page/tree/types.ts";
 
 /**
  * After toggling a note's temporary state, mirror the new deadline into the
@@ -30,6 +34,19 @@ export function syncTemporaryExpiresInCache(
       });
     }
   }
+  // Patch the in-memory sidebar tree node so its temporary clock marker
+  // appears/disappears immediately — WITHOUT a reload. The page cache update
+  // above only drives the in-page banner/menu; the sidebar reads
+  // `temporaryExpiresAt` straight off the `treeDataAtom` node. The app uses
+  // jotai's default store (no <Provider>), so `getDefaultStore()` is the same
+  // store the sidebar's hooks read from. `treeModel.update` returns the same
+  // reference (a no-op) when the page isn't in the currently loaded tree.
+  const store = getDefaultStore();
+  const prevTree = store.get(treeDataAtom);
+  const nextTree = treeModel.update(prevTree, page.id, {
+    temporaryExpiresAt,
+  } as Partial<SpaceTreeNode>);
+  if (nextTree !== prevTree) store.set(treeDataAtom, nextTree);
   queryClient.invalidateQueries({
     predicate: (item) =>
       ["sidebar-pages"].includes(item.queryKey[0] as string),

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

@@ -176,8 +176,8 @@ function PageActionMenu({ readOnly }: PageActionMenuProps) {
         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.
+      // Reflect the new deadline in the page cache (menu label + banner) AND in
+      // the sidebar tree node so its clock marker updates immediately, no reload.
       syncTemporaryExpiresInCache(page, res.temporaryExpiresAt);
       notifications.show({
         message: next

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

@@ -32,7 +32,7 @@ import {
 import { notifications } from "@mantine/notifications";
 import { IPagination, QueryParams } from "@/lib/types.ts";
 import { queryClient } from "@/main.tsx";
-import { buildTree } from "@/features/page/tree/utils";
+import { buildTree, pageToTreeNode } from "@/features/page/tree/utils";
 import { useEffect } from "react";
 import { validate as isValidUuid } from "uuid";
 import { useTranslation } from "react-i18next";
@@ -210,18 +210,15 @@ export function useRestorePageMutation() {
 
       // Check if the page already exists in the tree (it shouldn't)
       if (!treeModel.find(currentTree, restoredPage.id)) {
-        // Create the tree node data with hasChildren from backend
-        const nodeData: SpaceTreeNode = {
-          id: restoredPage.id,
-          slugId: restoredPage.slugId,
+        // Create the tree node data with hasChildren from backend. Routed
+        // through the canonical mapper so the field copy stays in lockstep with
+        // buildTree. The server NULLS `temporaryExpiresAt` on restore (a restored
+        // page is made permanent), so the mapper carries that null through and
+        // the node correctly shows no clock marker.
+        const nodeData: SpaceTreeNode = pageToTreeNode(restoredPage, {
           name: restoredPage.title || "Untitled",
-          icon: restoredPage.icon,
-          position: restoredPage.position,
-          spaceId: restoredPage.spaceId,
-          parentPageId: restoredPage.parentPageId,
           hasChildren: restoredPage.hasChildren || false,
-          children: [],
-        };
+        });
 
         // Determine the parent and index
         const parentId = restoredPage.parentPageId || null;
@@ -410,6 +407,11 @@ export function invalidateOnCreatePage(data: Partial<IPage>) {
     slugId: data.slugId,
     spaceId: data.spaceId,
     title: data.title,
+    // Carry the death-timer deadline so a note created as temporary keeps its
+    // sidebar clock marker when the tree is rebuilt from this cached entry
+    // (buildTree → mergeRootTrees). Omitting it overwrote the optimistic/socket
+    // node's marker with `undefined`, hiding it until a reload.
+    temporaryExpiresAt: data.temporaryExpiresAt,
   };
 
   let queryKey: QueryKey = null;

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

@@ -37,6 +37,7 @@ import {
 } 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 { pageToTreeNode } from "@/features/page/tree/utils";
 import { useTreeMutation } from "@/features/page/tree/hooks/use-tree-mutation.ts";
 import type { SpaceTreeNode } from "@/features/page/tree/types.ts";
 import classes from "@/features/page/tree/styles/tree.module.css";
@@ -130,18 +131,14 @@ export function NodeMenu({ node, canEdit }: NodeMenuProps) {
       const currentIndex = siblings?.index ?? 0;
       const newIndex = currentIndex + 1;
 
-      const treeNodeData: SpaceTreeNode = {
-        id: duplicatedPage.id,
-        slugId: duplicatedPage.slugId,
-        name: duplicatedPage.title,
-        position: duplicatedPage.position,
-        spaceId: duplicatedPage.spaceId,
-        parentPageId: duplicatedPage.parentPageId,
-        icon: duplicatedPage.icon,
-        hasChildren: duplicatedPage.hasChildren,
+      // Routed through the canonical mapper so the field copy stays in lockstep
+      // with buildTree. The server does NOT arm a death timer on duplicate (the
+      // copy's `temporaryExpiresAt` defaults to null = permanent), so the mapper
+      // carries that null through and the duplicated node correctly shows no
+      // clock marker — matching the server without a reload.
+      const treeNodeData: SpaceTreeNode = pageToTreeNode(duplicatedPage, {
         canEdit: true,
-        children: [],
-      };
+      });
 
       setData((prev) =>
         treeModel.insert(prev, parentId, treeNodeData, newIndex),

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

@@ -9,6 +9,7 @@ import { treeModel } from "@/features/page/tree/model/tree-model";
 import type { DropOp } from "@/features/page/tree/model/tree-model.types";
 import { dropOpToMovePayload } from "./drop-op-to-move-payload";
 import { SpaceTreeNode } from "@/features/page/tree/types.ts";
+import { pageToTreeNode } from "@/features/page/tree/utils";
 import { IPage } from "@/features/page/types/page.types.ts";
 import {
   useCreatePageMutation,
@@ -139,18 +140,15 @@ export function useTreeMutation(spaceId: string): UseTreeMutation {
         throw new Error("Failed to create page");
       }
 
-      const newNode: SpaceTreeNode = {
-        id: createdPage.id,
-        slugId: createdPage.slugId,
+      // Route through the canonical mapper so the field copy (esp.
+      // `temporaryExpiresAt`, which shows the temporary-note clock marker on
+      // optimistic insert) can't drift from buildTree. `name: ""` because a
+      // freshly created page is untitled; `hasChildren: false` because it has no
+      // children yet.
+      const newNode: SpaceTreeNode = pageToTreeNode(createdPage, {
         name: "",
-        position: createdPage.position,
-        spaceId: createdPage.spaceId,
-        parentPageId: createdPage.parentPageId,
         hasChildren: false,
-        // Show the temporary-note icon immediately on optimistic insert.
-        temporaryExpiresAt: createdPage.temporaryExpiresAt,
-        children: [],
-      };
+      });
 
       // Read latest tree at call time. Without this, callers that mutate the
       // tree (e.g. lazy-load children on expand) immediately before calling
@@ -173,7 +171,22 @@ export function useTreeMutation(spaceId: string): UseTreeMutation {
       // optimistic node's id IS the real created page id (createdPage.id), so
       // the ids match exactly regardless of which path runs first.
       setData((prev) => {
-        if (treeModel.find(prev, newNode.id)) return prev;
+        const existing = treeModel.find(prev, newNode.id);
+        if (existing) {
+          // The server `addTreeNode` broadcast won the race and already inserted
+          // this node. Older broadcasts could omit `temporaryExpiresAt`, leaving
+          // a temporary note WITHOUT its clock marker until reload; patch it on
+          // from the authoritative create response so the marker shows now.
+          if (
+            newNode.temporaryExpiresAt &&
+            !(existing as SpaceTreeNode).temporaryExpiresAt
+          ) {
+            return treeModel.update(prev, newNode.id, {
+              temporaryExpiresAt: newNode.temporaryExpiresAt,
+            } as Partial<SpaceTreeNode>);
+          }
+          return prev;
+        }
         return treeModel.insert(prev, parentId, newNode, lastIndex);
       });
 

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

@@ -393,6 +393,101 @@ describe("handleCreate optimistic-insert idempotency (find-then-skip)", () => {
   });
 });
 
+// handleCreate race-guard temporaryExpiresAt patch: when the server's
+// addTreeNode broadcast wins the race and inserts the node BEFORE the optimistic
+// updater runs, the updater must not re-insert. Two sub-branches:
+//  (a) the node the broadcast inserted carries NO deadline (an older broadcast
+//      omitted it) while the authoritative create response DOES → patch the
+//      deadline on so the clock marker shows now, without a reload.
+//  (b) the existing node ALREADY has a deadline → do NOT overwrite it; return
+//      `prev` by reference (a no-op write).
+describe("handleCreate race-guard temporaryExpiresAt patch", () => {
+  type TN = TreeNode<{ name: string; temporaryExpiresAt?: string | null }>;
+
+  // Mirrors the setData updater in use-tree-mutation handleCreate.
+  const applyOptimisticInsert = (
+    tree: TN[],
+    parentId: string | null,
+    node: TN,
+    index: number,
+  ): TN[] => {
+    const existing = treeModel.find(tree, node.id) as TN | null;
+    if (existing) {
+      if (node.temporaryExpiresAt && !existing.temporaryExpiresAt) {
+        return treeModel.update(tree, node.id, {
+          temporaryExpiresAt: node.temporaryExpiresAt,
+        });
+      }
+      return tree;
+    }
+    return treeModel.insert(tree, parentId, node, index);
+  };
+
+  const fixtureTN: TN[] = [
+    { id: "a", name: "A" },
+    { id: "b", name: "B" },
+  ];
+
+  const deadline = "2026-07-01T00:00:00.000Z";
+
+  it("(a) patches temporaryExpiresAt when the existing node has none + the response carries a deadline", () => {
+    // Server broadcast won the race and inserted the node WITHOUT a deadline.
+    const afterServer = treeModel.insert(fixtureTN, null, {
+      id: "new",
+      name: "",
+    });
+    expect((treeModel.find(afterServer, "new") as TN).temporaryExpiresAt).toBe(
+      undefined,
+    );
+
+    // The authoritative create response carries the deadline.
+    const created: TN = { id: "new", name: "", temporaryExpiresAt: deadline };
+    const patched = applyOptimisticInsert(
+      afterServer,
+      null,
+      created,
+      afterServer.length,
+    );
+
+    // A new reference (the patch wrote) and the node now has the deadline...
+    expect(patched).not.toBe(afterServer);
+    expect((treeModel.find(patched, "new") as TN).temporaryExpiresAt).toBe(
+      deadline,
+    );
+    // ...and still exactly one node (no duplicate re-insert).
+    expect(patched.filter((n) => n.id === "new")).toHaveLength(1);
+  });
+
+  it("(b) does NOT overwrite an existing deadline; returns prev by reference", () => {
+    const existingDeadline = deadline;
+    // The node already exists WITH a deadline (the broadcast carried it).
+    const afterServer = treeModel.insert(fixtureTN, null, {
+      id: "new",
+      name: "",
+      temporaryExpiresAt: existingDeadline,
+    });
+
+    // The create response carries a DIFFERENT deadline; the guard must ignore it.
+    const created: TN = {
+      id: "new",
+      name: "",
+      temporaryExpiresAt: "2099-01-01T00:00:00.000Z",
+    };
+    const after = applyOptimisticInsert(
+      afterServer,
+      null,
+      created,
+      afterServer.length,
+    );
+
+    // prev returned by reference (no write) and the original deadline is kept.
+    expect(after).toBe(afterServer);
+    expect((treeModel.find(after, "new") as TN).temporaryExpiresAt).toBe(
+      existingDeadline,
+    );
+  });
+});
+
 // moveTreeNode socket-handler semantics: the receiver must place the moved node
 // by `position` (NOT index 0) and apply the `pageData` the payload carries so a
 // moved node's title/icon/chevron stay correct. This mirrors the reducer in

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

@@ -9,26 +9,45 @@ export function sortPositionKeys(keys: any[]) {
   });
 }
 
+/**
+ * Single canonical `IPage -> SpaceTreeNode` field mapper. Every place that
+ * materialises a tree node from a page (buildTree, the optimistic insert in
+ * handleCreate, restore, duplicate) routes through here so the field copy —
+ * crucially `temporaryExpiresAt` — can never silently drift between sites. The
+ * `overrides` cover the small per-site differences (e.g. `name: ""` for an
+ * optimistic create, `name: title || "Untitled"` for restore, `canEdit: true`
+ * for duplicate). The default `temporaryExpiresAt` comes straight off the page,
+ * so restore (which the server nulls) stays permanent and a temporary create
+ * keeps its clock marker without a reload.
+ */
+export function pageToTreeNode(
+  page: IPage,
+  overrides?: Partial<SpaceTreeNode>,
+): SpaceTreeNode {
+  return {
+    id: page.id,
+    slugId: page.slugId,
+    name: page.title,
+    icon: page.icon,
+    position: page.position,
+    hasChildren: page.hasChildren,
+    spaceId: page.spaceId,
+    parentPageId: page.parentPageId,
+    canEdit: page.canEdit ?? page.permissions?.canEdit,
+    isTemplate: page.isTemplate,
+    temporaryExpiresAt: page.temporaryExpiresAt,
+    children: [],
+    ...overrides,
+  };
+}
+
 export function buildTree(pages: IPage[]): SpaceTreeNode[] {
   const pageMap: Record<string, SpaceTreeNode> = {};
 
   const tree: SpaceTreeNode[] = [];
 
   pages.forEach((page) => {
-    pageMap[page.id] = {
-      id: page.id,
-      slugId: page.slugId,
-      name: page.title,
-      icon: page.icon,
-      position: page.position,
-      hasChildren: page.hasChildren,
-      spaceId: page.spaceId,
-      parentPageId: page.parentPageId,
-      canEdit: page.canEdit ?? page.permissions?.canEdit,
-      isTemplate: page.isTemplate,
-      temporaryExpiresAt: page.temporaryExpiresAt,
-      children: [],
-    };
+    pageMap[page.id] = pageToTreeNode(page);
   });
 
   // Defense-in-depth: a duplicate id in `pages` would push two references to the

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

@@ -0,0 +1,149 @@
+import { describe, it, expect, vi, beforeEach } from "vitest";
+import { render, screen, fireEvent, waitFor } from "@testing-library/react";
+import { MantineProvider } from "@mantine/core";
+import type { IShareAlias } from "@/features/share/types/share.types";
+
+// matchMedia / storage are stubbed globally in vitest.setup.ts.
+
+// The mutation + query hooks reach react-query/network; the availability probe
+// hits the API. Stub them so the section renders in isolation and we can drive
+// the exact branches (taken name -> hint, 409 -> reassign modal).
+const setMutateAsync = vi.fn();
+let currentAlias: IShareAlias | null = null;
+let availabilityResult: {
+  valid: boolean;
+  available: boolean;
+  currentPageId: string | null;
+} = { valid: true, available: true, currentPageId: null };
+
+vi.mock("@/features/share/queries/share-query.ts", () => ({
+  useShareAliasForPageQuery: () => ({ data: currentAlias }),
+  useSetShareAliasMutation: () => ({
+    mutateAsync: setMutateAsync,
+    isPending: false,
+  }),
+  useRemoveShareAliasMutation: () => ({
+    mutateAsync: vi.fn(),
+    isPending: false,
+  }),
+}));
+
+vi.mock("@/features/share/services/share-service.ts", () => ({
+  checkShareAliasAvailability: vi.fn(async () => availabilityResult),
+}));
+
+import ShareAliasSection from "./share-alias-section";
+
+const aliasRow = (alias: string, pageId: string): IShareAlias => ({
+  id: `alias-${alias}`,
+  workspaceId: "ws-1",
+  alias,
+  pageId,
+  creatorId: "user-1",
+  createdAt: new Date().toISOString(),
+  updatedAt: new Date().toISOString(),
+});
+
+function renderSection(pageId = "page-Y") {
+  return render(
+    <MantineProvider>
+      <ShareAliasSection pageId={pageId} readOnly={false} />
+    </MantineProvider>,
+  );
+}
+
+describe("ShareAliasSection — taken-name handling is never a dead end", () => {
+  beforeEach(() => {
+    setMutateAsync.mockReset();
+    currentAlias = null;
+    availabilityResult = { valid: true, available: true, currentPageId: null };
+  });
+
+  it("shows a 'will move it here' HINT (not a terminal error) when the name belongs to another page, and keeps Save enabled", async () => {
+    // Page Y already owns "bee"; the user retypes a name owned by page X.
+    currentAlias = aliasRow("bee", "page-Y");
+    availabilityResult = {
+      valid: true,
+      available: false,
+      currentPageId: "page-X",
+    };
+
+    renderSection("page-Y");
+    const input = screen.getByPlaceholderText("my-page") as HTMLInputElement;
+    fireEvent.change(input, { target: { value: "test2" } });
+
+    // The reassign hint replaces the old dead-end red error.
+    await waitFor(
+      () =>
+        expect(
+          screen.getByText(
+            "This address is in use. Saving will move it to this page.",
+          ),
+        ).toBeDefined(),
+      { timeout: 2000 },
+    );
+    // The old terminal "already in use" error must NOT be shown.
+    expect(screen.queryByText("This address is already in use")).toBeNull();
+
+    // Save stays enabled so the confirm-reassign flow can run.
+    const saveBtn = screen.getByRole("button", {
+      name: "Save",
+    }) as HTMLButtonElement;
+    expect(saveBtn.disabled).toBe(false);
+  });
+
+  it("opens the reassign-confirm modal on a 409 ALIAS_REASSIGN_REQUIRED (path forward, not a dead end)", async () => {
+    currentAlias = aliasRow("bee", "page-Y");
+    availabilityResult = {
+      valid: true,
+      available: false,
+      currentPageId: "page-X",
+    };
+    // The server rejects the un-confirmed save asking the client to confirm.
+    setMutateAsync.mockRejectedValueOnce({
+      status: 409,
+      response: {
+        status: 409,
+        data: {
+          code: "ALIAS_REASSIGN_REQUIRED",
+          currentPageId: "page-X",
+          currentPageTitle: "Alias Test Page X",
+        },
+      },
+    });
+
+    renderSection("page-Y");
+    const input = screen.getByPlaceholderText("my-page") as HTMLInputElement;
+    fireEvent.change(input, { target: { value: "test2" } });
+
+    const saveBtn = screen.getByRole("button", {
+      name: "Save",
+    }) as HTMLButtonElement;
+    await waitFor(() => expect(saveBtn.disabled).toBe(false), {
+      timeout: 2000,
+    });
+    fireEvent.click(saveBtn);
+
+    // First save sent WITHOUT confirmReassign.
+    await waitFor(() =>
+      expect(setMutateAsync).toHaveBeenCalledWith(
+        expect.objectContaining({ alias: "test2", confirmReassign: false }),
+      ),
+    );
+
+    // The "Move custom address?" confirm modal must appear (the path forward).
+    await waitFor(() =>
+      expect(screen.getByText("Move custom address?")).toBeDefined(),
+    );
+    expect(screen.getByRole("button", { name: "Move here" })).toBeDefined();
+
+    // Confirming retries WITH confirmReassign: true.
+    setMutateAsync.mockResolvedValueOnce(aliasRow("test2", "page-Y"));
+    fireEvent.click(screen.getByRole("button", { name: "Move here" }));
+    await waitFor(() =>
+      expect(setMutateAsync).toHaveBeenCalledWith(
+        expect.objectContaining({ alias: "test2", confirmReassign: true }),
+      ),
+    );
+  });
+});

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

@@ -120,8 +120,13 @@ export default function ShareAliasSection({
   };
 
   const showInvalid = normalized.length > 0 && !isValid;
-  const showTaken =
-    isValid && !unchanged && availability && !availability.available;
+  // The typed name is already in use by ANOTHER page. This is NOT a dead end:
+  // hitting Save triggers the server's 409 `ALIAS_REASSIGN_REQUIRED` and opens
+  // the "Move custom address?" confirm modal that retargets the address here.
+  // So surface it as an informational hint (not a terminal red error) and keep
+  // Save enabled, instead of looking like the address is unusable.
+  const reassignable =
+    isValid && !unchanged && !!availability && !availability.available;
 
   // The slug prefix (e.g. "docs.example.com/l/") is static for the session.
   const prefixLabel = aliasPrefixLabel();
@@ -198,9 +203,12 @@ export default function ShareAliasSection({
         error={
           showInvalid
             ? t("Use 2-60 lowercase letters, digits and hyphens")
-            : showTaken
-              ? t("This address is already in use")
-              : undefined
+            : undefined
+        }
+        description={
+          reassignable
+            ? t("This address is in use. Saving will move it to this page.")
+            : undefined
         }
       />
 

apps/client/src/features/space/components/space-create-note-buttons.tsx

@@ -1,5 +1,5 @@
 import { useState } from "react";
-import { Button, Group } from "@mantine/core";
+import { Button, Stack } from "@mantine/core";
 import { IconHourglass, IconPlus } from "@tabler/icons-react";
 import { useParams } from "react-router-dom";
 import { useTranslation } from "react-i18next";
@@ -45,12 +45,16 @@ export default function SpaceCreateNoteButtons() {
       .finally(() => setPending(null));
   };
 
+  // Two full-width, vertically stacked buttons: a neutral regular note and an
+  // orange-tinted temporary note. Stacking full-width keeps the longer "New
+  // temporary note" label from clipping on narrow mobile widths.
   return (
-    <Group grow gap="sm">
+    <Stack gap="sm">
       <Button
         size="md"
         variant="light"
         color="gray"
+        fullWidth
         leftSection={<IconPlus size={18} />}
         loading={pending === "regular"}
         disabled={pending !== null}
@@ -61,7 +65,8 @@ export default function SpaceCreateNoteButtons() {
       <Button
         size="md"
         variant="light"
-        color="gray"
+        color="orange"
+        fullWidth
         leftSection={<IconHourglass size={18} />}
         loading={pending === "temporary"}
         disabled={pending !== null}
@@ -69,6 +74,6 @@ export default function SpaceCreateNoteButtons() {
       >
         {t("New temporary note")}
       </Button>
-    </Group>
+    </Stack>
   );
 }

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

@@ -323,4 +323,18 @@ describe("applyAddTreeNode", () => {
       "child",
     ]);
   });
+
+  it("carries temporaryExpiresAt onto the inserted node so the clock marker shows on create (no reload)", () => {
+    // A note created as temporary broadcasts addTreeNode with the death-timer
+    // deadline in its payload; the receiver's inserted node must keep it so
+    // space-tree-row renders the orange clock marker immediately.
+    const tree = roots();
+    const expiresAt = "2026-06-27T21:00:00.000Z";
+    const next = applyAddTreeNode(tree, {
+      parentId: null as unknown as string,
+      index: 0,
+      data: node("temp", { position: "a3", temporaryExpiresAt: expiresAt }),
+    });
+    expect(treeModel.find(next, "temp")?.temporaryExpiresAt).toBe(expiresAt);
+  });
 });

apps/client/src/features/workspace/components/settings/components/ai-agent-roles-catalog-modal.tsx

@@ -0,0 +1,407 @@
+import { useEffect, useMemo, useState } from "react";
+import {
+  Accordion,
+  Alert,
+  Badge,
+  Button,
+  Center,
+  Checkbox,
+  Group,
+  Loader,
+  Modal,
+  Radio,
+  Select,
+  Stack,
+  Text,
+} from "@mantine/core";
+import { IconAlertTriangle } from "@tabler/icons-react";
+import { useTranslation } from "react-i18next";
+import {
+  useAiRoleCatalogBundleQuery,
+  useAiRoleCatalogQuery,
+  useImportAiRolesFromCatalogMutation,
+  useUpdateAiRoleFromCatalogMutation,
+} from "@/features/ai-chat/queries/ai-chat-query.ts";
+import {
+  IAiRole,
+  IAiRoleCatalogBundleSummary,
+  IAiRoleCatalogRole,
+} from "@/features/ai-chat/types/ai-chat.types.ts";
+import { catalogRoleInstallState } from "@/features/ai-chat/utils/catalog-role-install-state.ts";
+
+interface AiAgentRolesCatalogModalProps {
+  opened: boolean;
+  onClose: () => void;
+  // The current admin role list (full view, including `source`). Used to compute
+  // each catalog role's install state (import / installed / update available).
+  roles: IAiRole[];
+}
+
+/** How a name collision with an existing role is handled on import. */
+type Conflict = "skip" | "rename";
+
+/**
+ * Admin modal: browse the curated role catalog, import roles, and update an
+ * imported role when the catalog ships a newer version.
+ *
+ * Import is per-bundle (the endpoint takes a single bundleId). Each bundle's
+ * Accordion panel has its own "Import" button that imports only that bundle's
+ * checked roles — the simplest mapping to the one-bundle-per-call API and the
+ * clearest UX. Selection state is tracked per bundle.
+ */
+export default function AiAgentRolesCatalogModal({
+  opened,
+  onClose,
+  roles,
+}: AiAgentRolesCatalogModalProps) {
+  const { t, i18n } = useTranslation();
+
+  // The user's i18n base subtag (e.g. "ru-RU" => "ru"); the preferred catalog
+  // language both when seeding and when reconciling against offered languages.
+  const baseLang = (i18n.language || "en").split("-")[0].toLowerCase();
+
+  // Fetch the catalog only while the modal is open. `language` drives both the
+  // catalog query (bundle names) and bundle reads (role content). Seed it
+  // synchronously from the base subtag so the first fetch already uses the
+  // user's language; the effect below still reconciles against the catalog's
+  // offered languages once they load.
+  const [language, setLanguage] = useState<string>(() => baseLang);
+  const catalogQuery = useAiRoleCatalogQuery(language || "en", opened);
+
+  // On name conflict: Skip (default) or Rename to a free " (N)" name.
+  const [conflict, setConflict] = useState<Conflict>("skip");
+
+  // The currently expanded bundle id (Accordion is single-open: one bundle's
+  // roles are fetched at a time).
+  const [expanded, setExpanded] = useState<string | null>(null);
+
+  // Per-bundle selected slugs (import-state roles checked for import).
+  const [selected, setSelected] = useState<Record<string, Set<string>>>({});
+
+  const languages = catalogQuery.data?.languages;
+
+  // Pick a sensible default language from the catalog once it loads: the i18n
+  // base subtag (e.g. "ru-RU" => "ru") if offered, else "en", else the first.
+  useEffect(() => {
+    if (!languages || languages.length === 0) return;
+    if (language && languages.includes(language)) return;
+    const preferred = languages.includes(baseLang)
+      ? baseLang
+      : languages.includes("en")
+        ? "en"
+        : languages[0];
+    setLanguage(preferred);
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [languages]);
+
+  // Reset per-language UI state when the language changes (the bundle content,
+  // hence the install computations, are language-specific).
+  useEffect(() => {
+    setExpanded(null);
+    setSelected({});
+  }, [language]);
+
+  return (
+    <Modal
+      opened={opened}
+      onClose={onClose}
+      title={t("Role catalog")}
+      size="lg"
+    >
+      <Stack>
+        <Select
+          label={t("Language")}
+          data={languages ?? []}
+          value={language || null}
+          onChange={(value) => value && setLanguage(value)}
+          allowDeselect={false}
+          disabled={!languages || languages.length === 0}
+          comboboxProps={{ withinPortal: true }}
+        />
+
+        <Radio.Group
+          label={t("On name conflict")}
+          value={conflict}
+          onChange={(value) => setConflict(value as Conflict)}
+        >
+          <Group mt="xs">
+            <Radio value="skip" label={t("Skip")} />
+            <Radio value="rename" label={t("Rename")} />
+          </Group>
+        </Radio.Group>
+
+        {catalogQuery.isLoading && (
+          <Center py="lg">
+            <Loader size="sm" />
+          </Center>
+        )}
+
+        {catalogQuery.isError && (
+          <Alert
+            color="red"
+            icon={<IconAlertTriangle size={16} />}
+            title={t("The role catalog is unavailable")}
+          >
+            {t("Please try again later.")}
+          </Alert>
+        )}
+
+        {catalogQuery.data && catalogQuery.data.bundles.length === 0 && (
+          <Text size="sm" c="dimmed">
+            {t("No bundles available")}
+          </Text>
+        )}
+
+        {catalogQuery.data && catalogQuery.data.bundles.length > 0 && (
+          <Accordion
+            variant="separated"
+            value={expanded}
+            onChange={setExpanded}
+          >
+            {catalogQuery.data.bundles.map((bundle) => (
+              <BundlePanel
+                key={bundle.id}
+                bundle={bundle}
+                language={language}
+                expanded={expanded === bundle.id}
+                roles={roles}
+                conflict={conflict}
+                selected={selected[bundle.id]}
+                onToggleSlug={(slug, checked) =>
+                  setSelected((prev) => {
+                    const next = new Set(prev[bundle.id] ?? []);
+                    if (checked) next.add(slug);
+                    else next.delete(slug);
+                    return { ...prev, [bundle.id]: next };
+                  })
+                }
+                onSetSelected={(slugs) =>
+                  setSelected((prev) => ({
+                    ...prev,
+                    [bundle.id]: new Set(slugs),
+                  }))
+                }
+              />
+            ))}
+          </Accordion>
+        )}
+
+        <Group justify="flex-end" mt="sm">
+          <Button variant="default" onClick={onClose}>
+            {t("Close")}
+          </Button>
+        </Group>
+      </Stack>
+    </Modal>
+  );
+}
+
+interface BundlePanelProps {
+  bundle: IAiRoleCatalogBundleSummary;
+  language: string;
+  expanded: boolean;
+  roles: IAiRole[];
+  conflict: Conflict;
+  selected: Set<string> | undefined;
+  onToggleSlug: (slug: string, checked: boolean) => void;
+  onSetSelected: (slugs: string[]) => void;
+}
+
+/** One catalog bundle: its roles (fetched when expanded) + a per-bundle import. */
+function BundlePanel({
+  bundle,
+  language,
+  expanded,
+  roles,
+  conflict,
+  selected,
+  onToggleSlug,
+  onSetSelected,
+}: BundlePanelProps) {
+  const { t } = useTranslation();
+
+  // Only fetch this bundle's roles once it is actually expanded.
+  const bundleQuery = useAiRoleCatalogBundleQuery(
+    bundle.id,
+    language,
+    expanded && !!language,
+  );
+
+  const importMutation = useImportAiRolesFromCatalogMutation();
+  const updateMutation = useUpdateAiRoleFromCatalogMutation();
+
+  // Compute each catalog role's install state against the current workspace
+  // roles (matched by source.slug + source.language). The decision lives in the
+  // pure `catalogRoleInstallState` helper so it is unit-tested directly.
+  const computed = useMemo(() => {
+    const list = bundleQuery.data?.roles ?? [];
+    return list.map((role) => ({
+      role,
+      ...catalogRoleInstallState(role, roles, language),
+    }));
+  }, [bundleQuery.data, roles, language]);
+
+  // Default-check every importable role once the bundle content arrives (unless
+  // the user already touched the selection for this bundle).
+  useEffect(() => {
+    if (!bundleQuery.data || selected !== undefined) return;
+    onSetSelected(
+      computed.filter((c) => c.state === "import").map((c) => c.role.slug),
+    );
+    // eslint-disable-next-line react-hooks/exhaustive-deps
+  }, [bundleQuery.data]);
+
+  const importableSlugs = computed
+    .filter((c) => c.state === "import")
+    .map((c) => c.role.slug);
+  const checkedSlugs = importableSlugs.filter((slug) => selected?.has(slug));
+
+  function handleImport() {
+    importMutation.mutate({
+      bundleId: bundle.id,
+      language,
+      slugs: checkedSlugs,
+      conflict,
+    });
+  }
+
+  return (
+    <Accordion.Item value={bundle.id}>
+      <Accordion.Control>
+        <Stack gap={2}>
+          <Text fw={500}>{bundle.name}</Text>
+          {bundle.description && (
+            <Text size="xs" c="dimmed">
+              {bundle.description}
+            </Text>
+          )}
+        </Stack>
+      </Accordion.Control>
+      <Accordion.Panel>
+        {bundleQuery.isLoading && (
+          <Center py="md">
+            <Loader size="sm" />
+          </Center>
+        )}
+
+        {bundleQuery.isError && (
+          <Alert
+            color="red"
+            icon={<IconAlertTriangle size={16} />}
+            title={t("The role catalog is unavailable")}
+          >
+            {t("Please try again later.")}
+          </Alert>
+        )}
+
+        {bundleQuery.data && (
+          <Stack gap="xs">
+            {computed.map((entry) => (
+              <CatalogRoleRow
+                key={entry.role.slug}
+                role={entry.role}
+                state={entry.state}
+                checked={
+                  entry.state === "import"
+                    ? !!selected?.has(entry.role.slug)
+                    : false
+                }
+                onToggle={(checked) => onToggleSlug(entry.role.slug, checked)}
+                fromVersion={
+                  entry.state === "update" ? entry.fromVersion : undefined
+                }
+                onUpdate={
+                  entry.state === "update"
+                    ? () => updateMutation.mutate(entry.installed.id)
+                    : undefined
+                }
+                updating={updateMutation.isPending}
+              />
+            ))}
+
+            <Group justify="flex-end" mt="xs">
+              <Button
+                size="xs"
+                onClick={handleImport}
+                loading={importMutation.isPending}
+                disabled={checkedSlugs.length === 0}
+              >
+                {t("Import")}
+              </Button>
+            </Group>
+          </Stack>
+        )}
+      </Accordion.Panel>
+    </Accordion.Item>
+  );
+}
+
+interface CatalogRoleRowProps {
+  role: IAiRoleCatalogRole;
+  state: "import" | "installed" | "update";
+  checked: boolean;
+  onToggle: (checked: boolean) => void;
+  // The installed role's current source version (only set in the "update" state).
+  fromVersion?: number;
+  onUpdate?: () => void;
+  updating: boolean;
+}
+
+/** A single catalog role row with its install-state affordance. */
+function CatalogRoleRow({
+  role,
+  state,
+  checked,
+  onToggle,
+  fromVersion,
+  onUpdate,
+  updating,
+}: CatalogRoleRowProps) {
+  const { t } = useTranslation();
+
+  return (
+    <Group justify="space-between" wrap="nowrap" align="flex-start">
+      <Group gap="xs" wrap="nowrap" align="flex-start" style={{ minWidth: 0 }}>
+        {state === "import" && (
+          <Checkbox
+            checked={checked}
+            onChange={(event) => onToggle(event.currentTarget.checked)}
+            aria-label={role.name}
+          />
+        )}
+        <Stack gap={2} style={{ minWidth: 0 }}>
+          <Text fw={500} truncate>
+            {role.emoji ? `${role.emoji} ` : ""}
+            {role.name}
+          </Text>
+          {role.description && (
+            <Text size="xs" c="dimmed">
+              {role.description}
+            </Text>
+          )}
+        </Stack>
+      </Group>
+
+      <Group gap="xs" wrap="nowrap" style={{ flex: "none" }}>
+        {state === "installed" && (
+          <Badge size="sm" variant="light" color="gray">
+            {t("Installed")}
+          </Badge>
+        )}
+        {state === "update" && (
+          <>
+            <Badge size="sm" variant="light" color="blue">
+              {t("v{{from}} → v{{to}}", {
+                from: fromVersion ?? 0,
+                to: role.version,
+              })}
+            </Badge>
+            <Button size="xs" variant="light" onClick={onUpdate} loading={updating}>
+              {t("Update")}
+            </Button>
+          </>
+        )}
+      </Group>
+    </Group>
+  );
+}

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

@@ -13,7 +13,12 @@ import {
 } from "@mantine/core";
 import { useDisclosure } from "@mantine/hooks";
 import { modals } from "@mantine/modals";
-import { IconPencil, IconPlus, IconTrash } from "@tabler/icons-react";
+import {
+  IconPackageImport,
+  IconPencil,
+  IconPlus,
+  IconTrash,
+} from "@tabler/icons-react";
 import { useTranslation } from "react-i18next";
 import useUserRole from "@/hooks/use-user-role.tsx";
 import {
@@ -23,6 +28,7 @@ import {
 } from "@/features/ai-chat/queries/ai-chat-query.ts";
 import { IAiRole } from "@/features/ai-chat/types/ai-chat.types.ts";
 import AiAgentRoleForm from "./ai-agent-role-form.tsx";
+import AiAgentRolesCatalogModal from "./ai-agent-roles-catalog-modal.tsx";
 
 /**
  * Admin section: list / add / edit / delete reusable agent roles. A role
@@ -39,6 +45,9 @@ export default function AiAgentRoles() {
   const deleteMutation = useDeleteAiRoleMutation();
 
   const [opened, { open, close }] = useDisclosure(false);
+  // Separate disclosure for the catalog (import/update) modal.
+  const [catalogOpened, { open: openCatalog, close: closeCatalog }] =
+    useDisclosure(false);
   // The role being edited; undefined => the modal is in "create" mode.
   const [editing, setEditing] = useState<IAiRole | undefined>(undefined);
 
@@ -86,14 +95,24 @@ export default function AiAgentRoles() {
           />
           <Text fw={600}>{t("Agent roles")}</Text>
         </Group>
-        <Button
-          leftSection={<IconPlus size={16} />}
-          variant="default"
-          size="xs"
-          onClick={openCreate}
-        >
-          {t("Add role")}
-        </Button>
+        <Group gap="xs" wrap="nowrap">
+          <Button
+            leftSection={<IconPackageImport size={16} />}
+            variant="default"
+            size="xs"
+            onClick={openCatalog}
+          >
+            {t("Import from catalog")}
+          </Button>
+          <Button
+            leftSection={<IconPlus size={16} />}
+            variant="default"
+            size="xs"
+            onClick={openCreate}
+          >
+            {t("Add role")}
+          </Button>
+        </Group>
       </Group>
       <Text size="xs" c="dimmed" mt={4}>
         {t(
@@ -102,9 +121,19 @@ export default function AiAgentRoles() {
       </Text>
 
       {!isLoading && (!roles || roles.length === 0) && (
-        <Text size="sm" c="dimmed" mt="sm">
-          {t("No roles configured")}
-        </Text>
+        <Group gap="sm" mt="sm" align="center">
+          <Text size="sm" c="dimmed">
+            {t("No roles configured")}
+          </Text>
+          <Button
+            leftSection={<IconPackageImport size={16} />}
+            variant="light"
+            size="xs"
+            onClick={openCatalog}
+          >
+            {t("Browse the catalog")}
+          </Button>
+        </Group>
       )}
 
       <Stack gap="xs" mt="sm">
@@ -170,6 +199,12 @@ export default function AiAgentRoles() {
         {/* Remount the form per target so its internal state re-hydrates. */}
         <AiAgentRoleForm key={editing?.id ?? "new"} role={editing} onClose={close} />
       </Modal>
+
+      <AiAgentRolesCatalogModal
+        opened={catalogOpened}
+        onClose={closeCatalog}
+        roles={roles ?? []}
+      />
     </Paper>
   );
 }

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

@@ -0,0 +1,44 @@
+import { AiChatController } from './ai-chat.controller';
+import type { User, Workspace } from '@docmost/db/types/entity.types';
+
+/**
+ * Wiring spec for the #191 `POST /ai-chat/bound-chat` endpoint. It must forward
+ * the requesting user + workspace + pageId to findLatestByPage and return the
+ * matched chat's id, or `{ chatId: null }` when there is none. The repo already
+ * scopes to the caller's OWN chats, so a foreign pageId simply yields no match
+ * (null) — no extra page-access check is needed. Exercised with hand-rolled
+ * mocks, no Nest graph and no DB.
+ */
+describe('AiChatController.boundChat', () => {
+  const user = { id: 'u1' } as User;
+  const workspace = { id: 'ws1' } as Workspace;
+
+  function makeController(chat: unknown) {
+    const aiChatRepo = {
+      findLatestByPage: jest.fn().mockResolvedValue(chat),
+    };
+    const controller = new AiChatController(
+      {} as never,
+      aiChatRepo as never,
+      {} as never,
+      {} as never,
+    );
+    return { controller, aiChatRepo };
+  }
+
+  it('returns the owned chat id and scopes the lookup to user + workspace + page', async () => {
+    const { controller, aiChatRepo } = makeController({
+      id: 'c1',
+      creatorId: 'u1',
+    });
+    const res = await controller.boundChat({ pageId: 'p1' }, user, workspace);
+    expect(aiChatRepo.findLatestByPage).toHaveBeenCalledWith('u1', 'ws1', 'p1');
+    expect(res).toEqual({ chatId: 'c1' });
+  });
+
+  it('returns { chatId: null } for a page with no owned chat (incl. foreign pageId)', async () => {
+    const { controller } = makeController(undefined);
+    const res = await controller.boundChat({ pageId: 'foreign' }, user, workspace);
+    expect(res).toEqual({ chatId: null });
+  });
+});

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

@@ -30,6 +30,7 @@ import { FileInterceptor } from '../../common/interceptors/file.interceptor';
 import { AiChatService, AiChatStreamBody } from './ai-chat.service';
 import { AiTranscriptionService } from './ai-transcription.service';
 import {
+  BoundChatDto,
   ChatIdDto,
   ExportChatDto,
   GeneratePageTitleDto,
@@ -67,6 +68,28 @@ export class AiChatController {
     return this.aiChatRepo.findByCreator(user.id, workspace.id, pagination);
   }
 
+  /**
+   * Resolve the chat bound to a document for the requesting user: the most-recent
+   * non-deleted chat created on that page (ai_chats.page_id). Returns
+   * { chatId: null } when the page has no owned chat (-> a fresh chat). No page
+   * access check needed: only the caller's OWN chats are matched, so a foreign
+   * pageId reveals nothing.
+   */
+  @HttpCode(HttpStatus.OK)
+  @Post('bound-chat')
+  async boundChat(
+    @Body() dto: BoundChatDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ): Promise<{ chatId: string | null }> {
+    const chat = await this.aiChatRepo.findLatestByPage(
+      user.id,
+      workspace.id,
+      dto.pageId,
+    );
+    return { chatId: chat?.id ?? null };
+  }
+
   /** Fetch the messages of a chat (oldest first, paginated). */
   @HttpCode(HttpStatus.OK)
   @Post('messages')

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

@@ -37,6 +37,12 @@ export class GetChatMessagesDto {
   cursor?: string;
 }
 
+/** Resolve the chat bound to a document (the page's most-recent owned chat). */
+export class BoundChatDto {
+  @IsString()
+  pageId: string;
+}
+
 /** Export a chat to Markdown (#183). `lang` localizes the few fixed
  *  role/tool-action labels; defaults to English server-side. */
 export class ExportChatDto {

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

@@ -39,6 +39,10 @@ describe('AiAgentRolesController admin gate', () => {
       create: jest.fn().mockResolvedValue({ id: 'r1' }),
       update: jest.fn().mockResolvedValue({ id: 'r1' }),
       remove: jest.fn().mockResolvedValue({ success: true }),
+      getCatalog: jest.fn().mockResolvedValue({ languages: [], bundles: [] }),
+      getCatalogBundle: jest.fn().mockResolvedValue({ roles: [] }),
+      importFromCatalog: jest.fn().mockResolvedValue({ created: 0 }),
+      updateFromCatalog: jest.fn().mockResolvedValue({ updated: false }),
     };
     const controller = new AiAgentRolesController(
       rolesService as never,
@@ -109,6 +113,90 @@ describe('AiAgentRolesController admin gate', () => {
     });
   });
 
+  // Catalog routes (browse + import) are ALL admin-only: a non-admin caller must
+  // get ForbiddenException with the service untouched; an admin delegates with
+  // the right arguments (import/update-from-catalog carry workspace.id).
+  describe('catalog routes admin gate', () => {
+    const catalogDto = { language: 'en' } as never;
+    const bundleDto = { bundleId: 'general', language: 'en' } as never;
+    const importDto = {
+      bundleId: 'general',
+      language: 'en',
+      conflict: 'skip',
+    } as never;
+    const updateDto = { id: 'r1' } as never;
+
+    describe('non-admin is rejected and the service is NOT called', () => {
+      it('catalog', async () => {
+        const { controller, rolesService } = makeController(false);
+        await expect(
+          controller.catalog(catalogDto, user, workspace),
+        ).rejects.toBeInstanceOf(ForbiddenException);
+        expect(rolesService.getCatalog).not.toHaveBeenCalled();
+      });
+
+      it('catalog/bundle', async () => {
+        const { controller, rolesService } = makeController(false);
+        await expect(
+          controller.catalogBundle(bundleDto, user, workspace),
+        ).rejects.toBeInstanceOf(ForbiddenException);
+        expect(rolesService.getCatalogBundle).not.toHaveBeenCalled();
+      });
+
+      it('import', async () => {
+        const { controller, rolesService } = makeController(false);
+        await expect(
+          controller.import(importDto, user, workspace),
+        ).rejects.toBeInstanceOf(ForbiddenException);
+        expect(rolesService.importFromCatalog).not.toHaveBeenCalled();
+      });
+
+      it('update-from-catalog', async () => {
+        const { controller, rolesService } = makeController(false);
+        await expect(
+          controller.updateFromCatalog(updateDto, user, workspace),
+        ).rejects.toBeInstanceOf(ForbiddenException);
+        expect(rolesService.updateFromCatalog).not.toHaveBeenCalled();
+      });
+    });
+
+    describe('admin delegates to the service', () => {
+      it('catalog passes the requested language', async () => {
+        const { controller, rolesService } = makeController(true);
+        await controller.catalog(catalogDto, user, workspace);
+        expect(rolesService.getCatalog).toHaveBeenCalledWith('en');
+      });
+
+      it('catalog/bundle passes bundleId + language', async () => {
+        const { controller, rolesService } = makeController(true);
+        await controller.catalogBundle(bundleDto, user, workspace);
+        expect(rolesService.getCatalogBundle).toHaveBeenCalledWith(
+          'general',
+          'en',
+        );
+      });
+
+      it('import passes workspace.id + user.id + dto', async () => {
+        const { controller, rolesService } = makeController(true);
+        await controller.import(importDto, user, workspace);
+        expect(rolesService.importFromCatalog).toHaveBeenCalledWith(
+          'ws-1',
+          'u1',
+          importDto,
+        );
+      });
+
+      it('update-from-catalog passes workspace.id + dto', async () => {
+        const { controller, rolesService } = makeController(true);
+        await controller.updateFromCatalog(updateDto, user, workspace);
+        expect(rolesService.updateFromCatalog).toHaveBeenCalledWith(
+          'ws-1',
+          updateDto,
+        );
+      });
+    });
+  });
+
   describe('list (member-reachable)', () => {
     it('non-admin reaches list and the service is asked for the picker view (isAdmin=false)', async () => {
       const { controller, rolesService } = makeController(false);

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

@@ -22,6 +22,12 @@ import {
   CreateAgentRoleDto,
   UpdateAgentRoleDto,
 } from './dto/agent-role.dto';
+import {
+  CatalogBundleDto,
+  CatalogQueryDto,
+  ImportFromCatalogDto,
+  UpdateFromCatalogDto,
+} from './dto/agent-role-catalog.dto';
 
 /** Path/body param for the per-role routes (update/delete). */
 class AgentRoleIdDto {
@@ -113,4 +119,54 @@ export class AiAgentRolesController {
     this.assertAdmin(user, workspace);
     return this.rolesService.remove(workspace.id, idDto.id);
   }
+
+  // --- Catalog (admin-only): browse + import + update imported roles. ---
+
+  /** Browse the curated catalog (localized to dto.language). */
+  @HttpCode(HttpStatus.OK)
+  @Post('catalog')
+  async catalog(
+    @Body() dto: CatalogQueryDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ) {
+    this.assertAdmin(user, workspace);
+    return this.rolesService.getCatalog(dto.language);
+  }
+
+  /** Open one catalog bundle in a language (role content + versions). */
+  @HttpCode(HttpStatus.OK)
+  @Post('catalog/bundle')
+  async catalogBundle(
+    @Body() dto: CatalogBundleDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ) {
+    this.assertAdmin(user, workspace);
+    return this.rolesService.getCatalogBundle(dto.bundleId, dto.language);
+  }
+
+  /** Import roles from a catalog bundle into the workspace. */
+  @HttpCode(HttpStatus.OK)
+  @Post('import')
+  async import(
+    @Body() dto: ImportFromCatalogDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ) {
+    this.assertAdmin(user, workspace);
+    return this.rolesService.importFromCatalog(workspace.id, user.id, dto);
+  }
+
+  /** Update an already-imported role from its catalog source. */
+  @HttpCode(HttpStatus.OK)
+  @Post('update-from-catalog')
+  async updateFromCatalog(
+    @Body() dto: UpdateFromCatalogDto,
+    @AuthUser() user: User,
+    @AuthWorkspace() workspace: Workspace,
+  ) {
+    this.assertAdmin(user, workspace);
+    return this.rolesService.updateFromCatalog(workspace.id, dto);
+  }
 }

apps/server/src/core/ai-chat/roles/ai-agent-roles.module.ts

@@ -1,16 +1,19 @@
 import { Module } from '@nestjs/common';
 import { AiAgentRolesController } from './ai-agent-roles.controller';
 import { AiAgentRolesService } from './ai-agent-roles.service';
+import { AiAgentRolesCatalogProvider } from './catalog/ai-agent-roles-catalog.provider';
 
 /**
  * Agent roles unit (v1). Admin CRUD + member-visible listing for the chat
- * role picker. AiAgentRoleRepo (DatabaseModule, global) and
- * WorkspaceAbilityFactory (CaslModule, global) are resolved without explicit
- * imports. The stream-time role resolution + model override live in
- * AiChatService / AiService; this module only hosts the management API.
+ * role picker, plus the admin catalog (browse/import/update). AiAgentRoleRepo
+ * (DatabaseModule, global), WorkspaceAbilityFactory (CaslModule, global) and
+ * EnvironmentService (EnvironmentModule, global — used by the catalog provider)
+ * are resolved without explicit imports. The stream-time role resolution +
+ * model override live in AiChatService / AiService; this module only hosts the
+ * management API.
  */
 @Module({
   controllers: [AiAgentRolesController],
-  providers: [AiAgentRolesService],
+  providers: [AiAgentRolesService, AiAgentRolesCatalogProvider],
 })
 export class AiAgentRolesModule {}

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

@@ -1,4 +1,9 @@
-import { BadRequestException, ConflictException } from '@nestjs/common';
+import {
+  BadGatewayException,
+  BadRequestException,
+  ConflictException,
+  Logger,
+} from '@nestjs/common';
 import { AiAgentRolesService } from './ai-agent-roles.service';
 import type { AiAgentRole } from '@docmost/db/types/entity.types';
 import type {
@@ -27,12 +32,22 @@ describe('AiAgentRolesService guards', () => {
       enabled: true,
       autoStart: true,
       launchMessage: null,
+      source: null,
       createdAt: new Date(),
       updatedAt: new Date(),
       ...over,
     } as AiAgentRole;
   }
 
+  // A stubbed catalog provider; the CRUD tests never reach it (they exercise
+  // create/update/remove/list only), so the methods just reject if hit.
+  function makeCatalog() {
+    return {
+      fetchIndex: jest.fn(),
+      fetchBundle: jest.fn(),
+    };
+  }
+
   function makeService(opts: { existing?: AiAgentRole | undefined } = {}) {
     const repo = {
       findById: jest.fn().mockResolvedValue(opts.existing),
@@ -41,8 +56,9 @@ describe('AiAgentRolesService guards', () => {
       softDelete: jest.fn().mockResolvedValue(undefined),
       listByWorkspace: jest.fn().mockResolvedValue([]),
     };
-    const service = new AiAgentRolesService(repo as never);
-    return { service, repo };
+    const catalog = makeCatalog();
+    const service = new AiAgentRolesService(repo as never, catalog as never);
+    return { service, repo, catalog };
   }
 
   describe('update', () => {
@@ -163,6 +179,7 @@ describe('AiAgentRolesService guards', () => {
         enabled: false,
         autoStart: true,
         launchMessage: null,
+        source: null,
         createdAt,
         updatedAt,
       });
@@ -397,7 +414,7 @@ describe('AiAgentRolesService guards', () => {
         softDelete: jest.fn(),
         listByWorkspace: jest.fn().mockResolvedValue(rows),
       };
-      const service = new AiAgentRolesService(repo as never);
+      const service = new AiAgentRolesService(repo as never, makeCatalog() as never);
       return { service, repo };
     }
 
@@ -461,4 +478,630 @@ describe('AiAgentRolesService guards', () => {
       ).rejects.toBeInstanceOf(ConflictException);
     });
   });
+
+  // ---------------------------------------------------------------------------
+  // Catalog: import (skip / rename / already-installed) and update reconciliation
+  // against a MOCKED catalog provider + mocked repo (mirrors the CRUD style).
+  // ---------------------------------------------------------------------------
+  describe('importFromCatalog', () => {
+    function catalogRole(over: Record<string, unknown> = {}) {
+      return {
+        slug: 'researcher',
+        name: 'Researcher',
+        instructions: 'be a researcher',
+        ...over,
+      };
+    }
+
+    function makeImportService(opts: {
+      indexRoles?: { slug: string; version: number }[];
+      bundleRoles?: Record<string, unknown>[];
+      existing?: AiAgentRole[];
+    }) {
+      const index = {
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: opts.indexRoles ?? [{ slug: 'researcher', version: 3 }],
+          },
+        ],
+      };
+      const bundle = {
+        schemaVersion: 1,
+        language: 'en',
+        roles: opts.bundleRoles ?? [catalogRole()],
+      };
+      const repo = {
+        findById: jest.fn(),
+        insert: jest.fn().mockImplementation((v) => Promise.resolve(makeRow(v))),
+        update: jest.fn().mockResolvedValue(undefined),
+        softDelete: jest.fn(),
+        listByWorkspace: jest.fn().mockResolvedValue(opts.existing ?? []),
+      };
+      const catalog = {
+        fetchIndex: jest.fn().mockResolvedValue(index),
+        fetchBundle: jest.fn().mockResolvedValue(bundle),
+      };
+      const service = new AiAgentRolesService(repo as never, catalog as never);
+      return { service, repo, catalog };
+    }
+
+    const dto = (over: Record<string, unknown> = {}) =>
+      ({
+        bundleId: 'general',
+        language: 'en',
+        conflict: 'skip',
+        ...over,
+      }) as never;
+
+    it('inserts a new role with source { slug, language, version } from the index', async () => {
+      const { service, repo } = makeImportService({});
+      const res = await service.importFromCatalog('ws-1', 'u1', dto());
+      expect(res).toMatchObject({ created: 1, skipped: 0, renamed: 0 });
+      expect(res.errors).toEqual([]);
+      const values = repo.insert.mock.calls[0][0];
+      expect(values.source).toEqual({
+        slug: 'researcher',
+        language: 'en',
+        version: 3,
+      });
+      expect(values.enabled).toBe(true);
+    });
+
+    it('already-installed catalog slug => skipped (no insert)', async () => {
+      const existing = [
+        makeRow({
+          id: 'r-existing',
+          name: 'Old researcher',
+          source: { slug: 'researcher', language: 'en', version: 1 } as never,
+        }),
+      ];
+      const { service, repo } = makeImportService({ existing });
+      const res = await service.importFromCatalog('ws-1', 'u1', dto());
+      expect(res).toMatchObject({ created: 0, skipped: 1, renamed: 0 });
+      expect(repo.insert).not.toHaveBeenCalled();
+    });
+
+    it('same slug installed in a DIFFERENT language => NOT skipped (separate install)', async () => {
+      // Installed as `ru`; importing the `en` variant of the same slug must
+      // still import (dedup key is slug+language, matching the client UI).
+      const existing = [
+        makeRow({
+          id: 'r-ru',
+          name: 'Исследователь',
+          source: { slug: 'researcher', language: 'ru', version: 1 } as never,
+        }),
+      ];
+      const { service, repo } = makeImportService({ existing });
+      const res = await service.importFromCatalog('ws-1', 'u1', dto());
+      expect(res).toMatchObject({ created: 1, skipped: 0, renamed: 0 });
+      expect(repo.insert).toHaveBeenCalledTimes(1);
+      expect(repo.insert.mock.calls[0][0].source).toEqual({
+        slug: 'researcher',
+        language: 'en',
+        version: 3,
+      });
+    });
+
+    it('name collision + conflict:skip => skipped (no insert)', async () => {
+      const existing = [makeRow({ id: 'r-x', name: 'Researcher' })];
+      const { service, repo } = makeImportService({ existing });
+      const res = await service.importFromCatalog(
+        'ws-1',
+        'u1',
+        dto({ conflict: 'skip' }),
+      );
+      expect(res).toMatchObject({ created: 0, skipped: 1, renamed: 0 });
+      expect(repo.insert).not.toHaveBeenCalled();
+    });
+
+    it('name collision + conflict:rename => inserts under " (2)"', async () => {
+      const existing = [makeRow({ id: 'r-x', name: 'Researcher' })];
+      const { service, repo } = makeImportService({ existing });
+      const res = await service.importFromCatalog(
+        'ws-1',
+        'u1',
+        dto({ conflict: 'rename' }),
+      );
+      expect(res).toMatchObject({ created: 1, skipped: 0, renamed: 1 });
+      expect(repo.insert.mock.calls[0][0].name).toBe('Researcher (2)');
+    });
+
+    it('dto.slugs filters; an unknown slug becomes an error entry', async () => {
+      const { service, repo } = makeImportService({
+        bundleRoles: [catalogRole()],
+      });
+      const res = await service.importFromCatalog(
+        'ws-1',
+        'u1',
+        dto({ slugs: ['researcher', 'ghost'] }),
+      );
+      expect(res.created).toBe(1);
+      expect(res.errors).toEqual([
+        { slug: 'ghost', message: 'Role not found in catalog bundle' },
+      ]);
+      expect(repo.insert).toHaveBeenCalledTimes(1);
+    });
+
+    it('insert unique-violation (23505) is recorded as an error, import continues', async () => {
+      const { service, repo } = makeImportService({
+        bundleRoles: [
+          catalogRole({ slug: 'a', name: 'A' }),
+          catalogRole({ slug: 'b', name: 'B' }),
+        ],
+        indexRoles: [
+          { slug: 'a', version: 1 },
+          { slug: 'b', version: 1 },
+        ],
+      });
+      repo.insert
+        .mockRejectedValueOnce({ code: '23505' })
+        .mockImplementationOnce((v) => Promise.resolve(makeRow(v)));
+      const res = await service.importFromCatalog('ws-1', 'u1', dto());
+      expect(res.created).toBe(1);
+      expect(res.errors).toEqual([
+        { slug: 'a', message: 'A role with this name already exists' },
+      ]);
+    });
+
+    it('source-uniqueness 23505 (concurrent import of same slug+language) => skipped, NOT an error, batch continues', async () => {
+      // Two parallel imports of the same bundle each build installedKeys from a
+      // stale snapshot, so both reach the insert for slug 'a'. The DB partial
+      // unique index on (workspace, source->>slug, source->>language) rejects the
+      // loser with a 23505 carrying the source-index constraint name. That must
+      // be treated as "already installed" (skip), not a per-role error, and the
+      // rest of the batch (slug 'b') must still import.
+      const { service, repo } = makeImportService({
+        bundleRoles: [
+          catalogRole({ slug: 'a', name: 'A' }),
+          catalogRole({ slug: 'b', name: 'B' }),
+        ],
+        indexRoles: [
+          { slug: 'a', version: 1 },
+          { slug: 'b', version: 1 },
+        ],
+      });
+      // The kysely-postgres-js driver surfaces the violated constraint on
+      // `constraint_name` (not node-postgres' `.constraint`), matching prod.
+      const sourceRace = Object.assign(new Error('duplicate key'), {
+        code: '23505',
+        constraint_name: 'ai_agent_roles_workspace_source_unique',
+      });
+      repo.insert
+        .mockRejectedValueOnce(sourceRace)
+        .mockImplementationOnce((v) => Promise.resolve(makeRow(v)));
+      const res = await service.importFromCatalog('ws-1', 'u1', dto());
+      // 'a' converged on the concurrent install (skip); 'b' imported; no errors.
+      expect(res).toMatchObject({ created: 1, skipped: 1, renamed: 0 });
+      expect(res.errors).toEqual([]);
+      // Both inserts were attempted (the batch did not abort on the 23505).
+      expect(repo.insert).toHaveBeenCalledTimes(2);
+    });
+
+    it('non-unique insert error => generic message, root cause logged, import continues', async () => {
+      const logSpy = jest
+        .spyOn(Logger.prototype, 'error')
+        .mockImplementation(() => undefined);
+      try {
+        const { service, repo } = makeImportService({
+          bundleRoles: [
+            catalogRole({ slug: 'a', name: 'A' }),
+            catalogRole({ slug: 'b', name: 'B' }),
+          ],
+          indexRoles: [
+            { slug: 'a', version: 1 },
+            { slug: 'b', version: 1 },
+          ],
+        });
+        // A non-23505 failure (e.g. a not-null violation) on the first insert.
+        const boom = Object.assign(new Error('null value in column'), {
+          code: '23502',
+        });
+        repo.insert
+          .mockRejectedValueOnce(boom)
+          .mockImplementationOnce((v) => Promise.resolve(makeRow(v)));
+        const res = await service.importFromCatalog('ws-1', 'u1', dto());
+        // The generic (non-409) user-facing message; the second role still imports.
+        expect(res.created).toBe(1);
+        expect(res.errors).toEqual([
+          { slug: 'a', message: 'Failed to import role' },
+        ]);
+        // The root cause was logged with the slug for diagnosis.
+        expect(logSpy).toHaveBeenCalledTimes(1);
+        expect(String(logSpy.mock.calls[0][0])).toContain('slug=a');
+      } finally {
+        logSpy.mockRestore();
+      }
+    });
+
+    it('bundleId absent from the index => BadGateway (no insert)', async () => {
+      // The requested bundle is not listed in the fetched index (a stale client
+      // or an index/bundle drift); the import must surface a 502 rather than
+      // silently doing nothing or dereferencing a missing meta.
+      const { service, repo } = makeImportService({});
+      await expect(
+        service.importFromCatalog('ws-1', 'u1', dto({ bundleId: 'missing' })),
+      ).rejects.toBeInstanceOf(BadGatewayException);
+      expect(repo.insert).not.toHaveBeenCalled();
+    });
+  });
+
+  describe('updateFromCatalog', () => {
+    function makeUpdateService(opts: {
+      role?: AiAgentRole;
+      indexBundles?: unknown[];
+      bundleRoles?: Record<string, unknown>[];
+      others?: AiAgentRole[];
+    }) {
+      const index = {
+        schemaVersion: 1,
+        bundles: opts.indexBundles ?? [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: [{ slug: 'researcher', version: 5 }],
+          },
+        ],
+      };
+      const bundle = {
+        schemaVersion: 1,
+        language: 'en',
+        roles: opts.bundleRoles ?? [
+          { slug: 'researcher', name: 'Researcher v5', instructions: 'new' },
+        ],
+      };
+      const repo = {
+        findById: jest.fn().mockResolvedValue(opts.role),
+        insert: jest.fn(),
+        update: jest.fn().mockResolvedValue(undefined),
+        softDelete: jest.fn(),
+        listByWorkspace: jest.fn().mockResolvedValue(opts.others ?? []),
+      };
+      const catalog = {
+        fetchIndex: jest.fn().mockResolvedValue(index),
+        fetchBundle: jest.fn().mockResolvedValue(bundle),
+      };
+      const service = new AiAgentRolesService(repo as never, catalog as never);
+      return { service, repo, catalog };
+    }
+
+    const imported = (version: number, over: Partial<AiAgentRole> = {}) =>
+      makeRow({
+        id: 'r1',
+        name: 'Researcher',
+        source: { slug: 'researcher', language: 'en', version } as never,
+        ...over,
+      });
+
+    it('role not imported from catalog (source null) => BadRequest', async () => {
+      const { service } = makeUpdateService({ role: makeRow({ source: null }) });
+      await expect(
+        service.updateFromCatalog('ws-1', { id: 'r1' } as never),
+      ).rejects.toBeInstanceOf(BadRequestException);
+    });
+
+    it('role not found => BadRequest', async () => {
+      const { service } = makeUpdateService({ role: undefined });
+      await expect(
+        service.updateFromCatalog('ws-1', { id: 'r1' } as never),
+      ).rejects.toBeInstanceOf(BadRequestException);
+    });
+
+    it('catalog version <= source.version => up-to-date (no update)', async () => {
+      const { service, repo } = makeUpdateService({ role: imported(5) });
+      const res = await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      expect(res).toEqual({ updated: false, reason: 'up-to-date' });
+      expect(repo.update).not.toHaveBeenCalled();
+    });
+
+    it('slug no longer listed in any bundle => not-in-catalog', async () => {
+      const { service, repo } = makeUpdateService({
+        role: imported(1),
+        indexBundles: [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: [{ slug: 'other', version: 9 }],
+          },
+        ],
+      });
+      const res = await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      expect(res).toEqual({ updated: false, reason: 'not-in-catalog' });
+      expect(repo.update).not.toHaveBeenCalled();
+    });
+
+    it('source.language no longer offered by the bundle => language-unavailable', async () => {
+      const { service, repo } = makeUpdateService({
+        role: imported(1, {
+          source: { slug: 'researcher', language: 'ru', version: 1 } as never,
+        }),
+        indexBundles: [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: [{ slug: 'researcher', version: 5 }],
+          },
+        ],
+      });
+      const res = await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      expect(res).toEqual({ updated: false, reason: 'language-unavailable' });
+      expect(repo.update).not.toHaveBeenCalled();
+    });
+
+    it('newer version => updates content + bumps source.version, returns versions', async () => {
+      const role = imported(1);
+      const { service, repo } = makeUpdateService({ role });
+      // The post-update re-fetch returns the bumped row.
+      repo.findById
+        .mockResolvedValueOnce(role)
+        .mockResolvedValueOnce(
+          imported(5, { name: 'Researcher v5', instructions: 'new' }),
+        );
+      const res = await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      expect(res).toMatchObject({
+        updated: true,
+        fromVersion: 1,
+        toVersion: 5,
+      });
+      const patch = repo.update.mock.calls[0][2];
+      expect(patch.source).toEqual({
+        slug: 'researcher',
+        language: 'en',
+        version: 5,
+      });
+      expect(patch.name).toBe('Researcher v5');
+      // enabled is never touched by an update-from-catalog.
+      expect('enabled' in patch).toBe(false);
+    });
+
+    it('slug listed in the index but missing from the bundle file => not-in-catalog', async () => {
+      // Index/bundle drift: the index still advertises a newer `researcher`
+      // (v5 > installed v1) in an offered language, but the fetched bundle file
+      // no longer contains that slug. The update must no-op as not-in-catalog,
+      // not throw or write a half-resolved role.
+      const { service, repo } = makeUpdateService({
+        role: imported(1),
+        bundleRoles: [
+          { slug: 'someone-else', name: 'Other', instructions: 'x' },
+        ],
+      });
+      const res = await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      expect(res).toEqual({ updated: false, reason: 'not-in-catalog' });
+      expect(repo.update).not.toHaveBeenCalled();
+    });
+
+    it('new catalog name collides with another live role => keeps current name', async () => {
+      const role = imported(1);
+      const other = makeRow({ id: 'r2', name: 'Researcher v5' });
+      const { service, repo } = makeUpdateService({ role, others: [role, other] });
+      repo.findById
+        .mockResolvedValueOnce(role)
+        .mockResolvedValueOnce(imported(5));
+      await service.updateFromCatalog('ws-1', { id: 'r1' } as never);
+      // The colliding catalog name is dropped; the current name is kept.
+      expect(repo.update.mock.calls[0][2].name).toBe('Researcher');
+    });
+  });
+
+  // ---------------------------------------------------------------------------
+  // Catalog browse (getCatalog / getCatalogBundle) against a MOCKED provider.
+  // Covers the localized() three-tier fallback (requested lang -> en -> first ->
+  // null), the sorted union of bundle languages, the missing-bundle BadGateway,
+  // and the role-version default.
+  // ---------------------------------------------------------------------------
+  describe('getCatalog', () => {
+    function makeBrowseService(index: unknown) {
+      const repo = {
+        findById: jest.fn(),
+        insert: jest.fn(),
+        update: jest.fn(),
+        softDelete: jest.fn(),
+        listByWorkspace: jest.fn(),
+      };
+      const catalog = {
+        fetchIndex: jest.fn().mockResolvedValue(index),
+        fetchBundle: jest.fn(),
+      };
+      const service = new AiAgentRolesService(repo as never, catalog as never);
+      return { service, catalog };
+    }
+
+    it('returns the sorted union of every bundle language', async () => {
+      const { service } = makeBrowseService({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'a',
+            name: { en: 'A' },
+            languages: ['ru', 'en'],
+            roles: [],
+          },
+          {
+            id: 'b',
+            name: { en: 'B' },
+            languages: ['en', 'de'],
+            roles: [],
+          },
+        ],
+      });
+      const res = await service.getCatalog('en');
+      expect(res.languages).toEqual(['de', 'en', 'ru']);
+    });
+
+    it('localized name uses the requested language when present', async () => {
+      const { service } = makeBrowseService({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'a',
+            name: { en: 'General', ru: 'Общие' },
+            description: { en: 'desc-en', ru: 'desc-ru' },
+            languages: ['en', 'ru'],
+            roles: [{ slug: 'researcher', version: 2 }],
+          },
+        ],
+      });
+      const res = await service.getCatalog('ru');
+      expect(res.bundles[0]).toMatchObject({
+        id: 'a',
+        name: 'Общие',
+        description: 'desc-ru',
+        languages: ['en', 'ru'],
+        roles: [{ slug: 'researcher', version: 2 }],
+      });
+    });
+
+    it('localized name falls back to en when the requested language is missing', async () => {
+      const { service } = makeBrowseService({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'a',
+            name: { en: 'General', ru: 'Общие' },
+            languages: ['en', 'ru'],
+            roles: [],
+          },
+        ],
+      });
+      const res = await service.getCatalog('fr');
+      expect(res.bundles[0].name).toBe('General');
+    });
+
+    it('localized name falls back to the first available locale when en is absent', async () => {
+      const { service } = makeBrowseService({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'a',
+            name: { ru: 'Общие', de: 'Allgemein' },
+            languages: ['ru', 'de'],
+            roles: [],
+          },
+        ],
+      });
+      const res = await service.getCatalog('fr');
+      // Neither 'fr' nor 'en' is present -> first available value.
+      expect(res.bundles[0].name).toBe('Общие');
+    });
+
+    it('empty name map => falls back to the bundle id; absent description => null', async () => {
+      const { service } = makeBrowseService({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'a',
+            name: {},
+            languages: ['en'],
+            roles: [],
+          },
+        ],
+      });
+      const res = await service.getCatalog('en');
+      expect(res.bundles[0].name).toBe('a');
+      expect(res.bundles[0].description).toBeNull();
+    });
+  });
+
+  describe('getCatalogBundle', () => {
+    function makeBundleService(opts: {
+      index: unknown;
+      bundle: unknown;
+    }) {
+      const repo = {
+        findById: jest.fn(),
+        insert: jest.fn(),
+        update: jest.fn(),
+        softDelete: jest.fn(),
+        listByWorkspace: jest.fn(),
+      };
+      const catalog = {
+        fetchIndex: jest.fn().mockResolvedValue(opts.index),
+        fetchBundle: jest.fn().mockResolvedValue(opts.bundle),
+      };
+      const service = new AiAgentRolesService(repo as never, catalog as never);
+      return { service, catalog };
+    }
+
+    const index = {
+      schemaVersion: 1,
+      bundles: [
+        {
+          id: 'general',
+          name: { en: 'General' },
+          languages: ['en'],
+          roles: [{ slug: 'researcher', version: 4 }],
+        },
+      ],
+    };
+
+    it('missing bundle in the index => BadGateway', async () => {
+      const { service, catalog } = makeBundleService({
+        index,
+        bundle: { schemaVersion: 1, language: 'en', roles: [] },
+      });
+      await expect(
+        service.getCatalogBundle('ghost', 'en'),
+      ).rejects.toBeInstanceOf(BadGatewayException);
+      expect(catalog.fetchBundle).not.toHaveBeenCalled();
+    });
+
+    it('maps role content with the version taken from the index', async () => {
+      const { service } = makeBundleService({
+        index,
+        bundle: {
+          schemaVersion: 1,
+          language: 'en',
+          roles: [
+            {
+              slug: 'researcher',
+              name: 'Researcher',
+              instructions: 'be a researcher',
+              emoji: '🔬',
+              autoStart: false,
+              launchMessage: 'go',
+            },
+          ],
+        },
+      });
+      const res = await service.getCatalogBundle('general', 'en');
+      expect(res).toMatchObject({ bundleId: 'general', language: 'en' });
+      expect(res.roles[0]).toEqual({
+        slug: 'researcher',
+        emoji: '🔬',
+        name: 'Researcher',
+        description: null,
+        instructions: 'be a researcher',
+        autoStart: false,
+        launchMessage: 'go',
+        version: 4,
+      });
+    });
+
+    it('role absent from the index meta => version defaults to 1; autoStart defaults to true', async () => {
+      const { service } = makeBundleService({
+        index,
+        bundle: {
+          schemaVersion: 1,
+          language: 'en',
+          roles: [
+            { slug: 'newcomer', name: 'Newcomer', instructions: 'hi' },
+          ],
+        },
+      });
+      const res = await service.getCatalogBundle('general', 'en');
+      expect(res.roles[0]).toMatchObject({
+        slug: 'newcomer',
+        version: 1,
+        autoStart: true,
+        emoji: null,
+        launchMessage: null,
+      });
+    });
+  });
 });

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

@@ -1,12 +1,24 @@
 import {
+  BadGatewayException,
   BadRequestException,
   ConflictException,
   Injectable,
+  Logger,
 } from '@nestjs/common';
-import { AiAgentRoleRepo } from '@docmost/db/repos/ai-agent-roles/ai-agent-roles.repo';
-import { AiAgentRole } from '@docmost/db/types/entity.types';
+import {
+  AiAgentRoleRepo,
+  parseSource,
+} from '@docmost/db/repos/ai-agent-roles/ai-agent-roles.repo';
+import { AiAgentRole, RoleSource } from '@docmost/db/types/entity.types';
 import { CreateAgentRoleDto, UpdateAgentRoleDto } from './dto/agent-role.dto';
+import { ImportFromCatalogDto, UpdateFromCatalogDto } from './dto/agent-role-catalog.dto';
 import { RoleModelConfig } from './role-model-config';
+import { AiAgentRolesCatalogProvider } from './catalog/ai-agent-roles-catalog.provider';
+import {
+  CatalogBundleFile,
+  CatalogBundleMeta,
+  CatalogRole,
+} from './catalog/catalog-types';
 
 /**
  * Full (admin) view of an agent role. There are no secret columns on this table
@@ -24,6 +36,10 @@ export interface AgentRoleView {
   enabled: boolean;
   autoStart: boolean;
   launchMessage: string | null;
+  // Catalog origin of an imported role, or null for a manually-created one. The
+  // admin UI uses `version` to offer an UPDATE when the catalog ships a newer
+  // revision. Admin-only (deliberately absent from AgentRolePickerView).
+  source: RoleSource | null;
   createdAt: Date;
   updatedAt: Date;
 }
@@ -56,7 +72,12 @@ export interface AgentRolePickerView {
  */
 @Injectable()
 export class AiAgentRolesService {
-  constructor(private readonly repo: AiAgentRoleRepo) {}
+  private readonly logger = new Logger(AiAgentRolesService.name);
+
+  constructor(
+    private readonly repo: AiAgentRoleRepo,
+    private readonly catalog: AiAgentRolesCatalogProvider,
+  ) {}
 
   /**
    * List the workspace's roles. Admins get the full view (the settings page needs
@@ -165,6 +186,316 @@ export class AiAgentRolesService {
     return { success: true };
   }
 
+  // -------------------------------------------------------------------------
+  // Catalog (admin-only). The catalog is curated, untrusted JSON fetched +
+  // validated by AiAgentRolesCatalogProvider; this layer resolves localized
+  // text and reconciles a bundle against the workspace's existing roles.
+  // -------------------------------------------------------------------------
+
+  /**
+   * Browse the catalog. Returns the union of every bundle's languages (sorted)
+   * plus per-bundle metadata with `name` / `description` resolved to the
+   * requested `language` (fallback: 'en', then the first available locale).
+   */
+  async getCatalog(language?: string): Promise<{
+    languages: string[];
+    bundles: {
+      id: string;
+      name: string;
+      description: string | null;
+      languages: string[];
+      roles: { slug: string; version: number }[];
+    }[];
+  }> {
+    const index = await this.catalog.fetchIndex();
+    const languages = Array.from(
+      new Set(index.bundles.flatMap((b) => b.languages)),
+    ).sort();
+    const bundles = index.bundles.map((b) => ({
+      id: b.id,
+      name: localized(b.name, language) ?? b.id,
+      description: b.description ? localized(b.description, language) : null,
+      languages: b.languages,
+      roles: b.roles.map((r) => ({ slug: r.slug, version: r.version })),
+    }));
+    return { languages, bundles };
+  }
+
+  /**
+   * Shared read prefix for the two bundle-by-id catalog paths (getCatalogBundle /
+   * importFromCatalog): fetch the index, resolve the requested bundle's meta
+   * (502 if the index does not list it), fetch its per-language file, and build
+   * the slug->version map from the meta. The callers keep their own response /
+   * write logic; only this duplicated read is factored out here.
+   */
+  private async loadBundleById(
+    bundleId: string,
+    language: string,
+  ): Promise<{
+    meta: CatalogBundleMeta;
+    file: CatalogBundleFile;
+    versions: Map<string, number>;
+  }> {
+    const index = await this.catalog.fetchIndex();
+    const meta = index.bundles.find((b) => b.id === bundleId);
+    if (!meta) {
+      throw new BadGatewayException('Catalog bundle not found');
+    }
+    const file = await this.catalog.fetchBundle(bundleId, language);
+    return { meta, file, versions: versionMap(meta) };
+  }
+
+  /**
+   * Open one bundle in a language: returns each role's content plus the version
+   * taken from the index (so the client can compare against an imported role's
+   * source.version). A missing bundle/language => BadGateway (catalog issue).
+   */
+  async getCatalogBundle(
+    bundleId: string,
+    language: string,
+  ): Promise<{
+    bundleId: string;
+    language: string;
+    roles: {
+      slug: string;
+      emoji: string | null;
+      name: string;
+      description: string | null;
+      instructions: string;
+      autoStart: boolean;
+      launchMessage: string | null;
+      version: number;
+    }[];
+  }> {
+    const { file, versions } = await this.loadBundleById(bundleId, language);
+    return {
+      bundleId,
+      language,
+      roles: file.roles.map((r) => ({
+        slug: r.slug,
+        emoji: r.emoji ?? null,
+        name: r.name,
+        description: r.description ?? null,
+        instructions: r.instructions,
+        autoStart: r.autoStart ?? true,
+        launchMessage: r.launchMessage ?? null,
+        version: versions.get(r.slug) ?? 1,
+      })),
+    };
+  }
+
+  /**
+   * Import a bundle's roles into the workspace. A role is "already installed"
+   * (and thus skipped — updates are a separate action) only when an existing
+   * role matches BOTH its `source.slug` AND `source.language`: this is a
+   * multilingual catalog, so a different language of the same slug (e.g. the
+   * `ru` variant of a slug already installed as `en`) is a SEPARATE install and
+   * still imports. A name collision with an existing role is either skipped or
+   * imported under a free " (N)" name, per `dto.conflict`. Inserts run
+   * sequentially (the repo exposes no batch insert and the volume is tiny); a
+   * unique-name race still surfaces as an error entry rather than aborting the
+   * whole import.
+   */
+  async importFromCatalog(
+    workspaceId: string,
+    creatorId: string,
+    dto: ImportFromCatalogDto,
+  ): Promise<{
+    created: number;
+    skipped: number;
+    renamed: number;
+    errors: { slug: string; message: string }[];
+  }> {
+    const { file, versions } = await this.loadBundleById(
+      dto.bundleId,
+      dto.language,
+    );
+
+    const errors: { slug: string; message: string }[] = [];
+
+    // Resolve the selected catalog roles (honor dto.slugs; flag unknown ones).
+    let selected = file.roles;
+    if (dto.slugs && dto.slugs.length > 0) {
+      const wanted = new Set(dto.slugs);
+      const present = new Set(file.roles.map((r) => r.slug));
+      for (const slug of dto.slugs) {
+        if (!present.has(slug)) {
+          errors.push({ slug, message: 'Role not found in catalog bundle' });
+        }
+      }
+      selected = file.roles.filter((r) => wanted.has(r.slug));
+    }
+
+    const existingRoles = await this.repo.listByWorkspace(workspaceId);
+    // Catalog roles already installed in this workspace, keyed by slug+language
+    // (skip; never duplicate). The key MUST match the client install-state and
+    // updateFromCatalog (both match by source.slug AND source.language): the
+    // `ru` variant of a slug already installed as `en` is a separate install.
+    const installedKeys = new Set(
+      existingRoles
+        .map((r) => parseSource(r.source))
+        .filter((s): s is RoleSource => s !== null)
+        .map((s) => `${s.slug}:${s.language}`),
+    );
+    // Live role names (lowercased) for collision detection. Mutated as we
+    // insert so two imported roles cannot both grab the same name.
+    const takenNames = new Set(
+      existingRoles.map((r) => r.name.trim().toLowerCase()),
+    );
+
+    let created = 0;
+    let skipped = 0;
+    let renamed = 0;
+
+    for (const role of selected) {
+      // Already installed from the catalog in THIS language => skip (use
+      // update-from-catalog). A different language of the same slug still imports.
+      const installKey = `${role.slug}:${dto.language}`;
+      if (installedKeys.has(installKey)) {
+        skipped++;
+        continue;
+      }
+
+      let name = role.name.trim();
+      let didRename = false;
+      if (takenNames.has(name.toLowerCase())) {
+        if (dto.conflict === 'skip') {
+          skipped++;
+          continue;
+        }
+        // conflict === 'rename': find a free " (N)" suffix.
+        name = freeName(name, takenNames);
+        didRename = true;
+      }
+
+      const version = versions.get(role.slug) ?? 1;
+      try {
+        await this.repo.insert({
+          workspaceId,
+          creatorId,
+          name,
+          ...catalogRoleContentFields(role),
+          enabled: true,
+          source: { slug: role.slug, language: dto.language, version },
+        });
+        created++;
+        if (didRename) renamed++;
+        takenNames.add(name.toLowerCase());
+        installedKeys.add(installKey);
+      } catch (err) {
+        // A 23505 from the source-uniqueness index means a CONCURRENT import
+        // already installed this exact slug+language between our snapshot
+        // (installedKeys) and this insert: the in-process snapshot cannot see a
+        // sibling request's writes, so the partial unique index is the backstop.
+        // Outcome is identical to the snapshot-based skip above — count it as
+        // skipped (already installed) and continue; do NOT abort or error.
+        if (isSourceUniqueViolation(err)) {
+          skipped++;
+          installedKeys.add(installKey);
+          continue;
+        }
+        // Otherwise: a unique-NAME race (23505 on the name index) is expected and
+        // self-explanatory (it becomes a friendly per-role error). Any OTHER
+        // insert failure is unexpected, so log the root cause with enough context
+        // to diagnose it — the user-facing message is deliberately generic.
+        if (!isUniqueViolation(err)) {
+          this.logger.error(
+            `Failed to import catalog role (workspaceId=${workspaceId} bundleId=${dto.bundleId} slug=${role.slug}): ${err instanceof Error ? err.stack ?? err.message : String(err)}`,
+          );
+        }
+        errors.push({ slug: role.slug, message: importErrorMessage(err) });
+      }
+    }
+
+    return { created, skipped, renamed, errors };
+  }
+
+  /**
+   * Update an already-imported role from its catalog source when the catalog
+   * ships a newer version. Returns a discriminated result so the UI can explain
+   * a no-op (up-to-date / removed from catalog / language no longer offered).
+   * Never touches `enabled`; keeps the current name if the catalog's new name
+   * would collide with another role (avoiding the unique-name 409).
+   */
+  async updateFromCatalog(
+    workspaceId: string,
+    dto: UpdateFromCatalogDto,
+  ): Promise<
+    | { updated: false; reason: 'not-in-catalog' | 'up-to-date' | 'language-unavailable' }
+    | { updated: true; fromVersion: number; toVersion: number; role: AgentRoleView }
+  > {
+    const role = await this.repo.findById(dto.id, workspaceId);
+    if (!role) throw new BadRequestException('Role not found');
+
+    const source = parseSource(role.source);
+    if (!source || !source.slug) {
+      throw new BadRequestException('Role was not imported from the catalog');
+    }
+
+    const index = await this.catalog.fetchIndex();
+    // Find the bundle whose meta lists this slug, and its catalog version.
+    let meta: CatalogBundleMeta | undefined;
+    let currentVersion: number | undefined;
+    for (const b of index.bundles) {
+      const m = b.roles.find((r) => r.slug === source.slug);
+      if (m) {
+        meta = b;
+        currentVersion = m.version;
+        break;
+      }
+    }
+    if (!meta || currentVersion === undefined) {
+      return { updated: false, reason: 'not-in-catalog' };
+    }
+    if (currentVersion <= source.version) {
+      return { updated: false, reason: 'up-to-date' };
+    }
+    if (!meta.languages.includes(source.language)) {
+      return { updated: false, reason: 'language-unavailable' };
+    }
+
+    const file = await this.catalog.fetchBundle(meta.id, source.language);
+    const fresh = file.roles.find((r) => r.slug === source.slug);
+    if (!fresh) {
+      return { updated: false, reason: 'not-in-catalog' };
+    }
+
+    // Keep the current name when the catalog's new name would collide with
+    // another live role (avoids the unique-name 409). Same-name (case-insensitive)
+    // means "no rename needed".
+    const newName = fresh.name.trim();
+    let name = newName;
+    if (newName.toLowerCase() !== role.name.trim().toLowerCase()) {
+      const others = await this.repo.listByWorkspace(workspaceId);
+      const collision = others.some(
+        (r) =>
+          r.id !== role.id &&
+          r.name.trim().toLowerCase() === newName.toLowerCase(),
+      );
+      if (collision) name = role.name;
+    }
+
+    await this.repo.update(dto.id, workspaceId, {
+      name,
+      ...catalogRoleContentFields(fresh),
+      // enabled is deliberately NOT changed.
+      source: {
+        slug: source.slug,
+        language: source.language,
+        version: currentVersion,
+      },
+    });
+
+    const updated = await this.repo.findById(dto.id, workspaceId);
+    if (!updated) throw new BadRequestException('Role not found');
+    return {
+      updated: true,
+      fromVersion: source.version,
+      toVersion: currentVersion,
+      role: this.toView(updated),
+    };
+  }
+
   private toView(row: AiAgentRole): AgentRoleView {
     return {
       id: row.id,
@@ -176,6 +507,9 @@ export class AiAgentRolesService {
       enabled: row.enabled,
       autoStart: row.autoStart,
       launchMessage: row.launchMessage ?? null,
+      // parseSource yields a fully-valid RoleSource | null (the row is already
+      // normalized; this also keeps the field type honest without a cast).
+      source: parseSource(row.source),
       createdAt: row.createdAt,
       updatedAt: row.updatedAt,
     };
@@ -205,11 +539,7 @@ export class AiAgentRolesService {
  * failures keep surfacing as 500s.
  */
 function rethrowDuplicateName(err: unknown, name: string): never {
-  if (
-    err &&
-    typeof err === 'object' &&
-    (err as { code?: unknown }).code === '23505'
-  ) {
+  if (isUniqueViolation(err)) {
     throw new ConflictException(
       `A role named "${name}" already exists in this workspace.`,
     );
@@ -217,13 +547,120 @@ function rethrowDuplicateName(err: unknown, name: string): never {
   throw err;
 }
 
-/** '' / whitespace-only / undefined => null; otherwise the trimmed value. */
-function emptyToNull(value: string | undefined): string | null {
-  if (value === undefined) return null;
+/** Whether `err` is a Postgres unique-violation (SQLSTATE 23505). */
+function isUniqueViolation(err: unknown): boolean {
+  return (
+    !!err &&
+    typeof err === 'object' &&
+    (err as { code?: unknown }).code === '23505'
+  );
+}
+
+/**
+ * The partial unique index name from the
+ * 20260626T160000-ai-agent-roles-catalog-source-unique migration: unique on
+ * (workspace_id, source->>'slug', source->>'language') for catalog-imported,
+ * non-deleted rows. A 23505 carrying this constraint name is a source-collision
+ * (concurrent import of the same slug+language), distinct from a name-collision.
+ */
+const SOURCE_UNIQUE_CONSTRAINT = 'ai_agent_roles_workspace_source_unique';
+
+/**
+ * Whether `err` is the 23505 raised by the SOURCE-uniqueness index specifically
+ * (vs the name-uniqueness index). The active driver (`kysely-postgres-js` over
+ * `postgres@3.4.8`) exposes the violated constraint name on `constraint_name`,
+ * so we key off that (accepting the node-postgres-style `.constraint` as a
+ * fallback for other drivers) — that way a source race is skipped while a name
+ * race still surfaces as a friendly per-role error. A 23505 with no constraint
+ * name (e.g. a wrapped/test error) is NOT treated as a source collision,
+ * preserving the existing name-race behavior.
+ */
+function isSourceUniqueViolation(err: unknown): boolean {
+  if (!isUniqueViolation(err)) return false;
+  const e = err as { constraint_name?: unknown; constraint?: unknown };
+  return (
+    e.constraint_name === SOURCE_UNIQUE_CONSTRAINT ||
+    e.constraint === SOURCE_UNIQUE_CONSTRAINT
+  );
+}
+
+/**
+ * The role-content fields shared by import (insert) and update (patch) of a
+ * catalog role: emoji/description/launchMessage normalized to null, model config
+ * normalized, autoStart defaulted. The caller adds the write-specific fields
+ * (`name`, `source`, and on insert `workspaceId`/`creatorId`/`enabled`).
+ */
+function catalogRoleContentFields(role: CatalogRole): {
+  emoji: string | null;
+  description: string | null;
+  instructions: string;
+  modelConfig: Record<string, unknown> | null;
+  autoStart: boolean;
+  launchMessage: string | null;
+} {
+  return {
+    emoji: emptyToNull(role.emoji),
+    description: emptyToNull(role.description),
+    instructions: role.instructions,
+    modelConfig: normalizeModelConfig(role.modelConfig) as
+      | Record<string, unknown>
+      | null,
+    autoStart: role.autoStart ?? true,
+    launchMessage: emptyToNull(role.launchMessage ?? undefined),
+  };
+}
+
+/** '' / whitespace-only / undefined / null => null; otherwise the trimmed value. */
+function emptyToNull(value: string | null | undefined): string | null {
+  if (value === undefined || value === null) return null;
   const trimmed = value.trim();
   return trimmed.length > 0 ? trimmed : null;
 }
 
+/** slug -> version map from a bundle's index metadata. */
+function versionMap(meta: CatalogBundleMeta): Map<string, number> {
+  return new Map(meta.roles.map((r) => [r.slug, r.version]));
+}
+
+/**
+ * Resolve a localized value `{ en, ru, ... }` to `language`, falling back to
+ * 'en', then the first available locale. Returns null only for an empty map.
+ */
+function localized(
+  map: Record<string, string>,
+  language?: string,
+): string | null {
+  if (language && typeof map[language] === 'string') return map[language];
+  if (typeof map.en === 'string') return map.en;
+  const first = Object.values(map)[0];
+  return typeof first === 'string' ? first : null;
+}
+
+/**
+ * Find a free display name by appending " (2)", " (3)", ... when `base` is
+ * already taken (case-insensitive against `taken`). Caller adds the result to
+ * `taken` after a successful insert.
+ */
+function freeName(base: string, taken: Set<string>): string {
+  // `taken` is finite, so within `taken.size + 2` iterations a candidate index
+  // is guaranteed free; the 1000 cap is a defensive upper bound far above any
+  // realistic per-name collision count. The throw below is therefore
+  // unreachable in practice and only satisfies the return-type checker.
+  for (let n = 2; n < 1000; n++) {
+    const candidate = `${base} (${n})`;
+    if (!taken.has(candidate.toLowerCase())) return candidate;
+  }
+  throw new BadRequestException(`Too many roles named "${base}"`);
+}
+
+/** A short, safe message for an import insert failure (409 vs other). */
+function importErrorMessage(err: unknown): string {
+  if (isUniqueViolation(err)) {
+    return 'A role with this name already exists';
+  }
+  return 'Failed to import role';
+}
+
 /**
  * Normalize an incoming modelConfig DTO to the persisted shape, or null when
  * there is no usable override (no driver and no chatModel). The DTO's @IsIn

apps/server/src/core/ai-chat/roles/catalog/ai-agent-roles-catalog.provider.spec.ts

@@ -0,0 +1,307 @@
+import { BadGatewayException, BadRequestException } from '@nestjs/common';
+import { AiAgentRolesCatalogProvider } from './ai-agent-roles-catalog.provider';
+
+/**
+ * Provider tests against a mocked remote source (no network). They cover the
+ * happy read path (fetchIndex / fetchBundle), the malformed-shape rejection,
+ * rejection of non-http(s) sources (local sources are gone), and — most
+ * importantly — the `^[a-z0-9-]+$` path-traversal guard that runs BEFORE any
+ * path/URL is built.
+ */
+describe('AiAgentRolesCatalogProvider', () => {
+  function makeProvider(source: string) {
+    const env = {
+      getAiAgentRolesCatalogSource: () => source,
+    };
+    return new AiAgentRolesCatalogProvider(env as never);
+  }
+
+  it('non-http(s) source => BadGateway (local sources removed)', async () => {
+    for (const source of ['', '/var/lib/agent-roles-catalog', './agent-roles-catalog']) {
+      const provider = makeProvider(source);
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    }
+  });
+
+  describe('remote fetch streaming size cap', () => {
+    const realFetch = global.fetch;
+    afterEach(() => {
+      global.fetch = realFetch;
+    });
+
+    /** A web ReadableStream that yields `chunks` (each a Uint8Array). */
+    function streamOf(chunks: Uint8Array[]): ReadableStream<Uint8Array> {
+      let i = 0;
+      return new ReadableStream<Uint8Array>({
+        pull(controller) {
+          if (i < chunks.length) controller.enqueue(chunks[i++]);
+          else controller.close();
+        },
+        // The provider cancels the reader on the too-large path; no-op here.
+        cancel() {},
+      });
+    }
+
+    /** A ReadableStream whose first read rejects (e.g. a mid-body AbortError). */
+    function errorStream(err: Error): ReadableStream<Uint8Array> {
+      return new ReadableStream<Uint8Array>({
+        pull() {
+          throw err;
+        },
+        cancel() {},
+      });
+    }
+
+    function mockResponse(opts: {
+      ok?: boolean;
+      status?: number;
+      headers?: Record<string, string>;
+      body: ReadableStream<Uint8Array> | null;
+      text?: string;
+    }): Response {
+      return {
+        ok: opts.ok ?? true,
+        status: opts.status ?? 200,
+        headers: { get: (k: string) => opts.headers?.[k.toLowerCase()] ?? null },
+        body: opts.body,
+        text: async () => opts.text ?? 'unused',
+      } as unknown as Response;
+    }
+
+    it('fetchBundle remote happy path => parses + validates', async () => {
+      const json = JSON.stringify({
+        schemaVersion: 1,
+        language: 'en',
+        roles: [
+          {
+            slug: 'researcher',
+            name: 'Researcher',
+            instructions: 'be a researcher',
+          },
+        ],
+      });
+      const body = streamOf([new TextEncoder().encode(json)]);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      const bundle = await provider.fetchBundle('general', 'en');
+      expect(bundle.roles[0].slug).toBe('researcher');
+    });
+
+    it('fetchBundle remote malformed (role missing instructions) => BadGateway', async () => {
+      const json = JSON.stringify({
+        schemaVersion: 1,
+        language: 'fr',
+        roles: [{ slug: 'researcher', name: 'Chercheur' }],
+      });
+      const body = streamOf([new TextEncoder().encode(json)]);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchBundle('general', 'fr')).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('declared Content-Length over the cap => BadGateway before reading the body', async () => {
+      global.fetch = jest.fn().mockResolvedValue(
+        mockResponse({
+          headers: { 'content-length': String(2_000_000) },
+          body: streamOf([new Uint8Array(10)]),
+        }),
+      ) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('streamed body exceeding the cap (no/under-reported Content-Length) => BadGateway', async () => {
+      // 1.5 MB streamed in 256 KB chunks, with no Content-Length header.
+      const chunks = Array.from(
+        { length: 6 },
+        () => new Uint8Array(256 * 1024),
+      );
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body: streamOf(chunks) })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('fetch rejects (network failure) => BadGateway (unavailable)', async () => {
+      global.fetch = jest
+        .fn()
+        .mockRejectedValue(new Error('ECONNREFUSED')) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('passes redirect:"error" to fetch (redirect-SSRF hardening)', async () => {
+      const fetchMock = jest
+        .fn()
+        .mockResolvedValue(
+          mockResponse({ body: streamOf([new Uint8Array(0)]) }),
+        );
+      global.fetch = fetchMock as never;
+      const provider = makeProvider('https://catalog.example.com');
+      // Body shape is irrelevant; an empty stream parses to invalid JSON and
+      // throws, but the fetch call (with its init) still happened.
+      await expect(provider.fetchIndex()).rejects.toBeDefined();
+      expect(fetchMock).toHaveBeenCalledWith(
+        expect.any(String),
+        expect.objectContaining({ redirect: 'error' }),
+      );
+    });
+
+    it('redirect response rejects (redirect:"error") => BadGateway', async () => {
+      // With redirect:"error", the platform fetch rejects on a 3xx instead of
+      // following it. Simulate that: the mock rejects when asked not to follow.
+      global.fetch = jest.fn().mockImplementation((_url, init) => {
+        if (init?.redirect === 'error') {
+          return Promise.reject(
+            new TypeError('fetch failed: unexpected redirect'),
+          );
+        }
+        return Promise.resolve(
+          mockResponse({ status: 302, body: null }),
+        );
+      }) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('non-ok response (503) => BadGateway carrying the status', async () => {
+      global.fetch = jest.fn().mockResolvedValue(
+        mockResponse({ ok: false, status: 503, body: null }),
+      ) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toThrow(/503/);
+    });
+
+    it('small streamed body parses normally (cap not hit)', async () => {
+      const json = JSON.stringify({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: [{ slug: 'researcher', version: 2 }],
+          },
+        ],
+      });
+      const body = streamOf([new TextEncoder().encode(json)]);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      const index = await provider.fetchIndex();
+      expect(index.bundles[0].id).toBe('general');
+    });
+
+    it('body read aborts mid-stream (AbortError) => BadGateway (not a generic 500)', async () => {
+      // The 10s timer aborts the whole request; on a slow/dripping source the
+      // body read (reader.read()) rejects with an AbortError AFTER fetch()
+      // resolved. The provider must map that to BadGateway, not let it escape.
+      const abortErr = Object.assign(new Error('The operation was aborted'), {
+        name: 'AbortError',
+      });
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body: errorStream(abortErr) })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('null body (no readable stream) => response.text() fallback parses', async () => {
+      const json = JSON.stringify({
+        schemaVersion: 1,
+        bundles: [
+          {
+            id: 'general',
+            name: { en: 'General' },
+            languages: ['en'],
+            roles: [{ slug: 'researcher', version: 2 }],
+          },
+        ],
+      });
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body: null, text: json })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      const index = await provider.fetchIndex();
+      expect(index.bundles[0].id).toBe('general');
+    });
+
+    it('null body + text() over the cap => BadGateway (too large)', async () => {
+      const oversized = 'a'.repeat(1_000_001);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(
+          mockResponse({ body: null, text: oversized }),
+        ) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('invalid JSON body => BadGateway (parse failure)', async () => {
+      const body = streamOf([new TextEncoder().encode('{not valid json')]);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toBeInstanceOf(
+        BadGatewayException,
+      );
+    });
+
+    it('malformed index.json (valid JSON, wrong shape) => BadGateway', async () => {
+      // Parses as JSON but fails isCatalogIndex (schemaVersion not a number).
+      const body = streamOf([
+        new TextEncoder().encode(
+          JSON.stringify({ schemaVersion: 'x', bundles: [] }),
+        ),
+      ]);
+      global.fetch = jest
+        .fn()
+        .mockResolvedValue(mockResponse({ body })) as never;
+      const provider = makeProvider('https://catalog.example.com');
+      await expect(provider.fetchIndex()).rejects.toThrow(/malformed/i);
+    });
+  });
+
+  describe('path-traversal / SSRF guard (^[a-z0-9-]+$)', () => {
+    const bad = ['../etc', 'a/b', 'A', 'foo.bar', 'foo_bar', '', '..'];
+
+    for (const value of bad) {
+      it(`rejects bundleId="${value}" with BadRequest`, async () => {
+        const provider = makeProvider('https://catalog.example.com');
+        await expect(
+          provider.fetchBundle(value, 'en'),
+        ).rejects.toBeInstanceOf(BadRequestException);
+      });
+
+      it(`rejects language="${value}" with BadRequest`, async () => {
+        const provider = makeProvider('https://catalog.example.com');
+        await expect(
+          provider.fetchBundle('general', value),
+        ).rejects.toBeInstanceOf(BadRequestException);
+      });
+    }
+  });
+});

apps/server/src/core/ai-chat/roles/catalog/ai-agent-roles-catalog.provider.ts

@@ -0,0 +1,311 @@
+import {
+  BadGatewayException,
+  BadRequestException,
+  Injectable,
+  Logger,
+} from '@nestjs/common';
+import { EnvironmentService } from '../../../../integrations/environment/environment.service';
+import {
+  CatalogBundleFile,
+  CatalogBundleMeta,
+  CatalogIndex,
+  CatalogRole,
+} from './catalog-types';
+
+/** Identifier shape allowed in any path/URL segment (bundleId, language). The
+ *  ONLY characters that can appear in a fetched path — the path-traversal and
+ *  SSRF guard. Anything else is rejected before a path/URL is built. */
+const SEGMENT_RE = /^[a-z0-9-]+$/;
+
+/** Remote fetch timeout and response-size cap. A curated catalog file is tiny;
+ *  the cap stops a hostile/misconfigured source from streaming unbounded data. */
+const FETCH_TIMEOUT_MS = 10_000;
+const MAX_BYTES = 1_000_000;
+
+/**
+ * Fetches + validates the agent-roles catalog from its configured source. The
+ * source (EnvironmentService.getAiAgentRolesCatalogSource()) is an http(s)://
+ * base URL — REMOTE only; local-filesystem sources are no longer supported. The
+ * value is baked into the Docker image at build time (set per-branch in CI).
+ *
+ * The catalog is UNTRUSTED input: every file is JSON-parsed and run through a
+ * hand-written type guard before any field is exposed, and every dynamic path
+ * segment is validated against SEGMENT_RE up front (path-traversal + SSRF).
+ */
+@Injectable()
+export class AiAgentRolesCatalogProvider {
+  private readonly logger = new Logger(AiAgentRolesCatalogProvider.name);
+
+  constructor(private readonly environmentService: EnvironmentService) {}
+
+  /** Read + validate the top-level index (`index.json`). */
+  async fetchIndex(): Promise<CatalogIndex> {
+    const raw = await this.readRelative('index.json');
+    const parsed = this.parseJson(raw, 'index.json');
+    if (!isCatalogIndex(parsed)) {
+      throw new BadGatewayException(
+        'Agent roles catalog index is malformed (index.json)',
+      );
+    }
+    return parsed;
+  }
+
+  /** Read + validate one language file (`bundles/<bundleId>/<language>.json`). */
+  async fetchBundle(
+    bundleId: string,
+    language: string,
+  ): Promise<CatalogBundleFile> {
+    // SECURITY: validate BEFORE building any path/URL (path-traversal + SSRF).
+    this.assertSegment(bundleId, 'bundleId');
+    this.assertSegment(language, 'language');
+    const rel = `bundles/${bundleId}/${language}.json`;
+    const raw = await this.readRelative(rel);
+    const parsed = this.parseJson(raw, rel);
+    if (!isCatalogBundleFile(parsed)) {
+      throw new BadGatewayException(
+        `Agent roles catalog bundle is malformed (${rel})`,
+      );
+    }
+    return parsed;
+  }
+
+  /** Reject a segment that is not a safe `[a-z0-9-]+` identifier. */
+  private assertSegment(value: string, field: string): void {
+    if (typeof value !== 'string' || !SEGMENT_RE.test(value)) {
+      throw new BadRequestException(`Invalid ${field}`);
+    }
+  }
+
+  /** JSON.parse with a clear BadGateway on malformed content. */
+  private parseJson(raw: string, rel: string): unknown {
+    try {
+      return JSON.parse(raw);
+    } catch (err) {
+      const reason = shortError(err);
+      this.logger.error(`Agent roles catalog JSON parse failed (${rel}): ${reason}`);
+      throw new BadGatewayException(
+        `Agent roles catalog file is not valid JSON (${rel}): ${reason}`,
+      );
+    }
+  }
+
+  /** Read a relative catalog path as text from the configured remote source. */
+  private async readRelative(rel: string): Promise<string> {
+    const source = this.environmentService
+      .getAiAgentRolesCatalogSource()
+      .trim();
+    if (!/^https?:\/\//i.test(source)) {
+      this.logger.error(
+        'Agent roles catalog source is not configured (expected an http(s):// base URL)',
+      );
+      throw new BadGatewayException(
+        'Agent roles catalog is unavailable: source is not configured',
+      );
+    }
+    return this.fetchRemote(source, rel);
+  }
+
+  /**
+   * Fetch a remote catalog file with a timeout + a STREAMING size cap. The body
+   * is never buffered in full before the check: we reject on a too-large
+   * Content-Length up front, then read the stream chunk-by-chunk and abort the
+   * moment the running total exceeds MAX_BYTES, so a hostile/misconfigured
+   * source cannot make us hold an unbounded body in memory.
+   */
+  private async fetchRemote(base: string, rel: string): Promise<string> {
+    const url = `${base.replace(/\/+$/, '')}/${rel}`;
+    const controller = new AbortController();
+    const timer = setTimeout(() => controller.abort(), FETCH_TIMEOUT_MS);
+    try {
+      let response: Response;
+      try {
+        // `redirect: 'error'` hardens against redirect-SSRF: a
+        // compromised-but-trusted upstream cannot 3xx the fetch into the
+        // internal network (e.g. http://169.254.169.254/...). A redirect
+        // response rejects here and is mapped to BadGateway below.
+        response = await fetch(url, {
+          signal: controller.signal,
+          redirect: 'error',
+        });
+      } catch (err) {
+        const reason = shortError(err);
+        this.logger.error(
+          `Agent roles catalog remote fetch failed (${rel}): ${reason}`,
+        );
+        throw new BadGatewayException(
+          `Agent roles catalog is unavailable: ${reason}`,
+        );
+      }
+      if (!response.ok) {
+        this.logger.error(
+          `Agent roles catalog remote returned ${response.status} (${rel})`,
+        );
+        throw new BadGatewayException(
+          `Agent roles catalog returned ${response.status}`,
+        );
+      }
+      // Reject a too-large declared size before reading any body bytes.
+      const declared = Number(response.headers.get('content-length'));
+      if (Number.isFinite(declared) && declared > MAX_BYTES) {
+        throw new BadGatewayException('Agent roles catalog file is too large');
+      }
+      // Bound the actual read: a missing/lying Content-Length is caught here.
+      // The 10s timer aborts the WHOLE request, so a slow/dripping hostile
+      // source rejects reader.read() (or response.text()) with an AbortError
+      // mid-body. Map that — and any other read failure — to a logged
+      // BadGateway so the admin endpoint returns 502 (not a generic 500). The
+      // cap's own BadGateway is rethrown as-is (no double-wrap).
+      try {
+        if (response.body) {
+          return await readStreamCapped(response.body, MAX_BYTES);
+        }
+        // Edge: no readable stream — fall back to a buffered read + length check.
+        const text = await response.text();
+        if (text.length > MAX_BYTES) {
+          throw new BadGatewayException('Agent roles catalog file is too large');
+        }
+        return text;
+      } catch (err) {
+        if (err instanceof BadGatewayException) throw err;
+        const reason = shortError(err);
+        this.logger.error(
+          `Agent roles catalog body read failed (${rel}): ${reason}`,
+        );
+        throw new BadGatewayException(
+          `Agent roles catalog is unavailable: ${reason}`,
+        );
+      }
+    } finally {
+      clearTimeout(timer);
+    }
+  }
+}
+
+/**
+ * Read a web ReadableStream into a UTF-8 string, throwing as soon as the
+ * accumulated byte count exceeds `maxBytes` (the reader is cancelled so the
+ * underlying connection is released). Never buffers more than the cap + the
+ * final chunk before bailing out.
+ */
+async function readStreamCapped(
+  body: ReadableStream<Uint8Array>,
+  maxBytes: number,
+): Promise<string> {
+  const reader = body.getReader();
+  const chunks: Uint8Array[] = [];
+  let total = 0;
+  try {
+    for (;;) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      if (!value) continue;
+      total += value.length;
+      if (total > maxBytes) {
+        throw new BadGatewayException('Agent roles catalog file is too large');
+      }
+      chunks.push(value);
+    }
+  } finally {
+    // Release the stream on both the normal and the too-large/abort paths.
+    await reader.cancel().catch(() => undefined);
+  }
+  return Buffer.concat(chunks).toString('utf8');
+}
+
+/**
+ * A short, non-sensitive error string for logging/propagation: only the first
+ * line of the message head is kept (upstream bodies / URLs are discarded).
+ */
+function shortError(err: unknown): string {
+  let message = '';
+  if (typeof err === 'string') {
+    message = err;
+  } else if (
+    err &&
+    typeof err === 'object' &&
+    typeof (err as { message?: unknown }).message === 'string'
+  ) {
+    // Read `.message` directly (works for Error instances and the realm-shifted
+    // Error-likes jest can hand back, where `instanceof Error` is false).
+    message = (err as { message: string }).message;
+  }
+  const head = (message || 'unknown error').split('\n')[0];
+  return head.length > 200 ? `${head.slice(0, 200)}…` : head;
+}
+
+// ---------------------------------------------------------------------------
+// Hand-written type guards (no zod / new deps). Each validates the exact wire
+// shape declared in catalog-types.ts; anything else is rejected by the caller.
+// ---------------------------------------------------------------------------
+
+function isObject(v: unknown): v is Record<string, unknown> {
+  return v !== null && typeof v === 'object' && !Array.isArray(v);
+}
+
+function isStringMap(v: unknown): v is Record<string, string> {
+  if (!isObject(v)) return false;
+  return Object.values(v).every((x) => typeof x === 'string');
+}
+
+function isStringArray(v: unknown): v is string[] {
+  return Array.isArray(v) && v.every((x) => typeof x === 'string');
+}
+
+export function isCatalogRole(v: unknown): v is CatalogRole {
+  if (!isObject(v)) return false;
+  if (typeof v.slug !== 'string') return false;
+  if (typeof v.name !== 'string') return false;
+  if (typeof v.instructions !== 'string') return false;
+  if (v.emoji !== undefined && typeof v.emoji !== 'string') return false;
+  if (v.description !== undefined && typeof v.description !== 'string') {
+    return false;
+  }
+  if (v.autoStart !== undefined && typeof v.autoStart !== 'boolean') {
+    return false;
+  }
+  if (
+    v.launchMessage !== undefined &&
+    v.launchMessage !== null &&
+    typeof v.launchMessage !== 'string'
+  ) {
+    return false;
+  }
+  if (
+    v.modelConfig !== undefined &&
+    v.modelConfig !== null &&
+    !isObject(v.modelConfig)
+  ) {
+    return false;
+  }
+  return true;
+}
+
+export function isCatalogBundleFile(v: unknown): v is CatalogBundleFile {
+  if (!isObject(v)) return false;
+  if (typeof v.schemaVersion !== 'number') return false;
+  if (typeof v.language !== 'string') return false;
+  if (!Array.isArray(v.roles)) return false;
+  return v.roles.every(isCatalogRole);
+}
+
+function isCatalogBundleMeta(v: unknown): v is CatalogBundleMeta {
+  if (!isObject(v)) return false;
+  if (typeof v.id !== 'string') return false;
+  if (!isStringMap(v.name)) return false;
+  if (v.description !== undefined && !isStringMap(v.description)) return false;
+  if (!isStringArray(v.languages)) return false;
+  if (!Array.isArray(v.roles)) return false;
+  return v.roles.every(
+    (r) =>
+      isObject(r) &&
+      typeof r.slug === 'string' &&
+      typeof r.version === 'number',
+  );
+}
+
+export function isCatalogIndex(v: unknown): v is CatalogIndex {
+  if (!isObject(v)) return false;
+  if (typeof v.schemaVersion !== 'number') return false;
+  if (!Array.isArray(v.bundles)) return false;
+  return v.bundles.every(isCatalogBundleMeta);
+}

apps/server/src/core/ai-chat/roles/catalog/catalog-types.ts

@@ -0,0 +1,47 @@
+/**
+ * Catalog wire shapes. The catalog is curated, untrusted JSON (a GitHub repo or
+ * a local folder), so every shape is validated by a hand-written type guard in
+ * the provider before any field is used — no zod / new deps on the server.
+ *
+ * Localized fields (`name` / `description` at the bundle level) are
+ * `Record<language, string>` so one bundle serves many UI languages; per-role
+ * `name` / `description` are already language-specific (the bundle file is keyed
+ * by language).
+ */
+
+/** One role's content as shipped in a per-language bundle file. */
+export interface CatalogRole {
+  slug: string;
+  emoji?: string;
+  name: string;
+  description?: string;
+  instructions: string;
+  autoStart?: boolean;
+  launchMessage?: string | null;
+  // Optional model override; same loose object shape as ai_agent_roles.model_config.
+  modelConfig?: Record<string, unknown> | null;
+}
+
+/** A single language file: `bundles/<id>/<language>.json`. */
+export interface CatalogBundleFile {
+  schemaVersion: number;
+  language: string;
+  roles: CatalogRole[];
+}
+
+/** Bundle metadata as listed in the top-level index. Versions live here (per
+ *  slug), so an UPDATE check needs only the index, not every language file. */
+export interface CatalogBundleMeta {
+  id: string;
+  // Localized display name/description: { en: '...', ru: '...' }.
+  name: Record<string, string>;
+  description?: Record<string, string>;
+  languages: string[];
+  roles: { slug: string; version: number }[];
+}
+
+/** Top-level catalog index: `index.json`. */
+export interface CatalogIndex {
+  schemaVersion: number;
+  bundles: CatalogBundleMeta[];
+}

apps/server/src/core/ai-chat/roles/dto/agent-role-catalog.dto.ts

@@ -0,0 +1,62 @@
+import {
+  IsArray,
+  IsIn,
+  IsOptional,
+  IsString,
+  IsUUID,
+  Matches,
+  MaxLength,
+} from 'class-validator';
+
+/** Safe identifier shape for any catalog path segment (bundleId / language).
+ *  Mirrors SEGMENT_RE in the catalog provider — the path-traversal/SSRF guard
+ *  is enforced both at the API boundary (here) and in the provider. */
+const SEGMENT_RE = /^[a-z0-9-]+$/;
+
+/** Browse the catalog, optionally localized to `language` (defaults applied in
+ *  the service: fall back to 'en', then the first available language). */
+export class CatalogQueryDto {
+  @IsOptional()
+  @IsString()
+  @MaxLength(16)
+  language?: string;
+}
+
+/** Open one catalog bundle in a specific language. */
+export class CatalogBundleDto {
+  @IsString()
+  @Matches(SEGMENT_RE)
+  bundleId: string;
+
+  @IsString()
+  @Matches(SEGMENT_RE)
+  language: string;
+}
+
+/** Import roles from a catalog bundle into the workspace. */
+export class ImportFromCatalogDto {
+  @IsString()
+  @Matches(SEGMENT_RE)
+  bundleId: string;
+
+  @IsString()
+  @Matches(SEGMENT_RE)
+  language: string;
+
+  // Omitted => import the whole bundle; otherwise only these slugs.
+  @IsOptional()
+  @IsArray()
+  @IsString({ each: true })
+  slugs?: string[];
+
+  // How to handle a name collision with an existing (non-catalog) role:
+  // 'skip' leaves it; 'rename' imports under a free " (N)" name.
+  @IsIn(['skip', 'rename'])
+  conflict: 'skip' | 'rename';
+}
+
+/** Update an already-imported role from its catalog source. */
+export class UpdateFromCatalogDto {
+  @IsUUID()
+  id: string;
+}

apps/server/src/core/page/services/page.service.footnote-canonicalize.spec.ts

@@ -0,0 +1,153 @@
+// Binding test for issue #228 must-fix #1 / test-coverage #12: footnote
+// canonicalization moved OUT of parseProsemirrorContent and is now applied only
+// on FULL-document writes (createPage, and updatePageContent with operation
+// 'replace'), NEVER on an append/prepend FRAGMENT.
+//
+// The Yjs encode / plain-text extract are stubbed (partial module mock keeps the
+// REAL canonicalizeFootnotes) and parseProsemirrorContent is spied to return the
+// raw fixture, so the test isolates the canonicalize BINDING from schema/Yjs.
+jest.mock('@docmost/editor-ext', () => {
+  const actual = jest.requireActual('@docmost/editor-ext');
+  return {
+    ...actual,
+    createYdocFromJson: jest.fn(() => Buffer.from([])),
+    jsonToText: jest.fn(() => ''),
+  };
+});
+
+import { PageService } from './page.service';
+
+const refNode = (id: string) => ({ type: 'footnoteReference', attrs: { id } });
+const defNode = (id: string, text: string) => ({
+  type: 'footnoteDefinition',
+  attrs: { id },
+  content: [{ type: 'paragraph', content: [{ type: 'text', text }] }],
+});
+const doc = (...content: any[]) => ({ type: 'doc', content });
+
+/** A full doc whose footnote definitions are OUT of reference order (b,a refs;
+ *  a,b defs) — canonicalization must reorder the definitions to [b, a]. */
+const outOfOrderFull = () =>
+  doc(
+    { type: 'paragraph', content: [{ type: 'text', text: 'x' }, refNode('b'), refNode('a')] },
+    { type: 'footnotesList', content: [defNode('a', 'A'), defNode('b', 'B')] },
+  );
+
+/** A definition-ONLY fragment (no references): canonicalizing it would drop the
+ *  whole footnotesList (referenceIds is empty) — i.e. LOSE the footnote. */
+const defOnlyFragment = () =>
+  doc({ type: 'footnotesList', content: [defNode('a', 'appended note')] });
+
+/** A reference-only fragment that REUSES an id defined elsewhere in the live
+ *  doc: canonicalizing it would synthesize a bogus empty footnotesList/def. */
+const refReuseFragment = () =>
+  doc({ type: 'paragraph', content: [{ type: 'text', text: 'more' }, refNode('a')] });
+
+function listDefIds(content: any): string[] {
+  const list = (content.content ?? []).find((n: any) => n.type === 'footnotesList');
+  return (list?.content ?? [])
+    .filter((n: any) => n.type === 'footnoteDefinition')
+    .map((n: any) => n.attrs?.id);
+}
+function hasFootnotesList(content: any): boolean {
+  return (content.content ?? []).some((n: any) => n.type === 'footnotesList');
+}
+
+describe('PageService footnote canonicalization binding (#228)', () => {
+  function makeService() {
+    let insertedContent: any = null;
+    let yjsPayload: any = null;
+
+    const pageRepo = {
+      insertPage: jest.fn(async (values: any) => {
+        insertedContent = values.content;
+        return { id: 'page-id', slugId: 'slug-id' };
+      }),
+    };
+    const generalQueue = { add: jest.fn().mockReturnValue({ catch: jest.fn() }) };
+    const collaborationGateway = {
+      handleYjsEvent: jest.fn(async (_evt: string, _name: string, payload: any) => {
+        yjsPayload = payload;
+      }),
+    };
+
+    const service = new PageService(
+      pageRepo as any,
+      {} as any, // pagePermissionRepo
+      {} as any, // attachmentRepo
+      {} as any, // db
+      {} as any, // storageService
+      {} as any, // attachmentQueue
+      {} as any, // aiQueue
+      generalQueue as any,
+      {} as any, // eventEmitter
+      collaborationGateway as any,
+      {} as any, // watcherService
+      {} as any, // transclusionService
+    );
+    // Isolate the canonicalize BINDING: return the raw fixture (a deep clone so
+    // canonicalize never mutates the caller's object) instead of running the
+    // real markdown/HTML/JSON parse + schema validation.
+    jest
+      .spyOn(service as any, 'parseProsemirrorContent')
+      .mockImplementation(async (content: any) => structuredClone(content));
+    jest.spyOn(service as any, 'nextPagePosition').mockResolvedValue('a0');
+
+    return { service, getInsertedContent: () => insertedContent, getYjsPayload: () => yjsPayload };
+  }
+
+  it('createPage (full write) canonicalizes footnotes into reference order', async () => {
+    const { service, getInsertedContent } = makeService();
+    await service.create('user-id', 'workspace-id', {
+      spaceId: 'space-id',
+      content: outOfOrderFull(),
+      format: 'json',
+    } as any);
+    // Definitions reordered to reference order [b, a].
+    expect(listDefIds(getInsertedContent())).toEqual(['b', 'a']);
+  });
+
+  it("updatePageContent operation 'replace' canonicalizes footnotes", async () => {
+    const { service, getYjsPayload } = makeService();
+    await service.updatePageContent(
+      'page-id',
+      outOfOrderFull(),
+      'replace' as any,
+      'json' as any,
+      { id: 'user-id' } as any,
+    );
+    expect(getYjsPayload().operation).toBe('replace');
+    expect(listDefIds(getYjsPayload().prosemirrorJson)).toEqual(['b', 'a']);
+  });
+
+  it("append of a definition-only fragment is NOT canonicalized (footnote preserved, not dropped)", async () => {
+    const { service, getYjsPayload } = makeService();
+    await service.updatePageContent(
+      'page-id',
+      defOnlyFragment(),
+      'append' as any,
+      'json' as any,
+      { id: 'user-id' } as any,
+    );
+    // Canonicalizing a reference-less fragment would DROP the whole list; the
+    // fragment must pass through untouched so the merge keeps the definition.
+    expect(getYjsPayload().operation).toBe('append');
+    expect(hasFootnotesList(getYjsPayload().prosemirrorJson)).toBe(true);
+    expect(listDefIds(getYjsPayload().prosemirrorJson)).toEqual(['a']);
+  });
+
+  it('prepend of a reference-reuse fragment is NOT canonicalized (no synthesized garbage list)', async () => {
+    const { service, getYjsPayload } = makeService();
+    await service.updatePageContent(
+      'page-id',
+      refReuseFragment(),
+      'prepend' as any,
+      'json' as any,
+      { id: 'user-id' } as any,
+    );
+    // Canonicalizing would synthesize a bogus empty footnotesList for the reused
+    // reference; the fragment must pass through with no list at all.
+    expect(getYjsPayload().operation).toBe('prepend');
+    expect(hasFootnotesList(getYjsPayload().prosemirrorJson)).toBe(false);
+  });
+});

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

@@ -52,7 +52,7 @@ import {
   INTERNAL_LINK_REGEX,
   extractPageSlugId,
 } from '../../../integrations/export/utils';
-import { markdownToHtml } from '@docmost/editor-ext';
+import { markdownToHtml, canonicalizeFootnotes } from '@docmost/editor-ext';
 import { WatcherService } from '../../watcher/watcher.service';
 import { sql } from 'kysely';
 import { TransclusionService } from '../transclusion/transclusion.service';
@@ -160,9 +160,14 @@ export class PageService {
     let ydoc = undefined;
 
     if (createPageDto?.content && createPageDto?.format) {
-      const prosemirrorJson = await this.parseProsemirrorContent(
-        createPageDto.content,
-        createPageDto.format,
+      // createPage always writes a FULL document, so canonicalize footnotes to
+      // the editor's invariant before persisting (issue #228). Pure + idempotent
+      // + shape-safe: a doc with no footnotes is returned unchanged.
+      const prosemirrorJson = canonicalizeFootnotes(
+        await this.parseProsemirrorContent(
+          createPageDto.content,
+          createPageDto.format,
+        ),
       );
 
       content = prosemirrorJson;
@@ -343,7 +348,17 @@ export class PageService {
     format: ContentFormat,
     user: User,
   ): Promise<void> {
-    const prosemirrorJson = await this.parseProsemirrorContent(content, format);
+    let prosemirrorJson = await this.parseProsemirrorContent(content, format);
+
+    // Canonicalize footnotes ONLY for a full-document write ('replace'). For an
+    // append/prepend FRAGMENT, canonicalizing is semantically wrong (it would
+    // drop a definition-only fragment's list, or synthesize a duplicate empty
+    // definition for a fragment reusing an existing id) — the fragment merges
+    // into the live doc where the editor's footnoteSyncPlugin keeps the invariant
+    // (issue #228, must-fix #1).
+    if (operation === 'replace') {
+      prosemirrorJson = canonicalizeFootnotes(prosemirrorJson);
+    }
 
     const documentName = `page.${pageId}`;
     await this.collaborationGateway.handleYjsEvent(
@@ -1301,6 +1316,24 @@ export class PageService {
       }
     }
 
+    // NOTE: footnote canonicalization is intentionally NOT done here. This
+    // method serves BOTH full writes (createPage / updatePageContent with
+    // operation 'replace') AND fragment writes (append / prepend). Canonicalizing
+    // a FRAGMENT is semantically wrong — e.g. a definition-only fragment has no
+    // references, so the canonicalizer would drop its whole footnotesList (lost
+    // footnotes), and a fragment reusing an existing id would synthesize an empty
+    // duplicate definition. The canonicalizer therefore runs only at the
+    // FULL-DOCUMENT callers (createPage, and updatePageContent for 'replace'),
+    // never on a fragment (issue #228, must-fix #1).
+    // (Future consolidation, architecture B: the import services persist via a
+    // different path; folding all of these into one "prepare JSON for persist"
+    // helper would centralize the canonicalize call — left as follow-up.)
+    //
+    // ENFORCEMENT RULE (#228): any NEW FULL-document persist path MUST call
+    // `canonicalizeFootnotes(json)` before writing (see createPage and
+    // updatePageContent 'replace'); append/prepend FRAGMENT writes MUST NOT (it
+    // would drop or duplicate footnotes — that is exactly why this is per-call-site
+    // rather than a single wrapper here).
     try {
       jsonToNode(prosemirrorJson);
     } catch (err) {

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

@@ -1,4 +1,5 @@
 import { BadRequestException, ConflictException } from '@nestjs/common';
+import { NoResultError } from 'kysely';
 import { ShareAliasService } from './share-alias.service';
 
 /**
@@ -355,6 +356,68 @@ describe('ShareAliasService', () => {
       }
     });
 
+    it('maps a concurrent-delete race in the SWAP branch to a retryable 409 (not a 200-without-alias)', async () => {
+      const { service, shareAliasRepo } = makeService();
+      // Name points at another page; reassign confirmed -> swap branch.
+      shareAliasRepo.findByAliasAndWorkspace.mockResolvedValue({
+        id: 'a-1',
+        alias: 'foo',
+        pageId: 'p-other',
+      });
+      // A concurrent removeAlias deleted the row between read and UPDATE, so the
+      // repo's executeTakeFirstOrThrow finds 0 rows and throws NoResultError.
+      shareAliasRepo.updatePageId.mockRejectedValue(
+        new NoResultError({} as any),
+      );
+
+      try {
+        await service.setAlias({
+          workspaceId: 'ws-1',
+          pageId: 'p-1',
+          creatorId: 'u-1',
+          alias: 'foo',
+          confirmReassign: true,
+        });
+        fail('expected ConflictException');
+      } catch (err) {
+        // Crucially NOT a resolved 200 carrying `undefined` as the alias.
+        expect(err).toBeInstanceOf(ConflictException);
+        expect((err as ConflictException).getResponse()).toMatchObject({
+          code: 'ALIAS_PAGE_RACE',
+        });
+      }
+    });
+
+    it('maps a concurrent-delete race in the RENAME branch to a retryable 409 (not a generic 400)', async () => {
+      const { service, shareAliasRepo } = makeService();
+      // New slug is free, but the page already owns an alias we rename in place.
+      shareAliasRepo.findByAliasAndWorkspace.mockResolvedValue(undefined);
+      shareAliasRepo.findByPageId.mockResolvedValue({
+        id: 'a-1',
+        alias: 'te',
+        pageId: 'p-1',
+      });
+      // The row vanished before the UPDATE; repo throws NoResultError rather
+      // than returning undefined (which would dereference undefined.id -> 400).
+      shareAliasRepo.updateAlias.mockRejectedValue(new NoResultError({} as any));
+
+      try {
+        await service.setAlias({
+          workspaceId: 'ws-1',
+          pageId: 'p-1',
+          creatorId: 'u-1',
+          alias: 'ted',
+        });
+        fail('expected ConflictException');
+      } catch (err) {
+        expect(err).toBeInstanceOf(ConflictException);
+        expect(err).not.toBeInstanceOf(BadRequestException);
+        expect((err as ConflictException).getResponse()).toMatchObject({
+          code: 'ALIAS_PAGE_RACE',
+        });
+      }
+    });
+
     it('maps a non-unique-violation db error to BadRequest (Failed to set alias)', async () => {
       const { service, shareAliasRepo } = makeService();
       shareAliasRepo.findByAliasAndWorkspace.mockResolvedValue(undefined);

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

@@ -11,21 +11,21 @@ import { Page, ShareAlias } from '@docmost/db/types/entity.types';
 import { isValidShareAlias, normalizeShareAlias } from './share-alias.util';
 import { InjectKysely } from 'nestjs-kysely';
 import { KyselyDB } from '@docmost/db/types/kysely.types';
-import { executeTx } from '@docmost/db/utils';
-
-/** Postgres unique_violation. Two unique indexes can raise it on this table. */
-const PG_UNIQUE_VIOLATION = '23505';
+import {
+  executeTx,
+  isUniqueViolation,
+  violatedConstraint,
+} from '@docmost/db/utils';
+import { NoResultError } from 'kysely';
 
 /**
- * Unique index names from the share_aliases migrations. The `postgres@3.x`
- * driver (kysely-postgres-js) surfaces the violated constraint as
- * `err.constraint_name` (NOT `.constraint`); we keep `.constraint` only as a
- * defensive fallback for other drivers.
- *   - ALIAS:  `(workspace_id, alias)`  -> the vanity NAME is taken.
+ * Unique index name from the share_aliases migrations whose violation we map to
+ * a DISTINCT, non-misleading outcome:
  *   - PAGE_ID: partial `(workspace_id, page_id) WHERE page_id IS NOT NULL`
  *             -> a concurrent writer already gave THIS page an alias.
+ * The `(workspace_id, alias)` index (the vanity NAME being taken) needs no
+ * constant: it is the default "Alias already taken" mapping.
  */
-const UNIQUE_ALIAS_INDEX = 'share_aliases_workspace_id_alias_unique';
 const UNIQUE_PAGE_ID_INDEX = 'share_aliases_workspace_id_page_id_unique';
 
 export interface ResolvedAliasTarget {
@@ -171,11 +171,23 @@ export class ShareAliasService {
       ) {
         throw err;
       }
+      // The row we read was deleted (concurrent `removeAlias`) before our UPDATE
+      // matched it, so `executeTakeFirstOrThrow` found no row. Surface a
+      // retryable conflict instead of a 200-without-alias (swap branch) or a
+      // generic 400 from dereferencing `undefined.id` (rename branch).
+      if (err instanceof NoResultError) {
+        this.logger.warn(
+          'share alias update matched no row (concurrent-delete race)',
+        );
+        throw new ConflictException({
+          message: 'The address changed concurrently, please retry',
+          code: 'ALIAS_PAGE_RACE',
+        });
+      }
       // A unique index fired. Which one decides the message — always log the
       // constraint so the race is diagnosable.
-      if (err?.code === PG_UNIQUE_VIOLATION) {
-        const constraint: string | undefined =
-          err?.constraint_name ?? err?.constraint;
+      if (isUniqueViolation(err)) {
+        const constraint = violatedConstraint(err);
         this.logger.warn(
           `share alias unique violation on ${constraint ?? '<unknown>'}`,
         );
@@ -189,13 +201,8 @@ export class ShareAliasService {
             code: 'ALIAS_PAGE_RACE',
           });
         }
-        // `(workspace_id, alias)` (UNIQUE_ALIAS_INDEX) or any other/unknown
-        // unique index: treat as the vanity name being claimed first.
-        if (constraint && constraint !== UNIQUE_ALIAS_INDEX) {
-          this.logger.warn(
-            `unexpected unique index ${constraint} mapped to "Alias already taken"`,
-          );
-        }
+        // `(workspace_id, alias)` or any other/unknown unique index: treat as
+        // the vanity name being claimed first.
         throw new ConflictException({ message: 'Alias already taken' });
       }
       this.logger.error(err);

apps/server/src/database/listeners/page.listener.ts

@@ -21,6 +21,41 @@ export interface TreeNodeSnapshot {
   position: string;
   spaceId: string;
   parentPageId: string | null;
+  // Death-timer deadline carried so the `addTreeNode` broadcast shows the
+  // temporary-note clock marker immediately on every client (incl. the author,
+  // whose optimistic insert can lose the race to this broadcast). null/absent =>
+  // permanent.
+  temporaryExpiresAt?: Date | string | null;
+}
+
+/**
+ * Single canonical builder for a `TreeNodeSnapshot` from a page-like row. Both
+ * the `PAGE_CREATED` event enrichment (`page.repo.insertPage`) and the
+ * `addTreeNode` broadcast (`WsTreeService.broadcastPageCreated`) build this same
+ * snapshot; routing both through here keeps the optional `temporaryExpiresAt`
+ * (and the `?? null` normalisation that pins a permanent note to an explicit
+ * null) from silently drifting between the two literals.
+ */
+export function toTreeNodeSnapshot(page: {
+  id: string;
+  slugId: string;
+  title: string | null;
+  icon: string | null;
+  position: string;
+  spaceId: string;
+  parentPageId: string | null;
+  temporaryExpiresAt?: Date | string | null;
+}): TreeNodeSnapshot {
+  return {
+    id: page.id,
+    slugId: page.slugId,
+    title: page.title,
+    icon: page.icon,
+    position: page.position,
+    spaceId: page.spaceId,
+    parentPageId: page.parentPageId,
+    temporaryExpiresAt: page.temporaryExpiresAt ?? null,
+  };
 }
 
 export class PageEvent {

apps/server/src/database/migrations/20260626T150000-ai-agent-roles-catalog-source.ts

@@ -0,0 +1,19 @@
+import { type Kysely } from 'kysely';
+
+export async function up(db: Kysely<any>): Promise<void> {
+  // `source` links an imported role back to its catalog origin
+  // `{ slug, language, version }`. Nullable: null => a manually-created role
+  // (no catalog provenance). The version lets the admin UI offer an UPDATE when
+  // the catalog ships a newer revision of the same slug.
+  await db.schema
+    .alterTable('ai_agent_roles')
+    .addColumn('source', 'jsonb', (col) => col)
+    .execute();
+}
+
+export async function down(db: Kysely<any>): Promise<void> {
+  await db.schema
+    .alterTable('ai_agent_roles')
+    .dropColumn('source')
+    .execute();
+}

apps/server/src/database/migrations/20260626T160000-ai-agent-roles-catalog-source-unique.ts

@@ -0,0 +1,31 @@
+import { type Kysely, sql } from 'kysely';
+
+export async function up(db: Kysely<any>): Promise<void> {
+  // A catalog-imported role is uniquely identified within a workspace by its
+  // `source.slug` + `source.language` (a multilingual catalog: the `ru` variant
+  // of a slug installed as `en` is a SEPARATE install — hence both keys). The
+  // import path skips a slug+language already installed using an in-memory
+  // snapshot (installedKeys), but two CONCURRENT imports of the same bundle each
+  // read a stale snapshot and would both insert the same slug+language,
+  // duplicating the role. This partial unique index is the database-level
+  // backstop: the second insert gets a 23505 the service treats as
+  // "already installed" (skip), so the two imports converge on ONE role.
+  //
+  // Partial on `source IS NOT NULL` so MANUALLY-created roles (source NULL) are
+  // unconstrained — there can be many of those. Also partial on
+  // `deleted_at IS NULL` (like the existing name-unique index) so a soft-deleted
+  // role does not block re-importing the same slug+language later, matching the
+  // app's snapshot (listByWorkspace filters out soft-deleted rows).
+  await sql`
+    CREATE UNIQUE INDEX IF NOT EXISTS ai_agent_roles_workspace_source_unique
+    ON ai_agent_roles (workspace_id, (source ->> 'slug'), (source ->> 'language'))
+    WHERE source IS NOT NULL AND deleted_at IS NULL
+  `.execute(db);
+}
+
+export async function down(db: Kysely<any>): Promise<void> {
+  await db.schema
+    .dropIndex('ai_agent_roles_workspace_source_unique')
+    .ifExists()
+    .execute();
+}

apps/server/src/database/repos/ai-agent-roles/ai-agent-roles.repo.spec.ts

@@ -1,4 +1,4 @@
-import { AiAgentRoleRepo } from './ai-agent-roles.repo';
+import { AiAgentRoleRepo, parseSource } from './ai-agent-roles.repo';
 import type { KyselyDB } from '../../types/kysely.types';
 
 /**
@@ -132,4 +132,77 @@ describe('AiAgentRoleRepo insert/update auto-start columns', () => {
     expect(set2.mock.calls[0][0].launchMessage).toBeNull();
     expect('autoStart' in set2.mock.calls[0][0]).toBe(false);
   });
+
+  it('insert binds `source` (jsonb); update sets it only when present', async () => {
+    const { repo, values } = makeInsertRepo();
+    await repo.insert({
+      workspaceId: 'ws-1',
+      name: 'R',
+      instructions: 'do',
+      source: { slug: 'researcher', language: 'en', version: 1 },
+    });
+    // jsonbBind returns a RawBuilder for a non-empty object (not null).
+    expect(values.mock.calls[0][0].source).not.toBeNull();
+
+    const { repo: repo2, set } = makeUpdateRepo();
+    await repo2.update('r-1', 'ws-1', { name: 'X' });
+    expect('source' in set.mock.calls[0][0]).toBe(false);
+
+    const { repo: repo3, set: set3 } = makeUpdateRepo();
+    await repo3.update('r-1', 'ws-1', {
+      source: { slug: 's', language: 'en', version: 2 },
+    });
+    expect('source' in set3.mock.calls[0][0]).toBe(true);
+  });
+});
+
+/**
+ * parseSource is THE single form validator for the `source` jsonb column: a
+ * JSON-string (legacy double-encoded) is parsed; a FULLY-VALID object
+ * ({ slug, language, version }) passes through as a typed RoleSource; anything
+ * partial or wrong-shaped degrades to null (= manual role). This is the
+ * stricter-than-before guard that closes the drift where a weak `{}`/`{slug:123}`
+ * value used to be stamped as a valid source by the read path.
+ */
+describe('parseSource', () => {
+  it('parses a legacy double-encoded JSON string into the typed source', () => {
+    expect(
+      parseSource('{"slug":"researcher","language":"en","version":1}'),
+    ).toEqual({ slug: 'researcher', language: 'en', version: 1 });
+  });
+
+  it('passes a fully-valid already-parsed object through', () => {
+    const obj = { slug: 's', language: 'en', version: 2 };
+    expect(parseSource(obj)).toEqual(obj);
+  });
+
+  it('returns the typed RoleSource (extra keys tolerated) for a valid shape', () => {
+    const src = parseSource({ slug: 's', language: 'ru', version: 3 });
+    expect(src).not.toBeNull();
+    // Narrowed to RoleSource: the fields are present and correctly typed.
+    expect(src?.slug).toBe('s');
+    expect(src?.language).toBe('ru');
+    expect(src?.version).toBe(3);
+  });
+
+  it('null / array / non-object / unparseable string => null', () => {
+    expect(parseSource(null)).toBeNull();
+    expect(parseSource([1, 2])).toBeNull();
+    expect(parseSource(42)).toBeNull();
+    expect(parseSource('not json')).toBeNull();
+  });
+
+  it('partial / wrong-typed shapes => null (no weak-but-typed-as-valid drift)', () => {
+    // Empty object: no slug/language/version.
+    expect(parseSource({})).toBeNull();
+    // slug present but not a string.
+    expect(parseSource({ slug: 123, language: 'en', version: 1 })).toBeNull();
+    // slug only, missing language + version.
+    expect(parseSource({ slug: 'a' })).toBeNull();
+    // empty-string slug / language are not valid catalog keys.
+    expect(parseSource({ slug: '', language: 'en', version: 1 })).toBeNull();
+    expect(parseSource({ slug: 'a', language: '', version: 1 })).toBeNull();
+    // version must be a number, not a numeric string.
+    expect(parseSource({ slug: 'a', language: 'en', version: '1' })).toBeNull();
+  });
 });

apps/server/src/database/repos/ai-agent-roles/ai-agent-roles.repo.ts

@@ -2,7 +2,7 @@ import { Injectable } from '@nestjs/common';
 import { InjectKysely } from 'nestjs-kysely';
 import { KyselyDB, KyselyTransaction } from '../../types/kysely.types';
 import { dbOrTx, jsonbBind, parseJsonbValue } from '../../utils';
-import { AiAgentRole } from '@docmost/db/types/entity.types';
+import { AiAgentRole, RoleSource } from '@docmost/db/types/entity.types';
 
 /** The jsonb shape persisted in `model_config` (loosely typed for the column). */
 type ModelConfigValue = Record<string, unknown> | null;
@@ -81,6 +81,8 @@ export class AiAgentRoleRepo {
       autoStart?: boolean;
       // null/'' => stored as null (client default launch message).
       launchMessage?: string | null;
+      // Catalog origin { slug, language, version } | null. null => manual role.
+      source?: Record<string, unknown> | null;
     },
     trx?: KyselyTransaction,
   ): Promise<AiAgentRole> {
@@ -103,6 +105,9 @@ export class AiAgentRoleRepo {
         autoStart: values.autoStart ?? true,
         // Empty string is treated as "no custom text" => null.
         launchMessage: values.launchMessage || null,
+        // Same cast reason as modelConfig (see above).
+        // eslint-disable-next-line @typescript-eslint/no-explicit-any
+        source: jsonbBind(values.source) as any,
       })
       .returningAll()
       .executeTakeFirst();
@@ -124,6 +129,8 @@ export class AiAgentRoleRepo {
       autoStart?: boolean;
       // undefined => unchanged; null/'' => clear to null; string => set.
       launchMessage?: string | null;
+      // undefined => unchanged; null => clear; object => set.
+      source?: Record<string, unknown> | null;
     },
     trx?: KyselyTransaction,
   ): Promise<void> {
@@ -142,6 +149,9 @@ export class AiAgentRoleRepo {
       // Empty string clears to null (client default launch message).
       set.launchMessage = patch.launchMessage || null;
     }
+    if (patch.source !== undefined) {
+      set.source = jsonbBind(patch.source);
+    }
     await db
       .updateTable('aiAgentRoles')
       .set(set)
@@ -192,14 +202,46 @@ export function parseModelConfig(
   );
 }
 
-/** Normalize a DB row so `modelConfig` is always an object or null. The cast
- *  bridges parseModelConfig's concrete `Record | null` to the column's broad
- *  generated `JsonValue` type (an object is a valid JsonValue at runtime). */
+/**
+ * THE single form validator for the `source` jsonb column: parse the value read
+ * from the DB into a fully-valid {@link RoleSource} or null. Same legacy
+ * double-encoding self-heal as {@link parseModelConfig} (a JSON string is parsed
+ * once), then validates the FULL shape — `slug` and `language` non-empty
+ * strings, `version` a number. A null / corrupt / partially-shaped value (e.g.
+ * `{}`, `{ slug: 123 }`, `{ slug: 'a' }` missing language/version) degrades to
+ * null (= manually created, no catalog provenance), so a bad row never breaks
+ * the read path AND never stamps a half-built object as a valid `RoleSource`.
+ * Both the repo read-path and the service share this so the contract cannot
+ * drift between layers.
+ */
+export function parseSource(value: unknown): RoleSource | null {
+  return parseJsonbValue(value, isRoleSource);
+}
+
+/** Full-shape guard for a persisted `source` jsonb value (see parseSource). */
+function isRoleSource(v: unknown): v is RoleSource {
+  if (v === null || typeof v !== 'object' || Array.isArray(v)) return false;
+  const obj = v as Record<string, unknown>;
+  return (
+    typeof obj.slug === 'string' &&
+    obj.slug.length > 0 &&
+    typeof obj.language === 'string' &&
+    obj.language.length > 0 &&
+    typeof obj.version === 'number'
+  );
+}
+
+/** Normalize a DB row so `modelConfig` and `source` are always a valid object or
+ *  null. The casts bridge the concrete parsed types (`Record | null`,
+ *  `RoleSource | null`) to the column's broad generated `JsonValue` type — both
+ *  are valid JsonValues at runtime; RoleSource lacks the JsonObject index
+ *  signature so it routes through `unknown`. */
 function normalizeRow(row: AiAgentRole): AiAgentRole {
   return {
     ...row,
     modelConfig: parseModelConfig(
       row.modelConfig,
     ) as AiAgentRole['modelConfig'],
+    source: parseSource(row.source) as unknown as AiAgentRole['source'],
   };
 }

apps/server/src/database/repos/ai-chat/ai-chat.repo.spec.ts

@@ -0,0 +1,85 @@
+import { AiChatRepo } from './ai-chat.repo';
+import type { KyselyDB } from '../../types/kysely.types';
+
+/**
+ * Unit test for AiChatRepo.findLatestByPage — the "bound chat" resolver behind
+ * #191 (auto-open the last chat created on a document). It builds the scoping
+ * query, so we assert the EXACT predicates/ordering the spec mandates over a
+ * chainable builder mock (no live DB): user + workspace + page scope, the
+ * deletedAt filter, newest-by-createdAt with an id tiebreaker, limit 1. A
+ * live-Postgres ordering test is out of scope for this pure unit test.
+ */
+describe('AiChatRepo.findLatestByPage', () => {
+  type Recorded = {
+    table?: string;
+    wheres: Array<[string, string, unknown]>;
+    orderBys: Array<[string, string]>;
+    limit?: number;
+  };
+
+  function makeDb(result: unknown): { db: KyselyDB; rec: Recorded } {
+    const rec: Recorded = { wheres: [], orderBys: [] };
+    const builder: Record<string, unknown> = {};
+    const chain = () => builder;
+    builder.selectAll = chain;
+    builder.where = (col: string, op: string, val: unknown) => {
+      rec.wheres.push([col, op, val]);
+      return builder;
+    };
+    builder.orderBy = (col: string, dir: string) => {
+      rec.orderBys.push([col, dir]);
+      return builder;
+    };
+    builder.limit = (n: number) => {
+      rec.limit = n;
+      return builder;
+    };
+    builder.executeTakeFirst = () => Promise.resolve(result);
+    const db = {
+      selectFrom: (table: string) => {
+        rec.table = table;
+        return builder;
+      },
+    } as unknown as KyselyDB;
+    return { db, rec };
+  }
+
+  it('returns the matched chat and scopes by user + workspace + page (deletedAt null)', async () => {
+    const chat = { id: 'c1', creatorId: 'u1', workspaceId: 'ws1', pageId: 'p1' };
+    const { db, rec } = makeDb(chat);
+    const repo = new AiChatRepo(db);
+
+    const res = await repo.findLatestByPage('u1', 'ws1', 'p1');
+
+    expect(res).toBe(chat);
+    expect(rec.table).toBe('aiChats');
+    expect(rec.wheres).toEqual(
+      expect.arrayContaining([
+        ['creatorId', '=', 'u1'],
+        ['workspaceId', '=', 'ws1'],
+        ['pageId', '=', 'p1'],
+        ['deletedAt', 'is', null],
+      ]),
+    );
+  });
+
+  it('orders newest-first by createdAt then id, limit 1', async () => {
+    const { db, rec } = makeDb(undefined);
+    const repo = new AiChatRepo(db);
+
+    await repo.findLatestByPage('u1', 'ws1', 'p1');
+
+    expect(rec.orderBys).toEqual([
+      ['createdAt', 'desc'],
+      ['id', 'desc'],
+    ]);
+    expect(rec.limit).toBe(1);
+  });
+
+  it('returns undefined when the page has no owned chat', async () => {
+    const { db } = makeDb(undefined);
+    const repo = new AiChatRepo(db);
+
+    await expect(repo.findLatestByPage('u1', 'ws1', 'p1')).resolves.toBeUndefined();
+  });
+});

apps/server/src/database/repos/ai-chat/ai-chat.repo.ts

@@ -80,6 +80,32 @@ export class AiChatRepo {
     });
   }
 
+  /**
+   * The "bound chat" for a document: the requesting user's most recently
+   * created, non-deleted chat whose origin page is `pageId`. Auto-opened when
+   * the AI chat window is opened on that page. Newest-by-createdAt wins, so a
+   * chat created later on the same page supersedes earlier ones — exactly how
+   * "new chat -> becomes the bound one" falls out for free. Scoped to the user +
+   * workspace, so a foreign pageId can only ever match the caller's own chats.
+   */
+  async findLatestByPage(
+    creatorId: string,
+    workspaceId: string,
+    pageId: string,
+  ): Promise<AiChat | undefined> {
+    return this.db
+      .selectFrom('aiChats')
+      .selectAll('aiChats')
+      .where('creatorId', '=', creatorId)
+      .where('workspaceId', '=', workspaceId)
+      .where('pageId', '=', pageId)
+      .where('deletedAt', 'is', null)
+      .orderBy('createdAt', 'desc')
+      .orderBy('id', 'desc') // stable tiebreaker, mirrors findByCreator's cursor
+      .limit(1)
+      .executeTakeFirst();
+  }
+
   async insert(
     insertable: InsertableAiChat,
     trx?: KyselyTransaction,

apps/server/src/database/repos/favorite/favorite.repo.ts

@@ -7,7 +7,7 @@ import { executeWithCursorPagination } from '@docmost/db/pagination/cursor-pagin
 import { jsonObjectFrom } from 'kysely/helpers/postgres';
 import { ExpressionBuilder, SelectQueryBuilder, sql } from 'kysely';
 import { DB } from '@docmost/db/types/db';
-import { dbOrTx } from '@docmost/db/utils';
+import { dbOrTx, isUniqueViolation } from '@docmost/db/utils';
 
 export const FavoriteType = {
   PAGE: 'page',
@@ -29,7 +29,8 @@ export class FavoriteRepo {
         .returningAll()
         .executeTakeFirst();
     } catch (err: any) {
-      if (err?.code === '23505') return undefined;
+      // Idempotent favorite: a duplicate (already-favorited) is not an error.
+      if (isUniqueViolation(err)) return undefined;
       throw err;
     }
   }

apps/server/src/database/repos/page/page.repo.ts

@@ -16,7 +16,10 @@ import { jsonArrayFrom, jsonObjectFrom } from 'kysely/helpers/postgres';
 import { SpaceMemberRepo } from '@docmost/db/repos/space/space-member.repo';
 import { EventEmitter2 } from '@nestjs/event-emitter';
 import { EventName } from '../../../common/events/event.contants';
-import { TreeUpdateSnapshot } from '../../listeners/page.listener';
+import {
+  TreeUpdateSnapshot,
+  toTreeNodeSnapshot,
+} from '../../listeners/page.listener';
 
 /**
  * Optional extras for the PAGE_UPDATED event emitted by updatePage(s). Lets the
@@ -200,17 +203,10 @@ export class PageRepo {
     this.eventEmitter.emit(EventName.PAGE_CREATED, {
       pageIds: [result.id],
       workspaceId: result.workspaceId,
-      pages: [
-        {
-          id: result.id,
-          slugId: result.slugId,
-          title: result.title,
-          icon: result.icon,
-          position: result.position,
-          spaceId: result.spaceId,
-          parentPageId: result.parentPageId,
-        },
-      ],
+      // Built via the shared snapshot helper so the field copy (and the
+      // death-timer deadline that shows the sidebar clock marker without a
+      // reload) can't drift from the `addTreeNode` broadcast literal.
+      pages: [toTreeNodeSnapshot(result)],
     });
 
     return result;

apps/server/src/database/repos/share-alias/share-alias.repo.spec.ts

@@ -94,7 +94,9 @@ describe('ShareAliasRepo', () => {
         return builder;
       }),
       returning: jest.fn(() => builder),
-      executeTakeFirst: jest.fn().mockResolvedValue({ id: 'a-1' }),
+      // Retarget uses executeTakeFirstOrThrow so a row reaped by a concurrent
+      // delete (0 rows matched) raises NoResultError instead of returning undefined.
+      executeTakeFirstOrThrow: jest.fn().mockResolvedValue({ id: 'a-1' }),
     };
     const db = { updateTable: jest.fn(() => builder) } as unknown as KyselyDB;
     const repo = new ShareAliasRepo(db);
@@ -121,7 +123,11 @@ describe('ShareAliasRepo', () => {
         return builder;
       }),
       returning: jest.fn(() => builder),
-      executeTakeFirst: jest.fn().mockResolvedValue({ id: 'a-1', alias: 'ted' }),
+      // Rename uses executeTakeFirstOrThrow so a row reaped by a concurrent
+      // delete (0 rows matched) raises NoResultError instead of returning undefined.
+      executeTakeFirstOrThrow: jest
+        .fn()
+        .mockResolvedValue({ id: 'a-1', alias: 'ted' }),
     };
     const db = { updateTable: jest.fn(() => builder) } as unknown as KyselyDB;
     const repo = new ShareAliasRepo(db);

apps/server/src/database/repos/share-alias/share-alias.repo.ts

@@ -92,6 +92,12 @@ export class ShareAliasRepo {
    * Rename an existing alias row in place (the vanity-slug edit, e.g.
    * `te` -> `ted`). Keeps the row's id/page_id/creator so the page's single
    * alias pointer is preserved — only the human-readable name changes.
+   *
+   * Uses `executeTakeFirstOrThrow`: if a concurrent `delete` reaps this row
+   * between the service's read and this UPDATE (READ COMMITTED), the UPDATE
+   * matches 0 rows and kysely throws `NoResultError` rather than returning
+   * `undefined` for a `Promise<ShareAlias>`. The service maps that to a
+   * retryable conflict instead of dereferencing `undefined.id`.
    */
   async updateAlias(
     id: string,
@@ -105,7 +111,7 @@ export class ShareAliasRepo {
       .where('id', '=', id)
       .where('workspaceId', '=', workspaceId)
       .returning(this.baseFields)
-      .executeTakeFirst();
+      .executeTakeFirstOrThrow();
   }
 
   /**
@@ -127,7 +133,15 @@ export class ShareAliasRepo {
       .execute();
   }
 
-  /** Retarget an existing alias to a new page (the "swap" operation). */
+  /**
+   * Retarget an existing alias to a new page (the "swap" operation).
+   *
+   * Uses `executeTakeFirstOrThrow`: if a concurrent `delete` reaps this row
+   * between the service's read and this UPDATE, the UPDATE matches 0 rows and
+   * kysely throws `NoResultError` instead of returning `undefined` into the 200
+   * response (a "success" with no alias). The service maps that to a retryable
+   * conflict.
+   */
   async updatePageId(
     id: string,
     pageId: string,
@@ -140,7 +154,7 @@ export class ShareAliasRepo {
       .where('id', '=', id)
       .where('workspaceId', '=', workspaceId)
       .returning(this.baseFields)
-      .executeTakeFirst();
+      .executeTakeFirstOrThrow();
   }
 
   async delete(

apps/server/src/database/types/db.d.ts

@@ -618,6 +618,8 @@ export interface AiAgentRoles {
   autoStart: Generated<boolean>;
   // Optional custom auto-start text. null/empty => client default launch message.
   launchMessage: string | null;
+  // Catalog origin of an imported role: { slug, language, version } | null. null => manually created.
+  source: Json | null;
   createdAt: Generated<Timestamp>;
   updatedAt: Generated<Timestamp>;
   deletedAt: Timestamp | null;

apps/server/src/database/types/entity.types.ts

@@ -81,6 +81,24 @@ export type UpdatableAiMcpServer = Updateable<Omit<AiMcpServersTable, 'id'>>;
 // A role replaces the persona layer of the system prompt (instructions) and may
 // optionally override the chat model (`modelConfig`). Soft-deletable.
 export type AiAgentRole = Selectable<AiAgentRoles>;
+
+/**
+ * The validated shape of the `source` jsonb column on ai_agent_roles: the
+ * catalog origin of an imported role. `version` lets the admin UI offer an
+ * UPDATE when the catalog ships a newer revision of the same slug; null `source`
+ * (not this type) means a manually-created role with no catalog provenance.
+ *
+ * THE single contract for that column, shared by the repo read-path
+ * (`parseSource`, the only form validator) and the service, so the persisted
+ * shape can never be validated weakly in one layer and strongly in another.
+ * Defined here (a leaf db-types module both already import `AiAgentRole` from) to
+ * avoid an import cycle between the repo and the service.
+ */
+export interface RoleSource {
+  slug: string;
+  language: string;
+  version: number;
+}
 export type InsertableAiAgentRole = Insertable<AiAgentRoles>;
 export type UpdatableAiAgentRole = Updateable<Omit<AiAgentRoles, 'id'>>;
 

apps/server/src/database/unique-violation.spec.ts

@@ -0,0 +1,51 @@
+import { isUniqueViolation, violatedConstraint } from './utils';
+
+/**
+ * Unit tests for the driver-bound Postgres unique-violation helpers extracted
+ * from the share-alias service (and now shared with favorite.repo). They encode
+ * two `kysely-postgres-js` / `postgres@3.x` quirks: the SQLSTATE is the string
+ * `'23505'`, and the violated index name arrives as `constraint_name` (with
+ * `constraint` only a fallback for other drivers).
+ */
+describe('isUniqueViolation', () => {
+  it('is true for a 23505 error', () => {
+    expect(isUniqueViolation({ code: '23505' })).toBe(true);
+  });
+
+  it('is false for any other code', () => {
+    expect(isUniqueViolation({ code: '08006' })).toBe(false);
+  });
+
+  it('is false when there is no code / not an object', () => {
+    expect(isUniqueViolation({})).toBe(false);
+    expect(isUniqueViolation(null)).toBe(false);
+    expect(isUniqueViolation(undefined)).toBe(false);
+    expect(isUniqueViolation(new Error('boom'))).toBe(false);
+  });
+});
+
+describe('violatedConstraint', () => {
+  it('reads the postgres@3.x `constraint_name` field', () => {
+    expect(
+      violatedConstraint({ code: '23505', constraint_name: 'idx_a' }),
+    ).toBe('idx_a');
+  });
+
+  it('falls back to `constraint` when `constraint_name` is absent', () => {
+    expect(violatedConstraint({ code: '23505', constraint: 'idx_b' })).toBe(
+      'idx_b',
+    );
+  });
+
+  it('prefers `constraint_name` over `constraint` when both are present', () => {
+    expect(
+      violatedConstraint({ constraint_name: 'idx_a', constraint: 'idx_b' }),
+    ).toBe('idx_a');
+  });
+
+  it('is undefined when neither field is present', () => {
+    expect(violatedConstraint({ code: '23505' })).toBeUndefined();
+    expect(violatedConstraint(null)).toBeUndefined();
+    expect(violatedConstraint(undefined)).toBeUndefined();
+  });
+});

apps/server/src/database/utils.ts

@@ -33,6 +33,35 @@ export function dbOrTx(
   }
 }
 
+/** Postgres `unique_violation` SQLSTATE — raised when a write hits a UNIQUE index. */
+const PG_UNIQUE_VIOLATION = '23505';
+
+/**
+ * Whether `err` is a Postgres unique-violation (SQLSTATE `23505`). THE single
+ * check so repos/services stop re-hardcoding the magic code.
+ *
+ * NOTE (#222): `core/ai-chat/roles/ai-agent-roles.service.ts` still carries its
+ * own inline `23505` check on a separate, unmerged branch; it should adopt this
+ * helper (and {@link violatedConstraint}) after #227 lands.
+ */
+export function isUniqueViolation(err: unknown): boolean {
+  return (err as { code?: unknown } | null | undefined)?.code === PG_UNIQUE_VIOLATION;
+}
+
+/**
+ * The name of the UNIQUE index/constraint a `23505` error violated, or
+ * undefined. The `kysely-postgres-js` / `postgres@3.x` driver surfaces it as
+ * `err.constraint_name` (NOT `.constraint`); `.constraint` is kept only as a
+ * defensive fallback for other drivers.
+ */
+export function violatedConstraint(err: unknown): string | undefined {
+  const e = err as
+    | { constraint_name?: string; constraint?: string }
+    | null
+    | undefined;
+  return e?.constraint_name ?? e?.constraint;
+}
+
 /**
  * Bind a JS array/object as a `jsonb` column value, working around a postgres
  * driver double-encoding quirk. THE single implementation — repos that persist

apps/server/src/integrations/environment/environment.service.ts

@@ -289,6 +289,18 @@ export class EnvironmentService {
   // provider/model/key config now lives solely in workspace settings +
   // ai_provider_credentials, with no env fallback. APP_SECRET stays (getAppSecret).
 
+  getAiAgentRolesCatalogSource(): string {
+    // Catalog location: an http(s):// base URL the catalog is fetched from.
+    // The image ships a per-branch default for this baked in at build time
+    // (Dockerfile ARG AI_AGENT_ROLES_CATALOG_URL, set per-branch in CI), but it
+    // is overridable at runtime via the env var (this getter returns that
+    // runtime value). Local-filesystem sources are no longer supported.
+    // Empty/unset => the catalog is unavailable (the provider returns 502).
+    // This is INFRA config (where the catalog lives), not provider/model
+    // config, so an env var is appropriate.
+    return this.configService.get<string>('AI_AGENT_ROLES_CATALOG_URL', '');
+  }
+
   getEventStoreDriver(): string {
     return this.configService
       .get<string>('EVENT_STORE_DRIVER', 'postgres')

apps/server/src/integrations/import/services/file-import-task.service.footnote-canonicalize.spec.ts

@@ -0,0 +1,150 @@
+// Importing FileImportTaskService transitively loads import-formatter.ts, which
+// imports the ESM-only @sindresorhus/slugify package (not in jest's transform
+// allowlist). slugify is irrelevant to the path under test, so it is mocked out
+// to keep the module graph loadable under ts-jest (mirrors the import.service spec).
+jest.mock('@sindresorhus/slugify', () => ({
+  __esModule: true,
+  default: (input: string) => String(input),
+}));
+// import-attachment.service.ts (loaded transitively for DI typing) imports the
+// ESM-only `p-limit` / `image-dimensions`; neither is exercised on the path under
+// test, so stub them so the module graph loads under ts-jest.
+jest.mock('p-limit', () => ({
+  __esModule: true,
+  default: () => (fn: any) => fn(),
+}));
+jest.mock('image-dimensions', () => ({
+  __esModule: true,
+  imageDimensionsFromData: () => undefined,
+}));
+
+import { promises as fs } from 'fs';
+import * as os from 'os';
+import * as path from 'path';
+import { FileImportTaskService } from './file-import-task.service';
+import { ImportService } from './import.service';
+
+/**
+ * Binding test for issue #228 / review #5: FileImportTaskService.processGenericImport
+ * is a NON-editor write path (markdownToHtml -> processHTML -> JSON, never runs
+ * footnoteSyncPlugin), so it canonicalizes footnotes before persisting. This pins
+ * that binding — the same one import.service has a spec for — which previously had
+ * NO spec at all.
+ *
+ * The markdown -> HTML -> ProseMirror conversion is REAL (a real ImportService,
+ * its createYdoc stubbed); the filesystem is a real temp dir with one .md file;
+ * the DB transaction is stubbed to capture the persisted page content.
+ */
+
+// Out-of-order references (c, a, b), a REUSED reference ([^a] twice), and an
+// ORPHAN definition ([^z], never referenced).
+const MARKDOWN = [
+  '# Title',
+  '',
+  'Body refs [^c] and [^a] and [^b] and again [^a].',
+  '',
+  '[^a]: note A',
+  '[^b]: note B',
+  '[^c]: note C',
+  '[^z]: orphan note',
+].join('\n');
+
+function footnoteListIds(content: any): string[] {
+  const list = (content?.content ?? []).find(
+    (n: any) => n.type === 'footnotesList',
+  );
+  return (list?.content ?? [])
+    .filter((n: any) => n.type === 'footnoteDefinition')
+    .map((n: any) => n.attrs?.id);
+}
+
+// A permissive chainable stub for the spaces lookup (selectFrom(...).select(...)
+// .where(...).executeTakeFirst()).
+function chainable(result: any): any {
+  const proxy: any = new Proxy(function () {}, {
+    get: (_t, prop) => {
+      if (prop === 'executeTakeFirst') return async () => result;
+      if (prop === 'execute') return async () => [];
+      return () => proxy;
+    },
+  });
+  return proxy;
+}
+
+describe('FileImportTaskService.processGenericImport — footnote canonicalization (#228)', () => {
+  it('orders footnotes by first reference, dedupes reuse, and drops orphans on zip import', async () => {
+    const extractDir = await fs.mkdtemp(path.join(os.tmpdir(), 'fit-canon-'));
+    await fs.writeFile(path.join(extractDir, 'note.md'), MARKDOWN, 'utf-8');
+
+    // Real ImportService for the html -> JSON conversion; stub the yjs encode.
+    const importService = new ImportService(
+      {} as any,
+      {} as any,
+      {} as any,
+      {} as any,
+    );
+    jest
+      .spyOn(importService as any, 'createYdoc')
+      .mockResolvedValue(Buffer.from([]) as any);
+
+    let captured: any = null;
+    const trx = {
+      insertInto: (table: string) => ({
+        values: (v: any) => {
+          if (table === 'pages') captured = v;
+          return { execute: async () => {} };
+        },
+      }),
+    };
+    const db: any = {
+      selectFrom: () => chainable({ slug: 'space-slug' }),
+      transaction: () => ({ execute: (fn: any) => fn(trx) }),
+    };
+
+    const importAttachmentService = {
+      processAttachments: async ({ html }: any) => html,
+    };
+    const backlinkRepo = { insertBacklink: jest.fn() };
+    const eventEmitter = { emit: jest.fn() };
+    const auditService = { logBatchWithContext: jest.fn() };
+
+    const pageService = { nextPagePosition: async () => 'a0' };
+
+    const service = new FileImportTaskService(
+      {} as any, // storageService
+      importService as any,
+      pageService as any,
+      backlinkRepo as any,
+      db,
+      importAttachmentService as any,
+      eventEmitter as any,
+      auditService as any,
+    );
+
+    const fileTask: any = {
+      id: 'task-1',
+      source: 'generic',
+      spaceId: 'space-1',
+      workspaceId: 'ws-1',
+      creatorId: 'user-1',
+    };
+
+    try {
+      await service.processGenericImport({ extractDir, fileTask });
+
+      expect(captured).toBeTruthy();
+      const content = captured.content;
+      // Reference order is c, a, b (NOT the markdown definition order a, b, c).
+      expect(footnoteListIds(content)).toEqual(['c', 'a', 'b']);
+      // Orphan [^z] dropped; reused [^a] collapses to one definition; one list.
+      expect(footnoteListIds(content)).not.toContain('z');
+      const lists = (content.content ?? []).filter(
+        (n: any) => n.type === 'footnotesList',
+      );
+      expect(lists).toHaveLength(1);
+      expect(footnoteListIds(content).filter((id) => id === 'a')).toHaveLength(1);
+    } finally {
+      await fs.rm(extractDir, { recursive: true, force: true });
+    }
+  });
+});

apps/server/src/integrations/import/services/file-import-task.service.ts

@@ -18,7 +18,7 @@ import { generateSlugId } from '../../../common/helpers';
 import { v7 } from 'uuid';
 import { generateJitteredKeyBetween } from 'fractional-indexing-jittered';
 import { FileTask, InsertablePage } from '@docmost/db/types/entity.types';
-import { markdownToHtml } from '@docmost/editor-ext';
+import { markdownToHtml, canonicalizeFootnotes } from '@docmost/editor-ext';
 import { getProsemirrorContent } from '../../../common/helpers/prosemirror/utils';
 import { formatImportHtml } from '../utils/import-formatter';
 import {
@@ -496,9 +496,19 @@ export class FileImportTaskService {
               await this.importService.processHTML(html),
             );
 
-            const { title, prosemirrorJson } =
+            const { title, prosemirrorJson: extractedJson } =
               this.importService.extractTitleAndRemoveHeading(pmState);
 
+            // Canonicalize footnote topology on this non-editor write path
+            // (markdownToHtml/processHTML never runs footnoteSyncPlugin), so a
+            // zip-imported page's footnotes are reference-ordered, deduped, and
+            // orphan-free like the editor's invariant (issue #228). Pure +
+            // idempotent + shape-safe; a footnote-free doc is unchanged.
+            // (Future consolidation, architecture B: like import.service, this
+            // path persists directly rather than via PageService — a shared
+            // "prepare JSON for persist" helper would centralize this call.)
+            const prosemirrorJson = canonicalizeFootnotes(extractedJson);
+
             const insertablePage: InsertablePage = {
               id: page.id,
               slugId: page.slugId,

apps/server/src/integrations/import/services/import.service.footnote-canonicalize.spec.ts

@@ -0,0 +1,139 @@
+// Importing ImportService transitively loads import-formatter.ts, which imports
+// the ESM-only @sindresorhus/slugify package (not in jest's transform
+// allowlist). slugify is irrelevant to the path under test, so it is mocked out
+// to keep the module graph loadable under ts-jest.
+jest.mock('@sindresorhus/slugify', () => ({
+  __esModule: true,
+  default: (input: string) => String(input),
+}));
+
+import { ImportService } from './import.service';
+import { canonicalizeFootnotes } from '@docmost/editor-ext';
+
+/**
+ * Integration-ish test for the USER-FACING markdown import path
+ * (`ImportService.importPage`). It exercises the REAL markdown -> HTML -> JSON
+ * conversion and asserts that the stored page content has its footnotes
+ * canonicalized — the gap that issue #228 fixes: the import path builds
+ * ProseMirror JSON directly (never running the editor's footnoteSyncPlugin), so
+ * before this wiring the stored footnotes kept the markdown's physical
+ * definition order (out of order vs. references), retained orphan definitions,
+ * and did not collapse reused references.
+ *
+ * The DB/ydoc side-effects are stubbed: `getNewPagePosition` (DB query) and
+ * `createYdoc` (Yjs encode) are spied, and `pageRepo.insertPage` captures the
+ * persisted `content`. Everything between markdown and persistence is REAL.
+ */
+
+// Out-of-order references (c, a, b), a REUSED reference ([^a] twice -> one
+// footnote), and an ORPHAN definition ([^z], never referenced).
+const MARKDOWN = [
+  '# Title',
+  '',
+  'Body refs [^c] and [^a] and [^b] and again [^a].',
+  '',
+  '[^a]: note A',
+  '[^b]: note B',
+  '[^c]: note C',
+  '[^z]: orphan note',
+].join('\n');
+
+function makeFile(filename: string, contents: string) {
+  return {
+    filename,
+    toBuffer: async () => Buffer.from(contents),
+  } as any;
+}
+
+function makeService() {
+  let captured: any = null;
+  const pageRepo = {
+    insertPage: jest.fn(async (values: any) => {
+      captured = values;
+      return { id: 'page-id', slugId: 'slug-id' };
+    }),
+  };
+  const service = new ImportService(
+    pageRepo as any,
+    {} as any,
+    {} as any,
+    {} as any,
+  );
+  jest.spyOn(service as any, 'getNewPagePosition').mockResolvedValue('a0');
+  jest
+    .spyOn(service as any, 'createYdoc')
+    .mockResolvedValue(Buffer.from([]) as any);
+  return { service, pageRepo, getCaptured: () => captured };
+}
+
+/** List the footnote-definition ids of the (single) footnotesList, in order. */
+function footnoteListIds(content: any): string[] {
+  const list = (content.content ?? []).find(
+    (n: any) => n.type === 'footnotesList',
+  );
+  if (!list) return [];
+  return (list.content ?? [])
+    .filter((n: any) => n.type === 'footnoteDefinition')
+    .map((n: any) => n.attrs?.id);
+}
+
+function definitionText(content: any, id: string): string | undefined {
+  const list = (content.content ?? []).find(
+    (n: any) => n.type === 'footnotesList',
+  );
+  const def = (list?.content ?? []).find(
+    (n: any) => n.type === 'footnoteDefinition' && n.attrs?.id === id,
+  );
+  return def?.content?.[0]?.content?.[0]?.text;
+}
+
+describe('ImportService.importPage — footnote canonicalization (#228)', () => {
+  it('orders footnotes by first reference, dedupes reuse, and drops orphans', async () => {
+    const { service, getCaptured } = makeService();
+
+    await service.importPage(
+      Promise.resolve(makeFile('note.md', MARKDOWN)),
+      'user-id',
+      'space-id',
+      'workspace-id',
+    );
+
+    const content = getCaptured().content;
+    expect(content).toBeTruthy();
+
+    // Reference order is c, a, b (NOT the markdown definition order a, b, c).
+    expect(footnoteListIds(content)).toEqual(['c', 'a', 'b']);
+
+    // Definitions preserved and attached to the right ids.
+    expect(definitionText(content, 'c')).toBe('note C');
+    expect(definitionText(content, 'a')).toBe('note A');
+    expect(definitionText(content, 'b')).toBe('note B');
+
+    // Orphan definition [^z] is dropped.
+    expect(footnoteListIds(content)).not.toContain('z');
+
+    // Reused [^a] yields exactly ONE definition, and exactly one list.
+    const lists = (content.content ?? []).filter(
+      (n: any) => n.type === 'footnotesList',
+    );
+    expect(lists).toHaveLength(1);
+    expect(footnoteListIds(content).filter((id) => id === 'a')).toHaveLength(1);
+  });
+
+  it('is idempotent: canonicalizing the stored output again is a no-op', async () => {
+    const { service, getCaptured } = makeService();
+    await service.importPage(
+      Promise.resolve(makeFile('note.md', MARKDOWN)),
+      'user-id',
+      'space-id',
+      'workspace-id',
+    );
+    const stored = getCaptured().content;
+
+    // The stored content is already canonical; running the canonicalizer a second
+    // time must not change it (safe to wire into every write path).
+    const second = canonicalizeFootnotes(stored);
+    expect(second).toEqual(stored);
+    expect(footnoteListIds(second)).toEqual(['c', 'a', 'b']);
+  });
+});

apps/server/src/integrations/import/services/import.service.ts

@@ -17,7 +17,7 @@ import {
 import { generateJitteredKeyBetween } from 'fractional-indexing-jittered';
 import { TiptapTransformer } from '@hocuspocus/transformer';
 import * as Y from 'yjs';
-import { markdownToHtml } from '@docmost/editor-ext';
+import { markdownToHtml, canonicalizeFootnotes } from '@docmost/editor-ext';
 import {
   FileTaskStatus,
   FileTaskType,
@@ -85,7 +85,17 @@ export class ImportService {
 
     const extracted = this.extractTitleAndRemoveHeading(prosemirrorState);
     const title = extracted.title;
-    const prosemirrorJson = extracted.prosemirrorJson;
+    // Imported markdown/HTML is built via markdownToHtml -> htmlToJson, which
+    // never runs the editor's footnoteSyncPlugin, so the footnote topology keeps
+    // the source's PHYSICAL definition order (out of order vs. references),
+    // retains orphan definitions, and is not deduped. Canonicalize before
+    // persisting so the stored page matches the editor's invariant (issue #228).
+    // Pure + idempotent + shape-safe: a doc with no footnotes is unchanged.
+    // (Future consolidation, architecture B: this import path persists directly
+    // via pageRepo.insertPage rather than through PageService.createPage, so the
+    // canonicalize call lives here; folding both into one "prepare JSON for
+    // persist" helper is a sensible follow-up.)
+    const prosemirrorJson = canonicalizeFootnotes(extracted.prosemirrorJson);
 
     const pageTitle = title || fileName;
 

apps/server/src/ws/ws-tree.service.spec.ts

@@ -83,6 +83,27 @@ describe('WsTreeService', () => {
     );
   });
 
+  it('broadcastPageCreated carries temporaryExpiresAt when the page is a temporary note', async () => {
+    const expiresAt = new Date('2026-07-01T00:00:00.000Z');
+    await service.broadcastPageCreated({ ...snapshot, temporaryExpiresAt: expiresAt });
+
+    const data =
+      wsService.emitTreeEvent.mock.calls[0][2].payload.data;
+    // The death-timer deadline reaches receivers so the clock marker renders
+    // immediately (incl. the author if this broadcast wins the optimistic race).
+    expect(data.temporaryExpiresAt).toBe(expiresAt);
+  });
+
+  it('broadcastPageCreated pins temporaryExpiresAt to null for a permanent page', async () => {
+    // Fixture omits temporaryExpiresAt; the `?? null` must send an explicit null
+    // (permanent) rather than undefined, so receivers clear any stale marker.
+    await service.broadcastPageCreated(snapshot);
+
+    const data =
+      wsService.emitTreeEvent.mock.calls[0][2].payload.data;
+    expect(data.temporaryExpiresAt).toBeNull();
+  });
+
   it('broadcastPageDeleted emits deleteTreeNode with the root node only', async () => {
     await service.broadcastPageDeleted({
       ...snapshot,

apps/server/src/ws/ws-tree.service.ts

@@ -5,6 +5,7 @@ import {
   PageMovedEvent,
   TreeNodeSnapshot,
   TreeUpdateSnapshot,
+  toTreeNodeSnapshot,
 } from '../database/listeners/page.listener';
 
 @Injectable()
@@ -28,15 +29,16 @@ export class WsTreeService {
         // Receivers place by `position` among already-loaded siblings, not by
         // this absolute index (sender's loaded set differs from receivers').
         index: 0,
+        // Built via the shared snapshot helper (same one page.repo uses to fill
+        // the event), then extended with the tree-only fields the client
+        // receiver consumes. The helper carries the death-timer deadline
+        // (normalised to null => permanent) so receivers — and the author, if
+        // this broadcast wins the race against the optimistic insert — render
+        // the temporary-note clock marker immediately, without it drifting from
+        // the event literal.
         data: {
-          id: page.id,
-          slugId: page.slugId,
+          ...toTreeNodeSnapshot(page),
           name: page.title ?? '',
-          title: page.title,
-          icon: page.icon,
-          position: page.position,
-          spaceId: page.spaceId,
-          parentPageId: page.parentPageId,
           hasChildren: false,
           children: [],
         },

packages/editor-ext/src/lib/footnote/footnote-canonicalize.test.ts

@@ -0,0 +1,371 @@
+import { describe, it, expect } from 'vitest';
+import { Editor, getSchema } from '@tiptap/core';
+import { Document } from '@tiptap/extension-document';
+import { Paragraph } from '@tiptap/extension-paragraph';
+import { Text } from '@tiptap/extension-text';
+import { FootnoteReference } from './footnote-reference';
+import { FootnotesList } from './footnotes-list';
+import { FootnoteDefinition } from './footnote-definition';
+import { canonicalizeFootnotes } from './footnote-canonicalize';
+import { FOOTNOTE_CORPUS } from './footnote-corpus';
+import {
+  collectReferenceIds,
+  computeFootnoteNumbers,
+  FOOTNOTE_REFERENCE_NAME,
+  FOOTNOTES_LIST_NAME,
+  FOOTNOTE_DEFINITION_NAME,
+} from './footnote-util';
+import { Node as PMNode } from '@tiptap/pm/model';
+
+const extensions = [
+  Document,
+  Paragraph,
+  Text,
+  FootnoteReference,
+  FootnotesList,
+  FootnoteDefinition,
+];
+
+const ref = (id: string) => ({ type: FOOTNOTE_REFERENCE_NAME, attrs: { id } });
+const def = (id: string, text?: string) => ({
+  type: FOOTNOTE_DEFINITION_NAME,
+  attrs: { id },
+  content: [
+    text
+      ? { type: 'paragraph', content: [{ type: 'text', text }] }
+      : { type: 'paragraph' },
+  ],
+});
+const list = (...defs: any[]) => ({ type: FOOTNOTES_LIST_NAME, content: defs });
+const para = (...inline: any[]) => ({ type: 'paragraph', content: inline });
+
+/** Find every node of `type`, document order. */
+function findAll(node: any, type: string, acc: any[] = []): any[] {
+  if (!node || typeof node !== 'object') return acc;
+  if (node.type === type) acc.push(node);
+  if (Array.isArray(node.content)) {
+    for (const c of node.content) findAll(c, type, acc);
+  }
+  return acc;
+}
+
+/** Physical id order of the definitions in the (single) footnotesList. */
+function defOrder(doc: any): string[] {
+  return findAll(doc, FOOTNOTE_DEFINITION_NAME).map((d) => d.attrs.id);
+}
+
+const schema = getSchema(extensions);
+/** Reference order (distinct, document order) computed via the shared util. */
+function refOrder(doc: any): string[] {
+  return collectReferenceIds(PMNode.fromJSON(schema, doc));
+}
+
+describe('canonicalizeFootnotes (pure JSON)', () => {
+  it('orders definitions by FIRST reference (out-of-order list -> 1..N)', () => {
+    // References appear b, a, d, c; the bottom list is in a different (import)
+    // order. The canonical list must follow reference order so reading it top to
+    // bottom yields numbers 1..N.
+    const doc = {
+      type: 'doc',
+      content: [
+        para(
+          { type: 'text', text: 'x' },
+          ref('b'),
+          ref('a'),
+          ref('d'),
+          ref('c'),
+        ),
+        list(def('a', 'A'), def('c', 'C'), def('b', 'B'), def('d', 'D')),
+      ],
+    };
+
+    const out = canonicalizeFootnotes(doc);
+    expect(defOrder(out)).toEqual(['b', 'a', 'd', 'c']);
+    // The physical definition order now matches reference order, so the derived
+    // numbers (1..N) run sequentially down the list.
+    expect(refOrder(out)).toEqual(['b', 'a', 'd', 'c']);
+    const numbers = computeFootnoteNumbers(PMNode.fromJSON(schema, out));
+    expect(numbers.get('b')).toBe(1);
+    expect(numbers.get('a')).toBe(2);
+    expect(numbers.get('d')).toBe(3);
+    expect(numbers.get('c')).toBe(4);
+  });
+
+  it('numbers run 1..N down the canonical list', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('b'), ref('a'), ref('c')),
+        list(def('a', 'A'), def('c', 'C'), def('b', 'B')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    // Definition order == reference order == 1,2,3 reading down.
+    expect(defOrder(out)).toEqual(['b', 'a', 'c']);
+  });
+
+  it('drops an orphan definition (no matching reference)', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('a')),
+        list(def('a', 'A'), def('orphan', 'O')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    expect(defOrder(out)).toEqual(['a']);
+    expect(findAll(out, FOOTNOTE_DEFINITION_NAME)).toHaveLength(1);
+  });
+
+  it('with NO references, removes the footnotesList entirely', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'plain' }),
+        list(def('orphan', 'O')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    expect(findAll(out, FOOTNOTES_LIST_NAME)).toHaveLength(0);
+    expect(findAll(out, FOOTNOTE_DEFINITION_NAME)).toHaveLength(0);
+  });
+
+  it('reuse: repeated references collapse to ONE definition/number', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para(ref('d'), { type: 'text', text: ' a ' }, ref('d'), ref('d')),
+        list(def('d', 'shared')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    // One definition; the three references keep id "d".
+    expect(defOrder(out)).toEqual(['d']);
+    expect(
+      findAll(out, FOOTNOTE_REFERENCE_NAME).map((r) => r.attrs.id),
+    ).toEqual(['d', 'd', 'd']);
+  });
+
+  it('duplicate definitions: first wins, the rest are dropped (never resurface as orphans)', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('d')),
+        list(def('d', 'first'), def('d', 'second'), def('d', 'third')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    const defs = findAll(out, FOOTNOTE_DEFINITION_NAME);
+    expect(defs.map((d) => d.attrs.id)).toEqual(['d']);
+    expect(defs[0].content[0].content[0].text).toBe('first');
+  });
+
+  it('synthesizes an empty definition for a reference that has none', () => {
+    const doc = {
+      type: 'doc',
+      content: [para({ type: 'text', text: 'x' }, ref('missing'))],
+    };
+    const out = canonicalizeFootnotes(doc);
+    expect(defOrder(out)).toEqual(['missing']);
+    const list0 = findAll(out, FOOTNOTES_LIST_NAME);
+    expect(list0).toHaveLength(1);
+  });
+
+  it('merges multiple footnotesList nodes into one', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'a' }, ref('x'), ref('y')),
+        list(def('x', 'X')),
+        para({ type: 'text', text: 'tail' }),
+        list(def('y', 'Y')),
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    expect(findAll(out, FOOTNOTES_LIST_NAME)).toHaveLength(1);
+    expect(defOrder(out)).toEqual(['x', 'y']);
+  });
+
+  it('places the single list before trailing empty paragraphs', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('a')),
+        list(def('a', 'A')),
+        { type: 'paragraph' },
+      ],
+    };
+    const out = canonicalizeFootnotes(doc);
+    const last = out.content[out.content.length - 1];
+    expect(last.type).toBe('paragraph');
+    expect(out.content[out.content.length - 2].type).toBe(FOOTNOTES_LIST_NAME);
+  });
+
+  it('is idempotent: canonicalize(canonicalize(x)) === canonicalize(x)', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('b'), ref('a')),
+        list(def('a', 'A'), def('b', 'B'), def('orphan', 'O')),
+      ],
+    };
+    const once = canonicalizeFootnotes(doc);
+    const twice = canonicalizeFootnotes(once);
+    expect(twice).toEqual(once);
+  });
+
+  it('does not mutate its input', () => {
+    const doc = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'x' }, ref('a')),
+        list(def('orphan', 'O')),
+      ],
+    };
+    const snapshot = JSON.parse(JSON.stringify(doc));
+    canonicalizeFootnotes(doc);
+    expect(doc).toEqual(snapshot);
+  });
+});
+
+/**
+ * GOLDEN PARITY against the live `footnoteSyncPlugin`. The server canonicalizer
+ * must produce EXACTLY what the editor keeps. For every editor-reachable steady
+ * state (the list is already reference-ordered there), driving a real editor to
+ * convergence and then running `canonicalizeFootnotes` on its JSON must be a
+ * byte-for-byte no-op — proving the server output is identical to the editor's.
+ */
+describe('canonicalizeFootnotes golden parity with footnoteSyncPlugin', () => {
+  function makeEditor(content: any) {
+    return new Editor({ extensions, content });
+  }
+
+  /** Load `content`, fire one local edit so the sync plugin converges, return JSON. */
+  function pluginSteadyState(content: any): any {
+    const editor = makeEditor(content);
+    // A local doc change triggers footnoteSyncPlugin.appendTransaction.
+    editor.commands.insertContentAt(1, ' ');
+    const json = editor.state.doc.toJSON();
+    editor.destroy();
+    return json;
+  }
+
+  const corpus: Array<{ name: string; content: any }> = [
+    {
+      name: 'plain ref + def',
+      content: {
+        type: 'doc',
+        content: [para({ type: 'text', text: 'a' }, ref('x')), list(def('x', 'X'))],
+      },
+    },
+    {
+      name: 'two refs, two defs in reference order',
+      content: {
+        type: 'doc',
+        content: [
+          para({ type: 'text', text: 'a' }, ref('x'), { type: 'text', text: 'b' }, ref('y')),
+          list(def('x', 'X'), def('y', 'Y')),
+        ],
+      },
+    },
+    {
+      name: 'orphan definition gets removed',
+      content: {
+        type: 'doc',
+        content: [para({ type: 'text', text: 'a' }, ref('x')), list(def('x', 'X'), def('orphan', 'O'))],
+      },
+    },
+    {
+      name: 'reference missing its definition (synth empty)',
+      content: {
+        type: 'doc',
+        content: [para({ type: 'text', text: 'a' }, ref('x'))],
+      },
+    },
+    {
+      name: 'reuse: repeated references, one definition',
+      content: {
+        type: 'doc',
+        content: [
+          para(ref('d'), { type: 'text', text: ' a ' }, ref('d'), ref('d')),
+          list(def('d', 'shared')),
+        ],
+      },
+    },
+    {
+      name: 'no footnotes at all',
+      content: {
+        type: 'doc',
+        content: [para({ type: 'text', text: 'just text' })],
+      },
+    },
+  ];
+
+  for (const { name, content } of corpus) {
+    it(`steady state is a canonicalize no-op: ${name}`, () => {
+      const steady = pluginSteadyState(content);
+      expect(canonicalizeFootnotes(steady)).toEqual(steady);
+    });
+  }
+
+  it('placement parity: the LIVE plugin leaves a list with NON-EMPTY content after it in place, and canonicalize agrees', () => {
+    // Drives the real footnoteSyncPlugin (not a hand-authored expected): a single
+    // canonical list with body content AFTER it must NOT be repositioned by the
+    // plugin, and the server canonicalizer must agree (step-6 placement parity).
+    const content = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'a' }, ref('x')),
+        list(def('x', 'X')),
+        para({ type: 'text', text: 'epilogue' }),
+      ],
+    };
+    const steady = pluginSteadyState(content);
+    // The plugin did NOT move the list to the end: a non-empty paragraph follows it.
+    const types = steady.content.map((n: any) => n.type);
+    const listPos = types.indexOf(FOOTNOTES_LIST_NAME);
+    expect(listPos).toBeGreaterThanOrEqual(0);
+    expect(listPos).toBeLessThan(types.length - 1);
+    const after = steady.content[listPos + 1];
+    expect(after.type).toBe('paragraph');
+    expect(JSON.stringify(after)).toContain('epilogue');
+    // The canonicalizer is a byte-for-byte no-op on that steady state (parity).
+    expect(canonicalizeFootnotes(steady)).toEqual(steady);
+  });
+
+  it('the canonicalizer and the editor agree on reference order and definition set', () => {
+    const content = {
+      type: 'doc',
+      content: [
+        para({ type: 'text', text: 'a' }, ref('x'), { type: 'text', text: 'b' }, ref('y')),
+        list(def('y', 'Y'), def('x', 'X')), // physically reversed
+      ],
+    };
+    const steady = pluginSteadyState(content);
+    const canon = canonicalizeFootnotes(content);
+    // Same reference order and same DEFINITION SET (ids) in both, even though the
+    // physical list order may differ (the plugin preserves node identity, the
+    // canonicalizer reorders). Numbering — derived from reference order — matches.
+    expect(refOrder(steady)).toEqual(['x', 'y']);
+    expect(defOrder(canon)).toEqual(['x', 'y']);
+    expect(new Set(defOrder(steady))).toEqual(new Set(defOrder(canon)));
+  });
+});
+
+/**
+ * SHARED golden corpus: this editor-ext copy of `canonicalizeFootnotes` and the
+ * MCP mirror (`packages/mcp/src/lib/footnote-canonicalize.ts`) are BOTH run
+ * against the identical { input -> expected } corpus. Pinning the same expected
+ * outputs in both suites makes "the two pure copies behave identically" a
+ * checkable property without coupling the packages (architecture item A). The
+ * MCP mirror of these assertions lives in `test/unit/footnote-corpus.test.mjs`.
+ */
+describe('canonicalizeFootnotes shared golden corpus (editor-ext copy)', () => {
+  for (const { name, input, expected } of FOOTNOTE_CORPUS) {
+    it(`matches the corpus expected output: ${name}`, () => {
+      expect(canonicalizeFootnotes(input)).toEqual(expected);
+      // Idempotent on the corpus too.
+      expect(canonicalizeFootnotes(expected)).toEqual(expected);
+    });
+  }
+});

packages/editor-ext/src/lib/footnote/footnote-canonicalize.ts

@@ -0,0 +1,272 @@
+import {
+  FOOTNOTE_REFERENCE_NAME,
+  FOOTNOTES_LIST_NAME,
+  FOOTNOTE_DEFINITION_NAME,
+} from './footnote-util';
+
+/**
+ * Server-side, EditorView-free port of the footnote integrity invariant that
+ * `footnoteSyncPlugin` maintains in the live editor. Where the plugin is an
+ * `appendTransaction` that only runs inside a ProseMirror `EditorView`, this is
+ * a PURE function over ProseMirror JSON: `canonicalizeFootnotes(doc) -> doc`.
+ *
+ * It exists because the NON-editor write paths served by THIS copy build
+ * ProseMirror JSON directly (never running the editor's plugins), so the
+ * canonical footnote topology was never enforced on those writes. The consumers
+ * of this editor-ext copy are: the server markdown/HTML import
+ * (`markdownToHtml -> htmlToJson` in import.service / file-import-task.service),
+ * `PageService` create/update (`parseProsemirrorContent` for the JSON/markdown/
+ * HTML REST write paths), and the client markdown PASTE path
+ * (`markdown-clipboard.ts`). (The MCP package mirrors this canonicalizer in
+ * `packages/mcp/src/lib/footnote-canonicalize.ts` for its own FULL-document write
+ * paths — `markdownToProseMirrorCanonical` (the page markdown-import path; the
+ * plain `markdownToProseMirror` primitive used for COMMENT bodies does NOT
+ * canonicalize), `update_page_json`, `docmost_transform`, `insert_footnote`,
+ * `copy_page_content` — see that file's header.) All of these are the root cause
+ * of the symptom in the issue: footnotes rendered out of order (`1, 4, 2, 3, …`),
+ * a raw trailing `[^id]: …` block, and orphan definitions, all of which are
+ * simply the result of content written PAST the canonicalizer.
+ *
+ * The desired end-state (identical to the plugin's) is:
+ *
+ *   1. Reference ids in DOCUMENT ORDER are the single source of truth for which
+ *      definitions exist and in what order (numbering is derived from this, see
+ *      `computeFootnoteNumbers`). Repeated references that share an id are REUSE
+ *      (one footnote, one number, one definition) — never re-id'd.
+ *   2. Exactly ONE `footnotesList`, holding one definition per referenced id in
+ *      REFERENCE order, reusing the existing definition node (content preserved)
+ *      or synthesizing an empty one when missing. The list sits after the last
+ *      meaningful block (only trailing empty paragraphs may follow it).
+ *   3. Orphan definitions (no matching reference) are dropped.
+ *   4. Duplicate DEFINITIONS (two nodes sharing an id) are resolved first-wins:
+ *      the first definition for an id is kept; later duplicates carry the SAME
+ *      id, so they can never be referenced separately and are simply dropped.
+ *      This matches the importer's first-wins rule ("one definition per id").
+ *      (The LIVE editor instead re-id's a duplicate definition so a paste/collab
+ *      merge cannot silently lose live user data; the artifacts this copy
+ *      sanitizes are agent/import-authored, so first-wins is the right policy —
+ *      see footnote-sync.ts `resolveCollisions`.)
+ *   5. Idempotent: a document that already satisfies the invariant is returned
+ *      structurally unchanged (the existing definition/list nodes are reused
+ *      verbatim), so re-running the canonicalizer — or running it on a write that
+ *      the editor already canonicalized — is a no-op. This is what makes it safe
+ *      to wire into EVERY write path without spurious mutations / git-sync churn.
+ *
+ * Divergence from the live plugin (intentional): the plugin preserves the
+ * PHYSICAL order of existing definition nodes to keep their Yjs/CRDT subtree
+ * identity stable across collaborators (numbering is decoration-derived, so the
+ * displayed numbers are correct regardless of physical order). This function has
+ * no live CRDT to protect, so when a REPAIR is needed it physically REORDERS the
+ * list into reference order — which is exactly the fix the out-of-order import
+ * needs.
+ *
+ * Placement PARITY with the plugin: when the document is already in the canonical
+ * single-list state, this function leaves that list EXACTLY where it sits (it
+ * does not move it to the end). The plugin behaves the same — it treats one
+ * footnotesList holding the canonical definition set as canonical regardless of
+ * whether content follows it (footnote-sync.ts: `primaryList` falls back to the
+ * last list and `noChangeNeeded` stays true). So on every editor-reachable steady
+ * state the two agree byte-for-byte, including when non-empty content follows the
+ * list; see the golden parity test and the shared corpus.
+ *
+ * Pure: deep-clones its input, never mutates the caller's object, and is
+ * deterministic (no `Math.random`/`Date.now`).
+ */
+export function canonicalizeFootnotes<T = any>(doc: T): T {
+  if (
+    doc == null ||
+    typeof doc !== 'object' ||
+    !Array.isArray((doc as any).content)
+  ) {
+    return doc;
+  }
+  const out = cloneJson(doc) as any;
+
+  // 1) Distinct reference ids in document order (deep — references can live in
+  //    callouts, tables, list items, ...). This is the ordering/numbering truth.
+  const referenceIds: string[] = [];
+  const seenRefIds = new Set<string>();
+  collectReferenceIds(out, referenceIds, seenRefIds);
+
+  // 2) Every definition node in document order (deep — defs normally live inside
+  //    one or more `footnotesList` blocks, but we tolerate stray placements).
+  const defNodes: any[] = [];
+  collectDefinitions(out, defNodes);
+
+  // 3) First definition per id wins. Later duplicates carry the SAME id, so they
+  //    can never be referenced separately and would be orphans — they are simply
+  //    dropped (first-wins; see the file header, item 4).
+  const defById = new Map<string, any>();
+  for (const d of defNodes) {
+    const id = d?.attrs?.id;
+    if (id && !defById.has(id)) defById.set(id, d);
+  }
+
+  // 4) Build the ordered definition list: one per referenced id, in REFERENCE
+  //    order, reusing the existing node (content preserved, id normalized) or
+  //    synthesizing an empty definition. Definitions whose id is NOT referenced
+  //    are orphans and are simply never added. The reused node is SHALLOW-copied
+  //    (id normalized): `out` is already a deep clone and the old lists are cut,
+  //    so a second per-definition deep clone is needless.
+  const orderedDefs: any[] = [];
+  for (const id of referenceIds) {
+    const existing = defById.get(id);
+    if (existing) {
+      orderedDefs.push({
+        ...existing,
+        attrs: { ...(existing.attrs ?? {}), id },
+      });
+    } else {
+      orderedDefs.push(emptyDefinition(id));
+    }
+  }
+
+  // 5) No references -> there must be NO list at all (at any depth).
+  if (referenceIds.length === 0) {
+    stripFootnotesListsDeep(out);
+    return out;
+  }
+
+  // 6) Placement parity with the live plugin: when the document is ALREADY in the
+  //    canonical single-list state, leave that list exactly where it sits instead
+  //    of cutting and re-inserting it at the end. The plugin never repositions a
+  //    sole correct list (footnote-sync.ts), so moving it here would silently
+  //    reorder any user content that follows the list on the first write. The doc
+  //    is in that state when there is exactly one top-level footnotesList, every
+  //    definition in the doc is referenced (no orphans / duplicates: the def count
+  //    equals the canonical count), and the list already holds exactly the
+  //    canonical definitions in reference order.
+  const topLevelLists = out.content.filter(
+    (n: any) => n && n.type === FOOTNOTES_LIST_NAME,
+  );
+  if (
+    topLevelLists.length === 1 &&
+    defNodes.length === orderedDefs.length &&
+    deepEqualJson(topLevelLists[0].content, orderedDefs)
+  ) {
+    return out;
+  }
+
+  // 7) Otherwise rebuild: strip every footnotesList AND every bare
+  //    footnoteDefinition at ANY depth (collectDefinitions gathers defs
+  //    recursively, so a list nested in a callout/blockquote — or a bare
+  //    definition outside any list — would otherwise have its defs copied into the
+  //    rebuilt list while the original survives in place → duplicates) and
+  //    re-insert exactly one list after the last meaningful (non-empty paragraph)
+  //    top-level block, so it coexists with a trailing-node empty paragraph. This
+  //    both repairs a non-canonical doc and (in the import case) physically
+  //    reorders the list into reference order.
+  stripFootnotesListsDeep(out);
+  stripFootnoteDefinitionsDeep(out);
+  const top: any[] = out.content;
+  let insertAt = top.length;
+  while (insertAt > 0 && isEmptyParagraph(top[insertAt - 1])) insertAt--;
+  top.splice(insertAt, 0, { type: FOOTNOTES_LIST_NAME, content: orderedDefs });
+  out.content = top;
+  return out;
+}
+
+/** Remove every `footnotesList` node at ANY depth (mutates the given clone). */
+function stripFootnotesListsDeep(node: any): void {
+  if (!node || typeof node !== 'object' || !Array.isArray(node.content)) return;
+  node.content = node.content.filter(
+    (c: any) => !(c && c.type === FOOTNOTES_LIST_NAME),
+  );
+  for (const child of node.content) stripFootnotesListsDeep(child);
+}
+
+/**
+ * Remove every BARE `footnoteDefinition` node at ANY depth (mutates the given
+ * clone). Runs only in the rebuild path AFTER the lists are stripped, so it
+ * targets definitions that were sitting outside a list (e.g. hand-authored via a
+ * raw-JSON write path and nested in a callout); their content was already copied
+ * into the rebuilt list, so leaving the originals would duplicate them.
+ */
+function stripFootnoteDefinitionsDeep(node: any): void {
+  if (!node || typeof node !== 'object' || !Array.isArray(node.content)) return;
+  node.content = node.content.filter(
+    (c: any) => !(c && c.type === FOOTNOTE_DEFINITION_NAME),
+  );
+  for (const child of node.content) stripFootnoteDefinitionsDeep(child);
+}
+
+/**
+ * Deep equality over plain JSON: arrays are compared POSITIONALLY
+ * (order-SENSITIVE), object keys order-insensitively. The array order-sensitivity
+ * is required for correctness here — a reordered `footnotesList.content` must
+ * compare UNEQUAL so the canonical rebuild fires instead of leaving it in place.
+ */
+function deepEqualJson(a: any, b: any): boolean {
+  if (a === b) return true;
+  if (a == null || b == null || typeof a !== typeof b) return false;
+  if (Array.isArray(a) || Array.isArray(b)) {
+    if (!Array.isArray(a) || !Array.isArray(b) || a.length !== b.length) {
+      return false;
+    }
+    for (let i = 0; i < a.length; i++) {
+      if (!deepEqualJson(a[i], b[i])) return false;
+    }
+    return true;
+  }
+  if (typeof a === 'object') {
+    const ka = Object.keys(a);
+    const kb = Object.keys(b);
+    if (ka.length !== kb.length) return false;
+    for (const k of ka) {
+      if (!Object.prototype.hasOwnProperty.call(b, k)) return false;
+      if (!deepEqualJson(a[k], b[k])) return false;
+    }
+    return true;
+  }
+  return false;
+}
+
+/** A fresh empty definition node for a referenced id with no definition. */
+function emptyDefinition(id: string): any {
+  return {
+    type: FOOTNOTE_DEFINITION_NAME,
+    attrs: { id },
+    content: [{ type: 'paragraph' }],
+  };
+}
+
+function isEmptyParagraph(node: any): boolean {
+  return (
+    !!node &&
+    node.type === 'paragraph' &&
+    (!Array.isArray(node.content) || node.content.length === 0)
+  );
+}
+
+/** Collect DISTINCT footnoteReference ids in document order (first appearance). */
+function collectReferenceIds(
+  node: any,
+  out: string[],
+  seen: Set<string>,
+): void {
+  if (!node || typeof node !== 'object') return;
+  if (node.type === FOOTNOTE_REFERENCE_NAME) {
+    const id = node?.attrs?.id;
+    if (id && !seen.has(id)) {
+      seen.add(id);
+      out.push(id);
+    }
+  }
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) collectReferenceIds(child, out, seen);
+  }
+}
+
+/** Collect every footnoteDefinition node in document order. */
+function collectDefinitions(node: any, out: any[]): void {
+  if (!node || typeof node !== 'object') return;
+  if (node.type === FOOTNOTE_DEFINITION_NAME) out.push(node);
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) collectDefinitions(child, out);
+  }
+}
+
+function cloneJson<T>(v: T): T {
+  if (typeof structuredClone === 'function') return structuredClone(v);
+  return JSON.parse(JSON.stringify(v)) as T;
+}

packages/editor-ext/src/lib/footnote/index.ts

@@ -4,3 +4,4 @@ export * from "./footnotes-list";
 export * from "./footnote-definition";
 export * from "./footnote-numbering";
 export * from "./footnote-sync";
+export * from "./footnote-canonicalize";

packages/editor-ext/tsconfig.json

@@ -22,5 +22,11 @@
     "noFallthroughCasesInSwitch": false
   },
   "include": ["src/**/*"],
-  "exclude": ["node_modules", "dist", "src/**/*.spec.ts", "src/**/*.test.ts"]
+  "exclude": [
+    "node_modules",
+    "dist",
+    "src/**/*.spec.ts",
+    "src/**/*.test.ts",
+    "src/lib/footnote/footnote-corpus.ts"
+  ]
 }

packages/mcp/build/client.js

@@ -7,7 +7,7 @@ import { TiptapTransformer } from "@hocuspocus/transformer";
 import * as Y from "yjs";
 import WebSocket from "ws";
 import { convertProseMirrorToMarkdown } from "./lib/markdown-converter.js";
-import { updatePageContentRealtime, replacePageContent, markdownToProseMirror, mutatePageContent, buildCollabWsUrl, assertYjsEncodable, applyDocToFragment, } from "./lib/collaboration.js";
+import { updatePageContentRealtime, replacePageContent, markdownToProseMirror, markdownToProseMirrorCanonical, mutatePageContent, buildCollabWsUrl, assertYjsEncodable, applyDocToFragment, } from "./lib/collaboration.js";
 import { footnoteWarningsField } from "./lib/footnote-analyze.js";
 import { buildPageTree } from "./lib/tree.js";
 import { serializeDocmostMarkdown, parseDocmostMarkdown, } from "./lib/markdown-document.js";
@@ -17,7 +17,7 @@ import { applyTextEdits, } from "./lib/json-edit.js";
 import { getCollabToken, performLogin } from "./lib/auth-utils.js";
 import { diffDocs, summarizeChange } from "./lib/diff.js";
 import { applyAnchorInDoc, canAnchorInDoc } from "./lib/comment-anchor.js";
-import { blockText, walk, getList, insertMarkerAfter, setCalloutRange, noteItem, mdToInlineNodes, commentsToFootnotes, } from "./lib/transforms.js";
+import { blockText, walk, getList, insertMarkerAfter, setCalloutRange, noteItem, mdToInlineNodes, commentsToFootnotes, canonicalizeFootnotes, insertInlineFootnote, } from "./lib/transforms.js";
 import vm from "node:vm";
 // Supported image types, kept as two lookup tables so both a local file
 // extension and a remote Content-Type can be mapped to the same canonical set.
@@ -1063,10 +1063,15 @@ export class DocmostClient {
         // the markdown link path (which TipTap sanitizes), raw JSON could otherwise
         // inject javascript:/data: link hrefs or media srcs straight into the doc.
         this.validateDocUrls(doc);
+        // Canonicalize footnotes (idempotent): an agent-authored JSON doc cannot
+        // leave footnotes out of order, orphaned, or in multiple lists — the bottom
+        // list + numbering are always derived from reference order. No-op when the
+        // footnotes are already canonical.
+        doc = canonicalizeFootnotes(doc);
         // Write the BODY first, then the title (#159 split-brain): a failed body
         // write (e.g. persist timeout) must not leave a new title over the old body.
         const collabToken = await this.getCollabTokenWithReauth();
-        const mutation = await replacePageContent(pageId, doc, collabToken, this.apiUrl);
+        const mutation = await this.replacePage(pageId, doc, collabToken, this.apiUrl);
         // Body persisted successfully — now it is safe to set the title.
         if (title) {
             await this.client.post("/pages/update", { pageId, title });
@@ -1079,6 +1084,73 @@ export class DocmostClient {
             verify: mutation.verify,
         };
     }
+    /**
+     * AUTHOR-INLINE footnote insertion. The agent supplies only WHERE
+     * (`anchorText`, a snippet of body text to attach the marker after) and WHAT
+     * (`text`, the footnote content as markdown). Numbering and the bottom
+     * `footnotesList` are derived deterministically server-side
+     * (`insertInlineFootnote` -> `canonicalizeFootnotes`): the agent never sees,
+     * assigns, or edits a footnote number or the list, so it CANNOT desync.
+     *
+     * Content DEDUP: when an existing definition has the same content, its id is
+     * reused (one number, one definition, several references). The write is atomic
+     * via `mutatePageContent` (single-writer, page-locked); if the anchor text is
+     * not found the transform aborts with a clear error and no write happens.
+     */
+    async insertFootnote(pageId, anchorText, text) {
+        await this.ensureAuthenticated();
+        if (!anchorText || !anchorText.trim()) {
+            throw new Error("insert_footnote: anchorText is required");
+        }
+        if (text == null || `${text}`.trim() === "") {
+            throw new Error("insert_footnote: text is required");
+        }
+        const collabToken = await this.getCollabTokenWithReauth();
+        let result = null;
+        const mutation = await this.mutatePage(pageId, collabToken, this.apiUrl, (liveDoc) => {
+            const r = insertInlineFootnote(liveDoc, { anchorText, text });
+            if (!r.inserted) {
+                // Abort the page-locked write by throwing: mutatePageContent does not
+                // persist when the transform throws, so a missing anchor leaves the
+                // page untouched (no partial write).
+                throw new Error(`insert_footnote: anchor text not found: ${JSON.stringify(anchorText.slice(0, 80))}`);
+            }
+            result = { footnoteId: r.footnoteId, reused: r.reused };
+            return r.doc;
+        });
+        // The not-found path throws inside the transform (aborting mutatePage), so by
+        // here `result` is always set.
+        const r = result;
+        return {
+            success: true,
+            modified: true,
+            pageId,
+            footnoteId: r.footnoteId,
+            reused: r.reused,
+            message: r.reused
+                ? "Footnote inserted (reused an existing same-content definition)."
+                : "Footnote inserted.",
+            verify: mutation.verify,
+        };
+    }
+    /**
+     * Page-locked write seam over collaboration.mutatePageContent. Production just
+     * delegates; it exists as an overridable method so the insert_footnote wrapper
+     * (transform abort-on-not-found + response shaping) can be unit-tested without
+     * standing up a live Hocuspocus collab socket.
+     */
+    mutatePage(pageId, collabToken, apiUrl, transform) {
+        return mutatePageContent(pageId, collabToken, apiUrl, transform);
+    }
+    /**
+     * Full-document write seam over collaboration.replacePageContent. Production
+     * just delegates; it exists as an overridable method so the full-doc write
+     * tools (update_page_json, copy_page_content) can have their footnote-
+     * canonicalization binding unit-tested without a live Hocuspocus collab socket.
+     */
+    replacePage(pageId, doc, collabToken, apiUrl) {
+        return replacePageContent(pageId, doc, collabToken, apiUrl);
+    }
     /**
      * Export a page to a single self-contained Docmost-flavoured markdown file:
      * meta block + body (with inline comment anchors + diagrams) + comment
@@ -1120,7 +1192,8 @@ export class DocmostClient {
     async importPageMarkdown(pageId, fullMarkdown) {
         await this.ensureAuthenticated();
         const { meta, body, comments } = parseDocmostMarkdown(fullMarkdown);
-        const doc = await markdownToProseMirror(body);
+        // PAGE import: canonicalize footnotes (see markdownToProseMirrorCanonical).
+        const doc = await markdownToProseMirrorCanonical(body);
         const collabToken = await this.getCollabTokenWithReauth();
         const mutation = await replacePageContent(pageId, doc, collabToken, this.apiUrl);
         // Collect distinct comment ids that actually became comment marks in the doc.
@@ -1200,13 +1273,18 @@ export class DocmostClient {
         // uses, so copying never lands a javascript:/data: href/src on the target
         // (parity with updatePageJson; harmless for already-stored source content).
         this.validateDocUrls(content);
+        // Defense-in-depth (#228): this is a FULL-document write, so canonicalize
+        // footnotes before copying — a no-op on already-canonical source content, but
+        // it guarantees a copy can never propagate a non-canonical footnote topology
+        // to the target (parity with the other full-doc write paths).
+        const canonical = canonicalizeFootnotes(content);
         const collabToken = await this.getCollabTokenWithReauth();
-        const mutation = await replacePageContent(targetPageId, content, collabToken, this.apiUrl);
+        const mutation = await this.replacePage(targetPageId, canonical, collabToken, this.apiUrl);
         return {
             success: true,
             sourcePageId,
             targetPageId,
-            copiedNodes: content.content.length,
+            copiedNodes: canonical.content.length,
             verify: mutation.verify,
         };
     }
@@ -1613,7 +1691,10 @@ export class DocmostClient {
                 }
             }
         }
-        // Convert through the full Docmost schema (consistent with page paths)
+        // Convert through the full Docmost schema. Deliberately the NON-canonicalizing
+        // variant: a comment body may carry a footnote definition with no matching
+        // reference, and canonicalization would drop it (data loss). See
+        // markdownToProseMirror vs markdownToProseMirrorCanonical.
         const jsonContent = await markdownToProseMirror(content);
         const payload = {
             pageId,
@@ -1701,6 +1782,7 @@ export class DocmostClient {
     }
     async updateComment(commentId, content) {
         await this.ensureAuthenticated();
+        // NON-canonicalizing on purpose (comment body — see createComment).
         const jsonContent = await markdownToProseMirror(content);
         await this.client.post("/comments/update", {
             commentId,
@@ -2422,6 +2504,8 @@ export class DocmostClient {
                 noteItem,
                 mdToInlineNodes,
                 commentsToFootnotes,
+                canonicalizeFootnotes,
+                insertInlineFootnote,
             },
         };
         // Captured oldDoc / newDoc for the diff (set inside runTransform).
@@ -2455,16 +2539,25 @@ export class DocmostClient {
             if (typeof fn !== "function") {
                 throw new Error("transform must evaluate to a function (doc, ctx) => doc");
             }
-            const result = vm.runInNewContext("f(d, c)", { f: fn, d: sandbox.doc, c: ctx }, { timeout: 5000 });
-            if (!result ||
-                typeof result !== "object" ||
-                result.type !== "doc" ||
-                !Array.isArray(result.content)) {
+            const raw = vm.runInNewContext("f(d, c)", { f: fn, d: sandbox.doc, c: ctx }, { timeout: 5000 });
+            if (!raw ||
+                typeof raw !== "object" ||
+                raw.type !== "doc" ||
+                !Array.isArray(raw.content)) {
                 throw new Error('transform must return a ProseMirror doc node ({ type:"doc", content:[...] })');
             }
-            // Validate the returned doc before it can be written.
-            this.validateDocStructure(result);
-            this.validateDocUrls(result);
+            // Validate the RAW transform output FIRST (structure — including the
+            // MAX_DEPTH guard — and URLs), mirroring updatePageJson. The canonicalizer
+            // recurses without a depth limiter, so validating after it would turn a
+            // too-deep doc into an opaque "Maximum call stack size exceeded" instead of
+            // the intended "nesting exceeds the maximum depth" error.
+            this.validateDocStructure(raw);
+            this.validateDocUrls(raw);
+            // Auto-canonicalize footnotes after the transform (idempotent): no write
+            // path can leave footnotes out of order / orphaned / in a raw `[^id]`
+            // block. In a dryRun preview this may surface footnote edits the script
+            // author did not write (the canonicalizer tidied them) — that is expected.
+            const result = canonicalizeFootnotes(raw);
             newDoc = result;
             return result;
         };

packages/mcp/build/index.js

@@ -637,8 +637,15 @@ export function createDocmostMcpServer(config) {
             "mark-safe), setCalloutRange(doc, n) (sync a [1]…[K] callout range to " +
             "[1]…[n]), noteItem(inlineNodes) (wrap inline nodes in a listItem with a " +
             "fresh id), mdToInlineNodes(markdown) (comment markdown -> inline nodes), " +
-            "and commentsToFootnotes(doc, comments, {notesHeading}) (turn inline " +
-            "comments into numbered footnotes). Footnote convention: markers are " +
+            "commentsToFootnotes(doc, comments, {notesHeading}) (turn inline " +
+            "comments into numbered footnotes), canonicalizeFootnotes(doc) (derive " +
+            "footnote numbering + the single bottom list from reference order, drop " +
+            "orphans/duplicates — runs AUTOMATICALLY on the transform RESULT, so the " +
+            "applied (and dryRun-previewed) doc is always footnote-canonical; a dryRun " +
+            "diff may therefore show footnote tidy-ups your script did not make, and " +
+            "it is idempotent after the first run), and " +
+            "insertInlineFootnote(doc, {anchorText, text}) (author-inline footnote: " +
+            "marker + dedup'd definition, list derived). Footnote convention: markers are " +
             "plain '[N]' text in the body; the notes are an orderedList under a " +
             "heading whose text is 'Примечания переводчика'. The transform runs " +
             "sandboxed (no require/process/fs/network, 5s timeout) and must return a " +
@@ -652,7 +659,8 @@ export function createDocmostMcpServer(config) {
                 "parenthesized function). It receives a clone of the live doc and " +
                 "ctx (comments, log, consume(id), helpers: blockText/walk/getList/" +
                 "insertMarkerAfter/setCalloutRange/noteItem/mdToInlineNodes/" +
-                "commentsToFootnotes) and must return a {type:'doc'} node."),
+                "commentsToFootnotes/canonicalizeFootnotes/insertInlineFootnote) " +
+                "and must return a {type:'doc'} node."),
             dryRun: z
                 .boolean()
                 .optional()
@@ -672,6 +680,33 @@ export function createDocmostMcpServer(config) {
         });
         return jsonContent(result);
     });
+    // Tool: insert_footnote
+    server.registerTool("insert_footnote", {
+        description: "Insert an AUTHOR-INLINE footnote: you specify only WHERE (anchorText) " +
+            "and WHAT (text). The footnote marker is placed right after anchorText in " +
+            "the body, and the bottom footnotes list + the numbering are derived " +
+            "deterministically server-side. You do NOT assign a number, and you " +
+            "never see or edit the footnotes list — so footnotes cannot end up out " +
+            "of order, orphaned, or as a raw '[^id]' block. If a footnote with the " +
+            "SAME text already exists, its number is REUSED (one definition, several " +
+            "references). The write is atomic and won't clobber concurrent edits; if " +
+            "anchorText is not found, nothing is written and an error is returned.",
+        inputSchema: {
+            pageId: z.string().min(1),
+            anchorText: z
+                .string()
+                .min(1)
+                .describe("A snippet of existing body text; the footnote marker is inserted " +
+                "immediately after its first occurrence (mark-safe)."),
+            text: z
+                .string()
+                .min(1)
+                .describe("The footnote content as markdown (becomes the definition)."),
+        },
+    }, async ({ pageId, anchorText, text }) => {
+        const result = await docmostClient.insertFootnote(pageId, anchorText, text);
+        return jsonContent(result);
+    });
     // Tool: diff_page_versions
     registerShared(SHARED_TOOL_SPECS.diffPageVersions, async ({ pageId, from, to }) => {
         const result = await docmostClient.diffPageVersions(pageId, from, to);

packages/mcp/build/lib/collaboration.js

@@ -11,6 +11,7 @@ import { docmostExtensions, docmostSchema } from "./docmost-schema.js";
 import { withPageLock } from "./page-lock.js";
 import { sanitizeForYjs, findUnstorableAttr } from "./node-ops.js";
 import { lexFootnoteLines } from "./footnote-lex.js";
+import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 import { summarizeChange } from "./diff.js";
 /**
  * Build the descriptive error for an opaque Yjs encode failure ("Unexpected
@@ -343,7 +344,20 @@ function extractFootnotes(markdown) {
         section: `<section data-footnotes>${inner}</section>`,
     };
 }
-/** Convert markdown to a ProseMirror doc using the full Docmost schema. */
+/**
+ * Convert markdown to a ProseMirror doc using the full Docmost schema.
+ *
+ * This conversion does NOT canonicalize footnotes — it is the shared, content-
+ * preserving primitive used by BOTH page write paths and COMMENT bodies
+ * (createComment / updateComment). Canonicalization MUST NOT run on a comment
+ * body: a comment may legitimately contain a footnote-definition line
+ * (`[^1]: text`) with no matching reference, and the canonicalizer drops a
+ * reference-less footnotesList — which would silently delete the comment's text.
+ *
+ * Page write paths that DO need the canonical footnote topology call
+ * `markdownToProseMirrorCanonical` instead (markdown import, update_page markdown
+ * path). Keep this function reference-loss-free.
+ */
 export async function markdownToProseMirror(markdownContent) {
     const withCallouts = await preprocessCallouts(markdownContent);
     const { body, section } = extractFootnotes(withCallouts);
@@ -351,6 +365,20 @@ export async function markdownToProseMirror(markdownContent) {
     const bridged = bridgeTaskLists(html);
     return generateJSON(bridged, docmostExtensions);
 }
+/**
+ * Page-write variant of `markdownToProseMirror`: converts markdown then enforces
+ * the canonical footnote topology. The footnote `section` markdown is emitted in
+ * DEFINITION order, but numbering derives from REFERENCE order, so without this
+ * the bottom list renders out of order (`1, 4, 2, 3, …`); orphan definitions and
+ * duplicate lists are also normalized. Idempotent — a no-op once canonical, and a
+ * no-op for footnote-free content.
+ *
+ * Use this ONLY for full-document PAGE writes (never for comment bodies, where it
+ * would drop a reference-less footnote definition — see `markdownToProseMirror`).
+ */
+export async function markdownToProseMirrorCanonical(markdownContent) {
+    return canonicalizeFootnotes(await markdownToProseMirror(markdownContent));
+}
 /**
  * Build the collaboration WebSocket URL from an API base URL:
  * switch http(s)->ws(s), strip a trailing /api, mount on /collab.
@@ -708,6 +736,8 @@ export async function replacePageContent(pageId, prosemirrorDoc, collabToken, ba
  * Tables and :::callout::: blocks survive thanks to the full schema.
  */
 export async function updatePageContentRealtime(pageId, markdownContent, collabToken, baseUrl) {
-    const tiptapJson = await markdownToProseMirror(markdownContent);
+    // PAGE write: canonicalize footnotes (markdown import builds the bottom list in
+    // definition order; numbering is reference-ordered).
+    const tiptapJson = await markdownToProseMirrorCanonical(markdownContent);
     return await mutatePageContent(pageId, collabToken, baseUrl, () => tiptapJson);
 }

packages/mcp/build/lib/footnote-authoring.js

@@ -0,0 +1,88 @@
+/**
+ * Inline-authoring helpers for footnotes (MCP).
+ *
+ * These build/identify footnote DEFINITION nodes for the author-inline tool
+ * (`insertInlineFootnote` in transforms.ts): a content key to de-duplicate notes
+ * by text, a definition-node factory, and a fresh uuidv7-style id generator.
+ *
+ * Split out of `footnote-canonicalize.ts` so that module stays a pure MIRROR of
+ * the editor-ext canonicalizer (compositionally symmetric to the editor-ext
+ * copy, which keeps its authoring helpers in `footnote-util.ts`). The pure
+ * canonicalizer has no dependency on these.
+ */
+const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
+function cloneJson(v) {
+    if (typeof structuredClone === "function")
+        return structuredClone(v);
+    return JSON.parse(JSON.stringify(v));
+}
+/**
+ * Normalized content key for de-duplicating footnote DEFINITIONS by their text.
+ *
+ * Two definitions with the same key are the SAME footnote — so the inline
+ * authoring tool reuses one id (one number, one definition, several references)
+ * instead of minting a second definition. Key = plaintext (whitespace-collapsed,
+ * trimmed) PLUS a signature of the inline mark types in order, so two notes that
+ * read the same but differ in formatting (one bold, one plain) are NOT merged.
+ * Conservative: only an exact match merges.
+ */
+export function footnoteContentKey(defNode) {
+    const parts = [];
+    const visit = (n) => {
+        if (!n || typeof n !== "object")
+            return;
+        if (n.type === "text" && typeof n.text === "string") {
+            const marks = Array.isArray(n.marks)
+                ? n.marks.map((m) => m?.type).filter(Boolean).sort().join(",")
+                : "";
+            parts.push(`${n.text}${marks}`);
+        }
+        if (Array.isArray(n.content))
+            for (const c of n.content)
+                visit(c);
+    };
+    visit(defNode);
+    // Collapse the assembled text's whitespace and trim, keeping the mark
+    // signature attached so formatting differences still distinguish notes.
+    return parts
+        .join("")
+        .replace(/[ \t\r\n]+/g, " ")
+        .trim();
+}
+/**
+ * Build a footnoteDefinition node from inline ProseMirror nodes, keyed by id.
+ */
+export function makeFootnoteDefinition(id, inlineNodes) {
+    const content = Array.isArray(inlineNodes) ? cloneJson(inlineNodes) : [];
+    return {
+        type: FOOTNOTE_DEFINITION_NAME,
+        attrs: { id },
+        content: [{ type: "paragraph", content }],
+    };
+}
+/**
+ * Generate a uuidv7-style id (time-ordered), matching editor-ext's
+ * `generateFootnoteId`. Used for a genuinely-new inline footnote id.
+ */
+export function generateFootnoteId() {
+    const now = Date.now();
+    const timeHex = now.toString(16).padStart(12, "0");
+    const rand = (length) => {
+        let s = "";
+        for (let i = 0; i < length; i++)
+            s += Math.floor(Math.random() * 16).toString(16);
+        return s;
+    };
+    const versioned = "7" + rand(3);
+    const variantNibble = (8 + Math.floor(Math.random() * 4)).toString(16);
+    const variant = variantNibble + rand(3);
+    return (timeHex.slice(0, 8) +
+        "-" +
+        timeHex.slice(8, 12) +
+        "-" +
+        versioned +
+        "-" +
+        variant +
+        "-" +
+        rand(12));
+}

packages/mcp/build/lib/footnote-canonicalize.js

@@ -0,0 +1,215 @@
+/**
+ * Server-side footnote canonicalizer (MCP mirror — PURE).
+ *
+ * `canonicalizeFootnotes(doc)` is a pure ProseMirror-JSON port of the editor's
+ * `footnoteSyncPlugin` end-state, identical in behaviour to
+ * `@docmost/editor-ext`'s `canonicalizeFootnotes`. It is mirrored here — rather
+ * than imported from editor-ext — for the SAME reason `footnote-lex.ts` and the
+ * `docmost-schema.ts` nodes are mirrored: the MCP package is deliberately
+ * decoupled from the browser/React-heavy editor barrel and operates on plain
+ * JSON. The editor-ext copy owns the golden test against the live plugin; this
+ * copy must stay behaviourally identical (a SHARED golden corpus, exercised by
+ * both test suites, pins that — see `test/unit/footnote-corpus.mjs`).
+ *
+ * This module is the pure MIRROR only. The inline-authoring helpers
+ * (`footnoteContentKey`, `makeFootnoteDefinition`, `generateFootnoteId`) used by
+ * `insertInlineFootnote` live in the sibling `footnote-authoring.ts`, so this
+ * file is compositionally symmetric to the editor-ext copy.
+ *
+ * Why it exists: every NON-editor write path (markdown import, update_page_json,
+ * docmost_transform, insert_footnote) builds ProseMirror JSON directly, so the
+ * editor's footnote plugins never run and the canonical topology (sequential
+ * numbering by first reference, one trailing list, no orphans, no raw `[^id]`)
+ * was never enforced. Running this at the end of every write path closes that
+ * gap; because it is idempotent, it is a no-op when the footnotes are already
+ * canonical (no spurious mutations / git-sync churn).
+ *
+ * ENFORCEMENT RULE (#228): any NEW FULL-document persist path MUST call
+ * `canonicalizeFootnotes(doc)` before writing — the current callers are
+ * `markdownToProseMirrorCanonical` (page markdown import/update; the plain
+ * `markdownToProseMirror` used for COMMENT bodies must NOT, or it would drop a
+ * reference-less definition), `update_page_json`, `docmost_transform`,
+ * `insert_footnote`, and `copy_page_content`. Append/prepend FRAGMENT writes MUST
+ * NOT canonicalize. This is deliberately per-call-site (the replace-vs-fragment
+ * and comment-vs-page nuances make a single naive wrapper unsafe).
+ */
+const FOOTNOTE_REFERENCE_NAME = "footnoteReference";
+const FOOTNOTES_LIST_NAME = "footnotesList";
+const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
+function cloneJson(v) {
+    if (typeof structuredClone === "function")
+        return structuredClone(v);
+    return JSON.parse(JSON.stringify(v));
+}
+function isEmptyParagraph(node) {
+    return (!!node &&
+        node.type === "paragraph" &&
+        (!Array.isArray(node.content) || node.content.length === 0));
+}
+function collectReferenceIds(node, out, seen) {
+    if (!node || typeof node !== "object")
+        return;
+    if (node.type === FOOTNOTE_REFERENCE_NAME) {
+        const id = node?.attrs?.id;
+        if (id && !seen.has(id)) {
+            seen.add(id);
+            out.push(id);
+        }
+    }
+    if (Array.isArray(node.content)) {
+        for (const child of node.content)
+            collectReferenceIds(child, out, seen);
+    }
+}
+function collectDefinitions(node, out) {
+    if (!node || typeof node !== "object")
+        return;
+    if (node.type === FOOTNOTE_DEFINITION_NAME)
+        out.push(node);
+    if (Array.isArray(node.content)) {
+        for (const child of node.content)
+            collectDefinitions(child, out);
+    }
+}
+function emptyDefinition(id) {
+    return {
+        type: FOOTNOTE_DEFINITION_NAME,
+        attrs: { id },
+        content: [{ type: "paragraph" }],
+    };
+}
+/**
+ * Deep equality over plain JSON: arrays are compared POSITIONALLY
+ * (order-SENSITIVE), object keys order-insensitively. The array order-sensitivity
+ * is required for correctness here — a reordered `footnotesList.content` must
+ * compare UNEQUAL so the canonical rebuild fires instead of leaving it in place.
+ */
+function deepEqualJson(a, b) {
+    if (a === b)
+        return true;
+    if (a == null || b == null || typeof a !== typeof b)
+        return false;
+    if (Array.isArray(a) || Array.isArray(b)) {
+        if (!Array.isArray(a) || !Array.isArray(b) || a.length !== b.length) {
+            return false;
+        }
+        for (let i = 0; i < a.length; i++) {
+            if (!deepEqualJson(a[i], b[i]))
+                return false;
+        }
+        return true;
+    }
+    if (typeof a === "object") {
+        const ka = Object.keys(a);
+        const kb = Object.keys(b);
+        if (ka.length !== kb.length)
+            return false;
+        for (const k of ka) {
+            if (!Object.prototype.hasOwnProperty.call(b, k))
+                return false;
+            if (!deepEqualJson(a[k], b[k]))
+                return false;
+        }
+        return true;
+    }
+    return false;
+}
+/**
+ * Canonicalize footnotes in a ProseMirror-JSON document. See the file header and
+ * the editor-ext twin for the full contract. Pure (deep-clones input,
+ * deterministic, idempotent).
+ */
+export function canonicalizeFootnotes(doc) {
+    if (doc == null ||
+        typeof doc !== "object" ||
+        !Array.isArray(doc.content)) {
+        return doc;
+    }
+    const out = cloneJson(doc);
+    // 1) Distinct reference ids in document order (deep — refs can live in
+    //    callouts, tables, list items, ...). The ordering/numbering truth.
+    const referenceIds = [];
+    collectReferenceIds(out, referenceIds, new Set());
+    // 2) Every definition node in document order (deep).
+    const defNodes = [];
+    collectDefinitions(out, defNodes);
+    // 3) First definition per id wins; later duplicates carry the SAME id, so they
+    //    cannot be referenced separately and would be orphans — they are dropped.
+    const defById = new Map();
+    for (const d of defNodes) {
+        const id = d?.attrs?.id;
+        if (id && !defById.has(id))
+            defById.set(id, d);
+    }
+    // 4) Build the ordered definition list: one per referenced id, in REFERENCE
+    //    order, reusing the existing node (shallow-copied, id normalized — `out` is
+    //    already deep-cloned and the old lists are cut) or synthesizing an empty
+    //    one. Definitions whose id is not referenced are orphans and never added.
+    const orderedDefs = [];
+    for (const id of referenceIds) {
+        const existing = defById.get(id);
+        if (existing) {
+            orderedDefs.push({
+                ...existing,
+                attrs: { ...(existing.attrs ?? {}), id },
+            });
+        }
+        else {
+            orderedDefs.push(emptyDefinition(id));
+        }
+    }
+    // 5) No references -> there must be NO list at all (at any depth).
+    if (referenceIds.length === 0) {
+        stripFootnotesListsDeep(out);
+        return out;
+    }
+    // 6) Placement parity with the live plugin: when the document is ALREADY in the
+    //    canonical single-list state, leave that list exactly where it sits rather
+    //    than cutting and re-inserting it at the end (the plugin never repositions a
+    //    sole correct list, so moving it would silently reorder any content that
+    //    follows the list on the first write).
+    const topLevelLists = out.content.filter((n) => n && n.type === FOOTNOTES_LIST_NAME);
+    if (topLevelLists.length === 1 &&
+        defNodes.length === orderedDefs.length &&
+        deepEqualJson(topLevelLists[0].content, orderedDefs)) {
+        return out;
+    }
+    // 7) Otherwise rebuild: strip every footnotesList AND every bare
+    //    footnoteDefinition at ANY depth (collectDefinitions gathers defs
+    //    recursively, so a list nested in a callout/blockquote — or a bare
+    //    definition outside any list — would otherwise have its defs copied into the
+    //    rebuilt list while the original survives in place → duplicates) and
+    //    re-insert exactly one list after the last meaningful (non-empty paragraph)
+    //    top-level block.
+    stripFootnotesListsDeep(out);
+    stripFootnoteDefinitionsDeep(out);
+    const top = out.content;
+    let insertAt = top.length;
+    while (insertAt > 0 && isEmptyParagraph(top[insertAt - 1]))
+        insertAt--;
+    top.splice(insertAt, 0, { type: FOOTNOTES_LIST_NAME, content: orderedDefs });
+    out.content = top;
+    return out;
+}
+/** Remove every `footnotesList` node at ANY depth (mutates the given clone). */
+function stripFootnotesListsDeep(node) {
+    if (!node || typeof node !== "object" || !Array.isArray(node.content))
+        return;
+    node.content = node.content.filter((c) => !(c && c.type === FOOTNOTES_LIST_NAME));
+    for (const child of node.content)
+        stripFootnotesListsDeep(child);
+}
+/**
+ * Remove every BARE `footnoteDefinition` node at ANY depth (mutates the given
+ * clone). Runs only in the rebuild path AFTER the lists are stripped, so it
+ * targets definitions that were sitting outside a list (e.g. hand-authored via a
+ * raw-JSON write path and nested in a callout); their content was already copied
+ * into the rebuilt list, so leaving the originals would duplicate them.
+ */
+function stripFootnoteDefinitionsDeep(node) {
+    if (!node || typeof node !== "object" || !Array.isArray(node.content))
+        return;
+    node.content = node.content.filter((c) => !(c && c.type === FOOTNOTE_DEFINITION_NAME));
+    for (const child of node.content)
+        stripFootnoteDefinitionsDeep(child);
+}

packages/mcp/build/lib/transforms.js

@@ -14,6 +14,9 @@
  *  - `marks` arrays are preserved verbatim when fragments are split/reordered.
  */
 import { blockPlainText } from "./node-ops.js";
+import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
+import { footnoteContentKey, makeFootnoteDefinition, generateFootnoteId, } from "./footnote-authoring.js";
+export { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 /** Deep-clone a JSON-serializable value without mutating the original. */
 function clone(value) {
     if (typeof structuredClone === "function") {
@@ -64,6 +67,36 @@ export function getList(doc, predicate) {
     });
     return found;
 }
+/**
+ * Textblocks that hold raw text but do NOT accept inline atom nodes. A
+ * `footnoteReference` is `group:"inline", atom:true`; `codeBlock` is
+ * `content:"text*"` (text only), so splicing a footnoteReference into it yields
+ * an invalid document. (paragraph/heading/detailsSummary are `inline*` and DO
+ * accept it; footnote definitions live inside a footnotesList which the
+ * footnote inserter excludes via `beforeBlock`.)
+ */
+const INLINE_ATOM_FORBIDDEN_BLOCKS = new Set(["codeBlock"]);
+/**
+ * Footnote-notes subtrees the inline footnote inserter must never split into (at
+ * any depth): a `footnotesList` and the `footnoteDefinition`s it holds. Anchoring
+ * a reference inside one of these would later be dropped as an orphan by the
+ * canonicalizer, taking the existing definition's text with it.
+ */
+const FOOTNOTE_NOTES_SUBTREES = new Set([
+    "footnotesList",
+    "footnoteDefinition",
+]);
+/** True if `node` IS, or contains at any depth, a footnotesList/footnoteDefinition. */
+function containsFootnoteNotes(node) {
+    if (!isObject(node))
+        return false;
+    if (FOOTNOTE_NOTES_SUBTREES.has(node.type))
+        return true;
+    if (Array.isArray(node.content)) {
+        return node.content.some((c) => containsFootnoteNotes(c));
+    }
+    return false;
+}
 /**
  * Insert `marker` as a PLAIN (unmarked) text run right after the first
  * occurrence of `anchor`.
@@ -83,6 +116,19 @@ export function getList(doc, predicate) {
  * false when the anchor text was not found in any in-scope block.
  */
 export function insertMarkerAfter(doc, anchor, marker, opts = {}) {
+    // A plain marker is a leading-space-padded unmarked text run.
+    return insertNodesAfterAnchor(doc, anchor, () => [{ type: "text", text: " " + marker }], opts);
+}
+/**
+ * Mark-safe insertion CORE: split the inline text run that holds the END of
+ * `anchor` (preserving the surrounding marks) and splice the nodes produced by
+ * `makeMiddle()` in at the split point. `insertMarkerAfter` (plain text marker)
+ * and `insertInlineFootnote` (a `footnoteReference` node) are both thin callers —
+ * the only difference is WHAT is inserted (a space-padded text run vs. a node
+ * that should hug the preceding word), which is exactly what `makeMiddle`
+ * decides. Operates on a clone; returns `{ doc, inserted }`.
+ */
+function insertNodesAfterAnchor(doc, anchor, makeMiddle, opts = {}) {
     const out = clone(doc);
     if (!isObject(out) || !Array.isArray(out.content) || !anchor) {
         return { doc: out, inserted: false };
@@ -111,10 +157,25 @@ export function insertMarkerAfter(doc, anchor, marker, opts = {}) {
             if (inserted || !isObject(container) || !Array.isArray(container.content)) {
                 return;
             }
+            // Skip a forbidden subtree entirely (e.g. footnotesList/footnoteDefinition):
+            // never split into it, but keep `offset` aligned for any sibling text after
+            // it within this block.
+            if (opts.skipSubtreeTypes && opts.skipSubtreeTypes.has(container.type)) {
+                offset += blockPlainText(container).length;
+                return;
+            }
             const inline = container.content;
             // Detect whether this array is an inline array (contains text nodes).
             const hasText = inline.some((n) => isObject(n) && n.type === "text");
             if (hasText) {
+                // Refuse a textblock whose content spec cannot hold the inserted nodes
+                // (e.g. a codeBlock for an inline atom). Keep `offset` aligned for any
+                // sibling textblocks in this same block, then bail so the search falls
+                // through to the next candidate block.
+                if (opts.forbidBlockTypes && opts.forbidBlockTypes.has(container.type)) {
+                    offset += blockPlainText(container).length;
+                    return;
+                }
                 for (let i = 0; i < inline.length; i++) {
                     const n = inline[i];
                     const len = isObject(n) ? blockPlainText(n).length : 0;
@@ -136,8 +197,9 @@ export function insertMarkerAfter(doc, anchor, marker, opts = {}) {
                         if (before.length > 0) {
                             parts.push({ ...n, text: before, marks: [...marks] });
                         }
-                        // Marker is a PLAIN run: no marks copied. Leading space separates it.
-                        parts.push({ type: "text", text: " " + marker });
+                        // The inserted nodes are caller-decided (a space-padded marker run,
+                        // or a node that hugs the word). They carry no copied marks.
+                        parts.push(...makeMiddle());
                         if (after.length > 0) {
                             parts.push({ ...n, text: after, marks: [...marks] });
                         }
@@ -227,14 +289,16 @@ export function noteItem(inlineNodes) {
  * Wrap inline ProseMirror nodes in a real footnoteDefinition node keyed by id:
  *   { type:"footnoteDefinition", attrs:{id}, content:[{ type:"paragraph", content }] }
  * (mirrors the editor-ext / docmost-schema FootnoteDefinition node).
+ *
+ * Built on the shared `makeFootnoteDefinition` factory (footnote-authoring.ts);
+ * the only extra is a fresh block id on the inner paragraph (Docmost stamps one,
+ * and the canonicalizer preserves attrs as-is). Single factory, one place to
+ * change the definition shape.
  */
 export function footnoteDefinition(id, inlineNodes) {
-    const content = Array.isArray(inlineNodes) ? clone(inlineNodes) : [];
-    return {
-        type: "footnoteDefinition",
-        attrs: { id },
-        content: [{ type: "paragraph", attrs: { id: freshId() }, content }],
-    };
+    const node = makeFootnoteDefinition(id, inlineNodes);
+    node.content[0].attrs = { id: freshId() };
+    return node;
 }
 /**
  * Replace every `[N]` body marker and `\u0000FN<i>\u0000` comment placeholder in
@@ -471,3 +535,97 @@ export function commentsToFootnotes(doc, comments, opts = {}) {
     const synced = setCalloutRange(working, definitions.length);
     return { doc: synced.doc, consumed };
 }
+/**
+ * AUTHOR-INLINE footnote insertion. The caller supplies WHERE (anchorText) and
+ * WHAT (markdown text); numbering and the bottom list are derived server-side by
+ * `canonicalizeFootnotes`. The caller never sees or edits `footnotesList`, never
+ * assigns a number, and cannot desync — orphans / out-of-order lists / raw
+ * `[^id]` markdown are structurally impossible.
+ *
+ * Content DEDUP (#3 in the issue): if an existing definition has the SAME
+ * normalized content key, its id is REUSED (the new reference points at it: one
+ * number, one definition, several references). Otherwise a fresh uuid id is
+ * minted and a new definition added. Conservative — only an exact content match
+ * merges.
+ *
+ * Mechanics: the `footnoteReference` node is inserted DIRECTLY at the anchor via
+ * the same mark-safe split as `insertMarkerAfter` (the shared
+ * `insertNodesAfterAnchor` core), so it hugs the preceding word with no text
+ * sentinel round-trip. The whole document is then canonicalized.
+ *
+ * Operates on a clone of `doc`. When the anchor is not found, returns the input
+ * unchanged with `inserted:false`.
+ */
+export function insertInlineFootnote(doc, opts) {
+    const inline = mdToInlineNodes(opts.text ?? "");
+    // footnoteContentKey only reads `.content`, so key off the inline array
+    // directly instead of building a throwaway definition node.
+    const key = footnoteContentKey({ content: inline });
+    // Content dedup: reuse an existing definition's id when its key matches.
+    let footnoteId = null;
+    let reused = false;
+    if (key !== "") {
+        walk(doc, (n) => {
+            if (footnoteId == null &&
+                isObject(n) &&
+                n.type === "footnoteDefinition" &&
+                n.attrs &&
+                typeof n.attrs.id === "string" &&
+                n.attrs.id !== "" &&
+                footnoteContentKey(n) === key) {
+                footnoteId = n.attrs.id;
+                reused = true;
+            }
+        });
+    }
+    if (footnoteId == null)
+        footnoteId = generateFootnoteId();
+    // Insert the footnoteReference node directly after the anchor (mark-safe
+    // split); it hugs the preceding word with no leading space. Two guards keep the
+    // inline atom out of the notes section and out of blocks that cannot hold it:
+    //  - beforeBlock bounds the search to the BODY, before the first top-level block
+    //    that IS or CONTAINS (at any depth) a footnotesList/footnoteDefinition — so
+    //    a NESTED list or a bare definition also bounds the search, not just a
+    //    top-level list;
+    //  - skipSubtreeTypes refuses to descend into any footnotesList/footnoteDefinition
+    //    subtree, so a reference is never glued inside an existing definition (which
+    //    the canonicalizer would then drop as an orphan, losing that definition's
+    //    prose); and forbidBlockTypes refuses codeBlocks (an inline atom there is a
+    //    schema-invalid doc; insert_footnote skips validateDocStructure).
+    // When the only anchor match is in such a place, the insert is refused and the
+    // write aborts cleanly (inserted:false) instead of destroying content.
+    const boundaryIdx = Array.isArray(doc?.content)
+        ? doc.content.findIndex((n) => containsFootnoteNotes(n))
+        : -1;
+    const r = insertNodesAfterAnchor(doc, (opts.anchorText ?? "").trimEnd(), () => [{ type: "footnoteReference", attrs: { id: footnoteId } }], {
+        ...(boundaryIdx >= 0 ? { beforeBlock: boundaryIdx } : {}),
+        forbidBlockTypes: INLINE_ATOM_FORBIDDEN_BLOCKS,
+        skipSubtreeTypes: FOOTNOTE_NOTES_SUBTREES,
+    });
+    if (!r.inserted) {
+        return { doc: clone(doc), inserted: false, footnoteId, reused };
+    }
+    let working = r.doc;
+    // Add a NEW definition (canonicalize will order/place it); a reused id needs
+    // no new definition (the existing one is shared).
+    if (!reused) {
+        appendDefinition(working, makeFootnoteDefinition(footnoteId, inline));
+    }
+    // Derive numbering + the single bottom list deterministically.
+    working = canonicalizeFootnotes(working);
+    return { doc: working, inserted: true, footnoteId, reused };
+}
+/**
+ * Append a definition node so the canonicalizer can order/place it: into the
+ * first existing footnotesList, or a new trailing list when none exists.
+ */
+function appendDefinition(doc, defNode) {
+    const existingList = getList(doc, (n) => isObject(n) && n.type === "footnotesList");
+    if (existingList && Array.isArray(existingList.content)) {
+        existingList.content.push(defNode);
+        return;
+    }
+    if (Array.isArray(doc.content)) {
+        doc.content.push({ type: "footnotesList", content: [defNode] });
+    }
+}

packages/mcp/src/client.ts

@@ -17,6 +17,7 @@ import {
   updatePageContentRealtime,
   replacePageContent,
   markdownToProseMirror,
+  markdownToProseMirrorCanonical,
   mutatePageContent,
   buildCollabWsUrl,
   assertYjsEncodable,
@@ -60,6 +61,8 @@ import {
   noteItem,
   mdToInlineNodes,
   commentsToFootnotes,
+  canonicalizeFootnotes,
+  insertInlineFootnote,
 } from "./lib/transforms.js";
 import vm from "node:vm";
 
@@ -1344,10 +1347,16 @@ export class DocmostClient {
     // inject javascript:/data: link hrefs or media srcs straight into the doc.
     this.validateDocUrls(doc);
 
+    // Canonicalize footnotes (idempotent): an agent-authored JSON doc cannot
+    // leave footnotes out of order, orphaned, or in multiple lists — the bottom
+    // list + numbering are always derived from reference order. No-op when the
+    // footnotes are already canonical.
+    doc = canonicalizeFootnotes(doc);
+
     // Write the BODY first, then the title (#159 split-brain): a failed body
     // write (e.g. persist timeout) must not leave a new title over the old body.
     const collabToken = await this.getCollabTokenWithReauth();
-    const mutation = await replacePageContent(
+    const mutation = await this.replacePage(
       pageId,
       doc,
       collabToken,
@@ -1368,6 +1377,95 @@ export class DocmostClient {
     };
   }
 
+  /**
+   * AUTHOR-INLINE footnote insertion. The agent supplies only WHERE
+   * (`anchorText`, a snippet of body text to attach the marker after) and WHAT
+   * (`text`, the footnote content as markdown). Numbering and the bottom
+   * `footnotesList` are derived deterministically server-side
+   * (`insertInlineFootnote` -> `canonicalizeFootnotes`): the agent never sees,
+   * assigns, or edits a footnote number or the list, so it CANNOT desync.
+   *
+   * Content DEDUP: when an existing definition has the same content, its id is
+   * reused (one number, one definition, several references). The write is atomic
+   * via `mutatePageContent` (single-writer, page-locked); if the anchor text is
+   * not found the transform aborts with a clear error and no write happens.
+   */
+  async insertFootnote(pageId: string, anchorText: string, text: string) {
+    await this.ensureAuthenticated();
+    if (!anchorText || !anchorText.trim()) {
+      throw new Error("insert_footnote: anchorText is required");
+    }
+    if (text == null || `${text}`.trim() === "") {
+      throw new Error("insert_footnote: text is required");
+    }
+    const collabToken = await this.getCollabTokenWithReauth();
+    let result: { footnoteId: string; reused: boolean } | null = null;
+    const mutation = await this.mutatePage(
+      pageId,
+      collabToken,
+      this.apiUrl,
+      (liveDoc: any) => {
+        const r = insertInlineFootnote(liveDoc, { anchorText, text });
+        if (!r.inserted) {
+          // Abort the page-locked write by throwing: mutatePageContent does not
+          // persist when the transform throws, so a missing anchor leaves the
+          // page untouched (no partial write).
+          throw new Error(
+            `insert_footnote: anchor text not found: ${JSON.stringify(
+              anchorText.slice(0, 80),
+            )}`,
+          );
+        }
+        result = { footnoteId: r.footnoteId, reused: r.reused };
+        return r.doc;
+      },
+    );
+    // The not-found path throws inside the transform (aborting mutatePage), so by
+    // here `result` is always set.
+    const r = result!;
+    return {
+      success: true,
+      modified: true,
+      pageId,
+      footnoteId: r.footnoteId,
+      reused: r.reused,
+      message: r.reused
+        ? "Footnote inserted (reused an existing same-content definition)."
+        : "Footnote inserted.",
+      verify: mutation.verify,
+    };
+  }
+
+  /**
+   * Page-locked write seam over collaboration.mutatePageContent. Production just
+   * delegates; it exists as an overridable method so the insert_footnote wrapper
+   * (transform abort-on-not-found + response shaping) can be unit-tested without
+   * standing up a live Hocuspocus collab socket.
+   */
+  protected mutatePage(
+    pageId: string,
+    collabToken: string,
+    apiUrl: string,
+    transform: (doc: any) => any,
+  ): Promise<{ doc?: any; verify?: any }> {
+    return mutatePageContent(pageId, collabToken, apiUrl, transform);
+  }
+
+  /**
+   * Full-document write seam over collaboration.replacePageContent. Production
+   * just delegates; it exists as an overridable method so the full-doc write
+   * tools (update_page_json, copy_page_content) can have their footnote-
+   * canonicalization binding unit-tested without a live Hocuspocus collab socket.
+   */
+  protected replacePage(
+    pageId: string,
+    doc: any,
+    collabToken: string,
+    apiUrl: string,
+  ): Promise<{ doc?: any; verify?: any }> {
+    return replacePageContent(pageId, doc, collabToken, apiUrl);
+  }
+
   /**
    * Export a page to a single self-contained Docmost-flavoured markdown file:
    * meta block + body (with inline comment anchors + diagrams) + comment
@@ -1408,7 +1506,8 @@ export class DocmostClient {
   async importPageMarkdown(pageId: string, fullMarkdown: string): Promise<any> {
     await this.ensureAuthenticated();
     const { meta, body, comments } = parseDocmostMarkdown(fullMarkdown);
-    const doc = await markdownToProseMirror(body);
+    // PAGE import: canonicalize footnotes (see markdownToProseMirrorCanonical).
+    const doc = await markdownToProseMirrorCanonical(body);
     const collabToken = await this.getCollabTokenWithReauth();
     const mutation = await replacePageContent(
       pageId,
@@ -1503,10 +1602,16 @@ export class DocmostClient {
     // (parity with updatePageJson; harmless for already-stored source content).
     this.validateDocUrls(content);
 
+    // Defense-in-depth (#228): this is a FULL-document write, so canonicalize
+    // footnotes before copying — a no-op on already-canonical source content, but
+    // it guarantees a copy can never propagate a non-canonical footnote topology
+    // to the target (parity with the other full-doc write paths).
+    const canonical = canonicalizeFootnotes(content);
+
     const collabToken = await this.getCollabTokenWithReauth();
-    const mutation = await replacePageContent(
+    const mutation = await this.replacePage(
       targetPageId,
-      content,
+      canonical,
       collabToken,
       this.apiUrl,
     );
@@ -1515,7 +1620,7 @@ export class DocmostClient {
       success: true,
       sourcePageId,
       targetPageId,
-      copiedNodes: content.content.length,
+      copiedNodes: canonical.content.length,
       verify: mutation.verify,
     };
   }
@@ -2033,7 +2138,10 @@ export class DocmostClient {
       }
     }
 
-    // Convert through the full Docmost schema (consistent with page paths)
+    // Convert through the full Docmost schema. Deliberately the NON-canonicalizing
+    // variant: a comment body may carry a footnote definition with no matching
+    // reference, and canonicalization would drop it (data loss). See
+    // markdownToProseMirror vs markdownToProseMirrorCanonical.
     const jsonContent = await markdownToProseMirror(content);
     const payload: Record<string, any> = {
       pageId,
@@ -2136,6 +2244,7 @@ export class DocmostClient {
 
   async updateComment(commentId: string, content: string) {
     await this.ensureAuthenticated();
+    // NON-canonicalizing on purpose (comment body — see createComment).
     const jsonContent = await markdownToProseMirror(content);
     await this.client.post("/comments/update", {
       commentId,
@@ -2986,6 +3095,8 @@ export class DocmostClient {
         noteItem,
         mdToInlineNodes,
         commentsToFootnotes,
+        canonicalizeFootnotes,
+        insertInlineFootnote,
       },
     };
 
@@ -3022,24 +3133,33 @@ export class DocmostClient {
           "transform must evaluate to a function (doc, ctx) => doc",
         );
       }
-      const result = vm.runInNewContext(
+      const raw = vm.runInNewContext(
         "f(d, c)",
         { f: fn, d: sandbox.doc, c: ctx },
         { timeout: 5000 },
       );
       if (
-        !result ||
-        typeof result !== "object" ||
-        result.type !== "doc" ||
-        !Array.isArray(result.content)
+        !raw ||
+        typeof raw !== "object" ||
+        raw.type !== "doc" ||
+        !Array.isArray(raw.content)
       ) {
         throw new Error(
           'transform must return a ProseMirror doc node ({ type:"doc", content:[...] })',
         );
       }
-      // Validate the returned doc before it can be written.
-      this.validateDocStructure(result);
-      this.validateDocUrls(result);
+      // Validate the RAW transform output FIRST (structure — including the
+      // MAX_DEPTH guard — and URLs), mirroring updatePageJson. The canonicalizer
+      // recurses without a depth limiter, so validating after it would turn a
+      // too-deep doc into an opaque "Maximum call stack size exceeded" instead of
+      // the intended "nesting exceeds the maximum depth" error.
+      this.validateDocStructure(raw);
+      this.validateDocUrls(raw);
+      // Auto-canonicalize footnotes after the transform (idempotent): no write
+      // path can leave footnotes out of order / orphaned / in a raw `[^id]`
+      // block. In a dryRun preview this may surface footnote edits the script
+      // author did not write (the canonicalizer tidied them) — that is expected.
+      const result = canonicalizeFootnotes(raw);
       newDoc = result;
       return result;
     };

packages/mcp/src/index.ts

@@ -892,8 +892,15 @@ server.registerTool(
       "mark-safe), setCalloutRange(doc, n) (sync a [1]…[K] callout range to " +
       "[1]…[n]), noteItem(inlineNodes) (wrap inline nodes in a listItem with a " +
       "fresh id), mdToInlineNodes(markdown) (comment markdown -> inline nodes), " +
-      "and commentsToFootnotes(doc, comments, {notesHeading}) (turn inline " +
-      "comments into numbered footnotes). Footnote convention: markers are " +
+      "commentsToFootnotes(doc, comments, {notesHeading}) (turn inline " +
+      "comments into numbered footnotes), canonicalizeFootnotes(doc) (derive " +
+      "footnote numbering + the single bottom list from reference order, drop " +
+      "orphans/duplicates — runs AUTOMATICALLY on the transform RESULT, so the " +
+      "applied (and dryRun-previewed) doc is always footnote-canonical; a dryRun " +
+      "diff may therefore show footnote tidy-ups your script did not make, and " +
+      "it is idempotent after the first run), and " +
+      "insertInlineFootnote(doc, {anchorText, text}) (author-inline footnote: " +
+      "marker + dedup'd definition, list derived). Footnote convention: markers are " +
       "plain '[N]' text in the body; the notes are an orderedList under a " +
       "heading whose text is 'Примечания переводчика'. The transform runs " +
       "sandboxed (no require/process/fs/network, 5s timeout) and must return a " +
@@ -908,7 +915,8 @@ server.registerTool(
             "parenthesized function). It receives a clone of the live doc and " +
             "ctx (comments, log, consume(id), helpers: blockText/walk/getList/" +
             "insertMarkerAfter/setCalloutRange/noteItem/mdToInlineNodes/" +
-            "commentsToFootnotes) and must return a {type:'doc'} node.",
+            "commentsToFootnotes/canonicalizeFootnotes/insertInlineFootnote) " +
+            "and must return a {type:'doc'} node.",
         ),
       dryRun: z
         .boolean()
@@ -934,6 +942,41 @@ server.registerTool(
   },
 );
 
+// Tool: insert_footnote
+server.registerTool(
+  "insert_footnote",
+  {
+    description:
+      "Insert an AUTHOR-INLINE footnote: you specify only WHERE (anchorText) " +
+      "and WHAT (text). The footnote marker is placed right after anchorText in " +
+      "the body, and the bottom footnotes list + the numbering are derived " +
+      "deterministically server-side. You do NOT assign a number, and you " +
+      "never see or edit the footnotes list — so footnotes cannot end up out " +
+      "of order, orphaned, or as a raw '[^id]' block. If a footnote with the " +
+      "SAME text already exists, its number is REUSED (one definition, several " +
+      "references). The write is atomic and won't clobber concurrent edits; if " +
+      "anchorText is not found, nothing is written and an error is returned.",
+    inputSchema: {
+      pageId: z.string().min(1),
+      anchorText: z
+        .string()
+        .min(1)
+        .describe(
+          "A snippet of existing body text; the footnote marker is inserted " +
+            "immediately after its first occurrence (mark-safe).",
+        ),
+      text: z
+        .string()
+        .min(1)
+        .describe("The footnote content as markdown (becomes the definition)."),
+    },
+  },
+  async ({ pageId, anchorText, text }) => {
+    const result = await docmostClient.insertFootnote(pageId, anchorText, text);
+    return jsonContent(result);
+  },
+);
+
 // Tool: diff_page_versions
 registerShared(
   SHARED_TOOL_SPECS.diffPageVersions,

packages/mcp/src/lib/collaboration.ts

@@ -11,6 +11,7 @@ import { docmostExtensions, docmostSchema } from "./docmost-schema.js";
 import { withPageLock } from "./page-lock.js";
 import { sanitizeForYjs, findUnstorableAttr } from "./node-ops.js";
 import { lexFootnoteLines } from "./footnote-lex.js";
+import { canonicalizeFootnotes } from "./footnote-canonicalize.js";
 import { summarizeChange, VerifyReport } from "./diff.js";
 
 /**
@@ -392,7 +393,20 @@ function extractFootnotes(markdown: string): {
   };
 }
 
-/** Convert markdown to a ProseMirror doc using the full Docmost schema. */
+/**
+ * Convert markdown to a ProseMirror doc using the full Docmost schema.
+ *
+ * This conversion does NOT canonicalize footnotes — it is the shared, content-
+ * preserving primitive used by BOTH page write paths and COMMENT bodies
+ * (createComment / updateComment). Canonicalization MUST NOT run on a comment
+ * body: a comment may legitimately contain a footnote-definition line
+ * (`[^1]: text`) with no matching reference, and the canonicalizer drops a
+ * reference-less footnotesList — which would silently delete the comment's text.
+ *
+ * Page write paths that DO need the canonical footnote topology call
+ * `markdownToProseMirrorCanonical` instead (markdown import, update_page markdown
+ * path). Keep this function reference-loss-free.
+ */
 export async function markdownToProseMirror(
   markdownContent: string,
 ): Promise<any> {
@@ -403,6 +417,23 @@ export async function markdownToProseMirror(
   return generateJSON(bridged, docmostExtensions);
 }
 
+/**
+ * Page-write variant of `markdownToProseMirror`: converts markdown then enforces
+ * the canonical footnote topology. The footnote `section` markdown is emitted in
+ * DEFINITION order, but numbering derives from REFERENCE order, so without this
+ * the bottom list renders out of order (`1, 4, 2, 3, …`); orphan definitions and
+ * duplicate lists are also normalized. Idempotent — a no-op once canonical, and a
+ * no-op for footnote-free content.
+ *
+ * Use this ONLY for full-document PAGE writes (never for comment bodies, where it
+ * would drop a reference-less footnote definition — see `markdownToProseMirror`).
+ */
+export async function markdownToProseMirrorCanonical(
+  markdownContent: string,
+): Promise<any> {
+  return canonicalizeFootnotes(await markdownToProseMirror(markdownContent));
+}
+
 /**
  * Build the collaboration WebSocket URL from an API base URL:
  * switch http(s)->ws(s), strip a trailing /api, mount on /collab.
@@ -801,7 +832,9 @@ export async function updatePageContentRealtime(
   collabToken: string,
   baseUrl: string,
 ): Promise<MutationResult> {
-  const tiptapJson = await markdownToProseMirror(markdownContent);
+  // PAGE write: canonicalize footnotes (markdown import builds the bottom list in
+  // definition order; numbering is reference-ordered).
+  const tiptapJson = await markdownToProseMirrorCanonical(markdownContent);
   return await mutatePageContent(
     pageId,
     collabToken,

packages/mcp/src/lib/footnote-authoring.ts

@@ -0,0 +1,91 @@
+/**
+ * Inline-authoring helpers for footnotes (MCP).
+ *
+ * These build/identify footnote DEFINITION nodes for the author-inline tool
+ * (`insertInlineFootnote` in transforms.ts): a content key to de-duplicate notes
+ * by text, a definition-node factory, and a fresh uuidv7-style id generator.
+ *
+ * Split out of `footnote-canonicalize.ts` so that module stays a pure MIRROR of
+ * the editor-ext canonicalizer (compositionally symmetric to the editor-ext
+ * copy, which keeps its authoring helpers in `footnote-util.ts`). The pure
+ * canonicalizer has no dependency on these.
+ */
+
+const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
+
+function cloneJson<T>(v: T): T {
+  if (typeof structuredClone === "function") return structuredClone(v);
+  return JSON.parse(JSON.stringify(v)) as T;
+}
+
+/**
+ * Normalized content key for de-duplicating footnote DEFINITIONS by their text.
+ *
+ * Two definitions with the same key are the SAME footnote — so the inline
+ * authoring tool reuses one id (one number, one definition, several references)
+ * instead of minting a second definition. Key = plaintext (whitespace-collapsed,
+ * trimmed) PLUS a signature of the inline mark types in order, so two notes that
+ * read the same but differ in formatting (one bold, one plain) are NOT merged.
+ * Conservative: only an exact match merges.
+ */
+export function footnoteContentKey(defNode: any): string {
+  const parts: string[] = [];
+  const visit = (n: any): void => {
+    if (!n || typeof n !== "object") return;
+    if (n.type === "text" && typeof n.text === "string") {
+      const marks = Array.isArray(n.marks)
+        ? n.marks.map((m: any) => m?.type).filter(Boolean).sort().join(",")
+        : "";
+      parts.push(`${n.text}${marks}`);
+    }
+    if (Array.isArray(n.content)) for (const c of n.content) visit(c);
+  };
+  visit(defNode);
+  // Collapse the assembled text's whitespace and trim, keeping the mark
+  // signature attached so formatting differences still distinguish notes.
+  return parts
+    .join("")
+    .replace(/[ \t\r\n]+/g, " ")
+    .trim();
+}
+
+/**
+ * Build a footnoteDefinition node from inline ProseMirror nodes, keyed by id.
+ */
+export function makeFootnoteDefinition(id: string, inlineNodes: any[]): any {
+  const content = Array.isArray(inlineNodes) ? cloneJson(inlineNodes) : [];
+  return {
+    type: FOOTNOTE_DEFINITION_NAME,
+    attrs: { id },
+    content: [{ type: "paragraph", content }],
+  };
+}
+
+/**
+ * Generate a uuidv7-style id (time-ordered), matching editor-ext's
+ * `generateFootnoteId`. Used for a genuinely-new inline footnote id.
+ */
+export function generateFootnoteId(): string {
+  const now = Date.now();
+  const timeHex = now.toString(16).padStart(12, "0");
+  const rand = (length: number) => {
+    let s = "";
+    for (let i = 0; i < length; i++)
+      s += Math.floor(Math.random() * 16).toString(16);
+    return s;
+  };
+  const versioned = "7" + rand(3);
+  const variantNibble = (8 + Math.floor(Math.random() * 4)).toString(16);
+  const variant = variantNibble + rand(3);
+  return (
+    timeHex.slice(0, 8) +
+    "-" +
+    timeHex.slice(8, 12) +
+    "-" +
+    versioned +
+    "-" +
+    variant +
+    "-" +
+    rand(12)
+  );
+}

packages/mcp/src/lib/footnote-canonicalize.ts

@@ -0,0 +1,225 @@
+/**
+ * Server-side footnote canonicalizer (MCP mirror — PURE).
+ *
+ * `canonicalizeFootnotes(doc)` is a pure ProseMirror-JSON port of the editor's
+ * `footnoteSyncPlugin` end-state, identical in behaviour to
+ * `@docmost/editor-ext`'s `canonicalizeFootnotes`. It is mirrored here — rather
+ * than imported from editor-ext — for the SAME reason `footnote-lex.ts` and the
+ * `docmost-schema.ts` nodes are mirrored: the MCP package is deliberately
+ * decoupled from the browser/React-heavy editor barrel and operates on plain
+ * JSON. The editor-ext copy owns the golden test against the live plugin; this
+ * copy must stay behaviourally identical (a SHARED golden corpus, exercised by
+ * both test suites, pins that — see `test/unit/footnote-corpus.mjs`).
+ *
+ * This module is the pure MIRROR only. The inline-authoring helpers
+ * (`footnoteContentKey`, `makeFootnoteDefinition`, `generateFootnoteId`) used by
+ * `insertInlineFootnote` live in the sibling `footnote-authoring.ts`, so this
+ * file is compositionally symmetric to the editor-ext copy.
+ *
+ * Why it exists: every NON-editor write path (markdown import, update_page_json,
+ * docmost_transform, insert_footnote) builds ProseMirror JSON directly, so the
+ * editor's footnote plugins never run and the canonical topology (sequential
+ * numbering by first reference, one trailing list, no orphans, no raw `[^id]`)
+ * was never enforced. Running this at the end of every write path closes that
+ * gap; because it is idempotent, it is a no-op when the footnotes are already
+ * canonical (no spurious mutations / git-sync churn).
+ *
+ * ENFORCEMENT RULE (#228): any NEW FULL-document persist path MUST call
+ * `canonicalizeFootnotes(doc)` before writing — the current callers are
+ * `markdownToProseMirrorCanonical` (page markdown import/update; the plain
+ * `markdownToProseMirror` used for COMMENT bodies must NOT, or it would drop a
+ * reference-less definition), `update_page_json`, `docmost_transform`,
+ * `insert_footnote`, and `copy_page_content`. Append/prepend FRAGMENT writes MUST
+ * NOT canonicalize. This is deliberately per-call-site (the replace-vs-fragment
+ * and comment-vs-page nuances make a single naive wrapper unsafe).
+ */
+
+const FOOTNOTE_REFERENCE_NAME = "footnoteReference";
+const FOOTNOTES_LIST_NAME = "footnotesList";
+const FOOTNOTE_DEFINITION_NAME = "footnoteDefinition";
+
+function cloneJson<T>(v: T): T {
+  if (typeof structuredClone === "function") return structuredClone(v);
+  return JSON.parse(JSON.stringify(v)) as T;
+}
+
+function isEmptyParagraph(node: any): boolean {
+  return (
+    !!node &&
+    node.type === "paragraph" &&
+    (!Array.isArray(node.content) || node.content.length === 0)
+  );
+}
+
+function collectReferenceIds(node: any, out: string[], seen: Set<string>): void {
+  if (!node || typeof node !== "object") return;
+  if (node.type === FOOTNOTE_REFERENCE_NAME) {
+    const id = node?.attrs?.id;
+    if (id && !seen.has(id)) {
+      seen.add(id);
+      out.push(id);
+    }
+  }
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) collectReferenceIds(child, out, seen);
+  }
+}
+
+function collectDefinitions(node: any, out: any[]): void {
+  if (!node || typeof node !== "object") return;
+  if (node.type === FOOTNOTE_DEFINITION_NAME) out.push(node);
+  if (Array.isArray(node.content)) {
+    for (const child of node.content) collectDefinitions(child, out);
+  }
+}
+
+function emptyDefinition(id: string): any {
+  return {
+    type: FOOTNOTE_DEFINITION_NAME,
+    attrs: { id },
+    content: [{ type: "paragraph" }],
+  };
+}
+
+/**
+ * Deep equality over plain JSON: arrays are compared POSITIONALLY
+ * (order-SENSITIVE), object keys order-insensitively. The array order-sensitivity
+ * is required for correctness here — a reordered `footnotesList.content` must
+ * compare UNEQUAL so the canonical rebuild fires instead of leaving it in place.
+ */
+function deepEqualJson(a: any, b: any): boolean {
+  if (a === b) return true;
+  if (a == null || b == null || typeof a !== typeof b) return false;
+  if (Array.isArray(a) || Array.isArray(b)) {
+    if (!Array.isArray(a) || !Array.isArray(b) || a.length !== b.length) {
+      return false;
+    }
+    for (let i = 0; i < a.length; i++) {
+      if (!deepEqualJson(a[i], b[i])) return false;
+    }
+    return true;
+  }
+  if (typeof a === "object") {
+    const ka = Object.keys(a);
+    const kb = Object.keys(b);
+    if (ka.length !== kb.length) return false;
+    for (const k of ka) {
+      if (!Object.prototype.hasOwnProperty.call(b, k)) return false;
+      if (!deepEqualJson(a[k], b[k])) return false;
+    }
+    return true;
+  }
+  return false;
+}
+
+/**
+ * Canonicalize footnotes in a ProseMirror-JSON document. See the file header and
+ * the editor-ext twin for the full contract. Pure (deep-clones input,
+ * deterministic, idempotent).
+ */
+export function canonicalizeFootnotes<T = any>(doc: T): T {
+  if (
+    doc == null ||
+    typeof doc !== "object" ||
+    !Array.isArray((doc as any).content)
+  ) {
+    return doc;
+  }
+  const out = cloneJson(doc) as any;
+
+  // 1) Distinct reference ids in document order (deep — refs can live in
+  //    callouts, tables, list items, ...). The ordering/numbering truth.
+  const referenceIds: string[] = [];
+  collectReferenceIds(out, referenceIds, new Set<string>());
+
+  // 2) Every definition node in document order (deep).
+  const defNodes: any[] = [];
+  collectDefinitions(out, defNodes);
+
+  // 3) First definition per id wins; later duplicates carry the SAME id, so they
+  //    cannot be referenced separately and would be orphans — they are dropped.
+  const defById = new Map<string, any>();
+  for (const d of defNodes) {
+    const id = d?.attrs?.id;
+    if (id && !defById.has(id)) defById.set(id, d);
+  }
+
+  // 4) Build the ordered definition list: one per referenced id, in REFERENCE
+  //    order, reusing the existing node (shallow-copied, id normalized — `out` is
+  //    already deep-cloned and the old lists are cut) or synthesizing an empty
+  //    one. Definitions whose id is not referenced are orphans and never added.
+  const orderedDefs: any[] = [];
+  for (const id of referenceIds) {
+    const existing = defById.get(id);
+    if (existing) {
+      orderedDefs.push({
+        ...existing,
+        attrs: { ...(existing.attrs ?? {}), id },
+      });
+    } else {
+      orderedDefs.push(emptyDefinition(id));
+    }
+  }
+
+  // 5) No references -> there must be NO list at all (at any depth).
+  if (referenceIds.length === 0) {
+    stripFootnotesListsDeep(out);
+    return out;
+  }
+
+  // 6) Placement parity with the live plugin: when the document is ALREADY in the
+  //    canonical single-list state, leave that list exactly where it sits rather
+  //    than cutting and re-inserting it at the end (the plugin never repositions a
+  //    sole correct list, so moving it would silently reorder any content that
+  //    follows the list on the first write).
+  const topLevelLists = out.content.filter(
+    (n: any) => n && n.type === FOOTNOTES_LIST_NAME,
+  );
+  if (
+    topLevelLists.length === 1 &&
+    defNodes.length === orderedDefs.length &&
+    deepEqualJson(topLevelLists[0].content, orderedDefs)
+  ) {
+    return out;
+  }
+
+  // 7) Otherwise rebuild: strip every footnotesList AND every bare
+  //    footnoteDefinition at ANY depth (collectDefinitions gathers defs
+  //    recursively, so a list nested in a callout/blockquote — or a bare
+  //    definition outside any list — would otherwise have its defs copied into the
+  //    rebuilt list while the original survives in place → duplicates) and
+  //    re-insert exactly one list after the last meaningful (non-empty paragraph)
+  //    top-level block.
+  stripFootnotesListsDeep(out);
+  stripFootnoteDefinitionsDeep(out);
+  const top: any[] = out.content;
+  let insertAt = top.length;
+  while (insertAt > 0 && isEmptyParagraph(top[insertAt - 1])) insertAt--;
+  top.splice(insertAt, 0, { type: FOOTNOTES_LIST_NAME, content: orderedDefs });
+  out.content = top;
+  return out;
+}
+
+/** Remove every `footnotesList` node at ANY depth (mutates the given clone). */
+function stripFootnotesListsDeep(node: any): void {
+  if (!node || typeof node !== "object" || !Array.isArray(node.content)) return;
+  node.content = node.content.filter(
+    (c: any) => !(c && c.type === FOOTNOTES_LIST_NAME),
+  );
+  for (const child of node.content) stripFootnotesListsDeep(child);
+}
+
+/**
+ * Remove every BARE `footnoteDefinition` node at ANY depth (mutates the given
+ * clone). Runs only in the rebuild path AFTER the lists are stripped, so it
+ * targets definitions that were sitting outside a list (e.g. hand-authored via a
+ * raw-JSON write path and nested in a callout); their content was already copied
+ * into the rebuilt list, so leaving the originals would duplicate them.
+ */
+function stripFootnoteDefinitionsDeep(node: any): void {
+  if (!node || typeof node !== "object" || !Array.isArray(node.content)) return;
+  node.content = node.content.filter(
+    (c: any) => !(c && c.type === FOOTNOTE_DEFINITION_NAME),
+  );
+  for (const child of node.content) stripFootnoteDefinitionsDeep(child);
+}