Merge pull request 'Батч: бейдж контекста (#189) + e2e в CI (#187) + inline-тест MCP (#170)' (#197) from batch/issues-189-187-170 into develop

Reviewed-on: #197

.github/workflows/develop.yml

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

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

@@ -715,6 +715,8 @@
   "Test": "Test",
   "Available tools": "Available tools",
   "No tools available": "No tools available",
+  "Failed": "Failed",
+  "OK · {{n}}": "OK · {{n}}",
   "Created successfully": "Created successfully",
   "Deleted successfully": "Deleted successfully",
   "Clear": "Clear",
@@ -1167,8 +1169,9 @@
   "Pick an agent role whose persona the public assistant adopts. The safety rules always still apply.": "Pick an agent role whose persona the public assistant adopts. The safety rules always still apply.",
   "Built-in assistant persona": "Built-in assistant persona",
   "Minimize": "Minimize",
-  "Current context size": "Current context size",
+  "Context size / model limit": "Context size / model limit",
-  "Tokens generated this turn": "Tokens generated this turn",
+  "Context window (tokens)": "Context window (tokens)",
+  "Shown as used / total in the chat header. Leave empty to hide the limit.": "Shown as used / total in the chat header. Leave empty to hide the limit.",
   "AI agent": "AI agent",
   "Take a look at the current document": "Take a look at the current document",
   "AI agent is typing…": "AI agent is typing…",

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

@@ -704,13 +704,19 @@
   "Ask the AI agent…": "Спросите AI-агента…",
   "Copy chat": "Копировать чат",
   "Created successfully": "Успешно создано",
-  "Current context size": "Текущий размер контекста",
+  "Context size / model limit": "Размер контекста / лимит модели",
-  "Tokens generated this turn": "Токенов сгенерировано за ход",
+  "Context window (tokens)": "Окно контекста (токены)",
+  "Shown as used / total in the chat header. Leave empty to hide the limit.": "Показывается в шапке чата как использовано / всего. Пусто — лимит скрыт.",
   "Delete this chat?": "Удалить этот чат?",
   "Deleted successfully": "Успешно удалено",
   "Edited by AI agent on behalf of {{name}}": "Отредактировано AI-агентом от имени {{name}}",
   "Failed to delete chat": "Не удалось удалить чат",
   "Failed to rename chat": "Не удалось переименовать чат",
+  "Failed": "Ошибка",
+  "OK · {{n}}": "OK · {{n}}",
+  "Test": "Тест",
+  "No tools available": "Инструменты недоступны",
+  "Available tools": "Доступные инструменты",
   "Minimize": "Свернуть",
   "No chats yet.": "Чатов пока нет.",
   "Send": "Отправить",

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

@@ -45,6 +45,7 @@ import {
   shouldCollapseOnOutsidePointer,
   isHeaderClick,
 } from "@/features/ai-chat/utils/collapse-helpers.ts";
+import { selectContextBadge } from "@/features/ai-chat/utils/context-badge.ts";
 import { useClipboard } from "@/hooks/use-clipboard";
 import { notifications } from "@mantine/notifications";
 import classes from "@/features/ai-chat/components/ai-chat-window.module.css";
@@ -161,12 +162,6 @@ export default function AiChatWindow() {
   const { data: messageRows, isLoading: messagesLoading } =
     useAiChatMessagesQuery(activeChatId ?? undefined);
 
-  // Live turn-token total (reasoning + output) for the in-flight turn, pushed up
-  // (THROTTLED to ~8 Hz inside ChatThread) so the header badge ticks mid-stream.
-  // `null` means no turn is in flight -> the badge falls back to the persisted
-  // context size below.
-  const [liveTurnTokens, setLiveTurnTokens] = useState<number | null>(null);
-
   // The page the user is currently viewing. AiChatWindow lives in a pathless
   // parent layout route, so useParams() can't see :pageSlug. Match the full
   // pathname against the authenticated page route instead so "the current page"
@@ -301,24 +296,19 @@ export default function AiChatWindow() {
   // shipped; older rows fall back to that turn's `usage` total. NOTE: reflects
   // PERSISTED rows (updates on chat open/switch); it does not tick live
   // mid-stream — acceptable for v1.
-  const contextTokens = useMemo(() => {
+  //
-    if (!activeChatId || !messageRows) return 0;
+  // The denominator `maxContextTokens` (the model's configured max window) is
-    for (let i = messageRows.length - 1; i >= 0; i--) {
+  // derived in the SAME backward scan: it is stamped alongside `contextTokens`
-      const meta = messageRows[i].metadata;
+  // on a completed turn, but the numerator and denominator are taken from the
-      if (!meta) continue;
+  // most recent row carrying EACH value independently — they may land on
-      if (typeof meta.contextTokens === "number" && meta.contextTokens > 0) {
+  // different rows (e.g. a fresh error row can carry contextTokens but not
-        return meta.contextTokens;
+  // maxContextTokens), so we keep scanning for whichever is still unset. 0 when
-      }
+  // no row has it (older rows, or no admin-configured limit) — the badge then
-      const usage = meta.usage;
+  // shows just the current size with no denominator.
-      if (usage) {
+  const { contextTokens, maxContextTokens } = useMemo(
-        const fallback =
+    () => selectContextBadge(activeChatId ? messageRows : undefined),
-          usage.totalTokens ??
+    [activeChatId, messageRows],
-          (usage.inputTokens ?? 0) + (usage.outputTokens ?? 0);
+  );
-        if (fallback > 0) return fallback;
-      }
-    }
-    return 0;
-  }, [activeChatId, messageRows]);
 
   // On (re)open, settle the geometry before paint (useLayoutEffect → no
   // first-frame jump): compute an initial top-right placement the first time,
@@ -509,20 +499,17 @@ export default function AiChatWindow() {
         )}
 
         <div style={{ flex: 1, display: "flex", justifyContent: "center" }}>
-          {/* While a turn streams, show the LIVE turn-token count (ticks ~8 Hz);
+          {/* Always show the persisted "current / max" context. The denominator
-              once it finishes, fall back to the persisted context size. Require
+              (the admin-configured model limit) is appended only when known;
-              > 0 so the very first emit (an empty tail message, count 0) does not
+              not clamped when current > max (shown as-is, e.g. "210k / 200k").
-              flash a "0" badge before any token streams in (#151 review). */}
+              Hidden entirely until a turn has recorded a context figure. */}
-          {liveTurnTokens !== null && liveTurnTokens > 0 ? (
+          {contextTokens > 0 ? (
-            <Tooltip label={t("Tokens generated this turn")} withArrow>
+            <Tooltip label={t("Context size / model limit")} withArrow>
-              <span className={classes.badge}>
-                {formatTokens(liveTurnTokens)}
-              </span>
-            </Tooltip>
-          ) : contextTokens > 0 ? (
-            <Tooltip label={t("Current context size")} withArrow>
               <span className={classes.badge}>
                 {formatTokens(contextTokens)}
+                {maxContextTokens > 0
+                  ? ` / ${formatTokens(maxContextTokens)}`
+                  : ""}
               </span>
             </Tooltip>
           ) : null}
@@ -649,7 +636,6 @@ export default function AiChatWindow() {
               assistantName={currentRole?.name}
               onTurnFinished={onTurnFinished}
               onServerChatId={onServerChatId}
-              onLiveTurnTokens={setLiveTurnTokens}
             />
           )}
         </div>

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

@@ -20,7 +20,6 @@ import {
 } from "@/features/ai-chat/utils/role-launch.ts";
 import { describeChatError } from "@/features/ai-chat/utils/error-message.ts";
 import { extractServerChatId } from "@/features/ai-chat/utils/adopt-chat-id.ts";
-import { liveTurnTokens } from "@/features/ai-chat/utils/count-stream-tokens.ts";
 import {
   dequeue,
   enqueueMessage,
@@ -82,12 +81,6 @@ interface ChatThreadProps {
    *  Copy/export button available mid-stream). Distinct from onTurnFinished,
    *  which fires only at the terminal outcome. */
   onServerChatId?: (serverChatId?: string) => void;
-  /** Reports the live turn-token total (reasoning + output) for the in-flight
-   *  turn so the parent can show a header badge that ticks mid-stream. THROTTLED
-   *  here (~8 Hz) so the parent re-renders a handful of times a second, not on
-   *  every streamed delta. Called with `null` when no turn is in flight (the
-   *  parent then reverts the badge to the persisted context size). */
-  onLiveTurnTokens?: (tokens: number | null) => void;
 }
 
 /**
@@ -133,7 +126,6 @@ export default function ChatThread({
   assistantName,
   onTurnFinished,
   onServerChatId,
-  onLiveTurnTokens,
 }: ChatThreadProps) {
   const { t } = useTranslation();
 
@@ -348,53 +340,6 @@ export default function ChatThread({
   // the SAME on-screen banner text can be mirrored into the export (issue #160).
   const errorView = error ? describeChatError(error.message ?? "", t) : null;
 
-  // Report the live turn-token total to the parent header badge, THROTTLED to
-  // ~8 Hz so the parent re-renders a few times a second instead of on every
-  // streamed delta. The tail assistant message's reasoning+output (estimate while
-  // streaming, authoritative once a step reports usage) is the live figure. When
-  // the turn ends we emit a final exact value, then `null` so the parent reverts
-  // the badge to the persisted context size.
-  const lastEmitRef = useRef(0);
-  const emitTimerRef = useRef<ReturnType<typeof setTimeout> | null>(null);
-  useEffect(() => {
-    if (!onLiveTurnTokens) return;
-    if (!isStreaming) {
-      // Turn ended (or never started): clear any pending throttle and revert.
-      if (emitTimerRef.current) {
-        clearTimeout(emitTimerRef.current);
-        emitTimerRef.current = null;
-      }
-      lastEmitRef.current = 0;
-      onLiveTurnTokens(null);
-      return;
-    }
-    const tail = messages[messages.length - 1];
-    const live = tail?.role === "assistant" ? liveTurnTokens(tail) : null;
-    const total = live ? live.reasoning + live.output : 0;
-    const now = Date.now();
-    const MIN_INTERVAL = 120; // ms (~8 Hz)
-    const elapsed = now - lastEmitRef.current;
-    if (elapsed >= MIN_INTERVAL) {
-      lastEmitRef.current = now;
-      onLiveTurnTokens(total);
-    } else if (!emitTimerRef.current) {
-      // Schedule a trailing emit so the FINAL value of a burst is not dropped.
-      emitTimerRef.current = setTimeout(() => {
-        emitTimerRef.current = null;
-        lastEmitRef.current = Date.now();
-        onLiveTurnTokens(total);
-      }, MIN_INTERVAL - elapsed);
-    }
-  }, [messages, isStreaming, onLiveTurnTokens]);
-
-  // Clear any pending throttle timer on unmount (chat switch via `key`) so a
-  // trailing emit can't fire into a torn-down thread's parent.
-  useEffect(() => {
-    return () => {
-      if (emitTimerRef.current) clearTimeout(emitTimerRef.current);
-    };
-  }, []);
-
   // A role was picked with autoStart=false: the role is bound but NOTHING was
   // sent, so chatId stays null and the empty state would keep showing the cards.
   // This flag hides the cards and reveals the composer (with the role indicated)

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

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

@@ -275,11 +275,12 @@ describe('flushAssistant', () => {
     expect(f.toolCalls).not.toBeNull();
   });
 
-  it('completed: attaches finishReason + normalized usage + contextTokens', () => {
+  it('completed: attaches finishReason + normalized usage + contextTokens + maxContextTokens', () => {
     const f = flushAssistant([toolStep], '', 'completed', {
       finishReason: 'stop',
       usage: { inputTokens: 10, outputTokens: 5, totalTokens: 15 },
       contextTokens: 15,
+      maxContextTokens: 200000,
     });
     expect(f.status).toBe('completed');
     expect(f.metadata.finishReason).toBe('stop');
@@ -290,6 +291,23 @@ describe('flushAssistant', () => {
       reasoningTokens: undefined,
     });
     expect(f.metadata.contextTokens).toBe(15);
+    expect(f.metadata.maxContextTokens).toBe(200000);
+  });
+
+  it('completed: omits maxContextTokens when unset or 0', () => {
+    // No maxContextTokens in the extra (admin set no context window).
+    const f = flushAssistant([toolStep], '', 'completed', {
+      finishReason: 'stop',
+      contextTokens: 15,
+    });
+    expect('maxContextTokens' in f.metadata).toBe(false);
+    // Explicit 0 is treated the same as unset (no limit -> key omitted).
+    const f0 = flushAssistant([toolStep], '', 'completed', {
+      finishReason: 'stop',
+      contextTokens: 15,
+      maxContextTokens: 0,
+    });
+    expect('maxContextTokens' in f0.metadata).toBe(false);
   });
 
   it('error: records the error and a derived finishReason', () => {

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

@@ -616,6 +616,10 @@ export class AiChatService implements OnModuleInit {
               contextTokens:
                 (usage?.inputTokens ?? 0) + (usage?.outputTokens ?? 0) ||
                 undefined,
+              // Max context window for the chat header badge denominator;
+              // resolved from the admin-configured provider settings (in
+              // closure scope here). Omitted/0 = no limit.
+              maxContextTokens: resolved?.chatContextWindow,
             }),
           );
           // Lifecycle: release the external MCP clients leased for this turn.
@@ -1212,8 +1216,9 @@ export async function applyFinalize(
  * `metadata.parts` is built by assistantParts over the finished steps, then the
  * in-progress text appended as a trailing text part, so rowToUiMessage /
  * findRecent keep replaying the turn unchanged. `metadata.finishReason`,
- * `metadata.error`, `metadata.usage` and `metadata.contextTokens` are attached
+ * `metadata.error`, `metadata.usage`, `metadata.contextTokens` and
- * only when provided/relevant, matching the pre-#183 onFinish/onError records.
+ * `metadata.maxContextTokens` are attached only when provided/relevant, matching
+ * the pre-#183 onFinish/onError records.
  */
 export function flushAssistant(
   capturedSteps: ReadonlyArray<StepLike> | undefined,
@@ -1223,6 +1228,7 @@ export function flushAssistant(
     finishReason?: string;
     usage?: ChatStreamUsage | StreamUsage | undefined;
     contextTokens?: number;
+    maxContextTokens?: number;
     error?: string;
   },
 ): AssistantFlush {
@@ -1253,6 +1259,8 @@ export function flushAssistant(
       normalizeStreamUsage(extra.usage as StreamUsage) ?? extra.usage;
   }
   if (extra?.contextTokens) metadata.contextTokens = extra.contextTokens;
+  if (extra?.maxContextTokens)
+    metadata.maxContextTokens = extra.maxContextTokens;
   if (extra?.error) metadata.error = extra.error;
 
   return {

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

@@ -20,6 +20,7 @@ import { DB, Workspaces } from '@docmost/db/types/db';
 export const AI_PROVIDER_SETTINGS_ALLOWED: readonly string[] = [
   'driver',
   'chatModel',
+  'chatContextWindow',
   'chatApiStyle',
   'embeddingModel',
   'baseUrl',

apps/server/src/integrations/ai/ai-provider-settings-keys.spec.ts

@@ -41,3 +41,35 @@ describe('UpdateAiSettingsDto.chatApiStyle', () => {
     expect(errs.find((e) => e.property === 'chatApiStyle')).toBeUndefined();
   });
 });
+
+/** DTO validation for the new chatContextWindow field (@IsInt @Min(0)). */
+describe('UpdateAiSettingsDto.chatContextWindow', () => {
+  const errorsFor = async (chatContextWindow: unknown) =>
+    validate(plainToInstance(UpdateAiSettingsDto, { chatContextWindow }));
+
+  it('accepts a non-negative integer (incl. 0 = clear the limit)', async () => {
+    for (const v of [0, 200000]) {
+      const errs = await errorsFor(v);
+      expect(
+        errs.find((e) => e.property === 'chatContextWindow'),
+      ).toBeUndefined();
+    }
+  });
+
+  it('rejects a negative value', async () => {
+    const errs = await errorsFor(-1);
+    expect(errs.find((e) => e.property === 'chatContextWindow')).toBeDefined();
+  });
+
+  it('rejects a non-integer value', async () => {
+    const errs = await errorsFor(1.5);
+    expect(errs.find((e) => e.property === 'chatContextWindow')).toBeDefined();
+  });
+
+  it('accepts the field being omitted (optional)', async () => {
+    const errs = await validate(plainToInstance(UpdateAiSettingsDto, {}));
+    expect(
+      errs.find((e) => e.property === 'chatContextWindow'),
+    ).toBeUndefined();
+  });
+});

apps/server/src/integrations/ai/ai-settings.service.spec.ts

@@ -0,0 +1,43 @@
+import { parsePositiveInt } from './ai-settings.service';
+
+/**
+ * Round-trip coercion for numeric `::text` provider settings (e.g.
+ * chatContextWindow). Values are stored as text and read back as strings, so
+ * this guards the read path the DTO write-validation does not cover: a silent
+ * loss of `Math.floor` or a `> 0` → `>= 0` drift would otherwise go unnoticed.
+ */
+describe('parsePositiveInt', () => {
+  it('keeps a valid positive integer string', () => {
+    expect(parsePositiveInt('200000')).toBe(200000);
+  });
+
+  it('floors a fractional string', () => {
+    expect(parsePositiveInt('1.9')).toBe(1);
+    expect(parsePositiveInt('1.0')).toBe(1);
+  });
+
+  it('returns undefined for zero', () => {
+    expect(parsePositiveInt('0')).toBeUndefined();
+  });
+
+  it('returns undefined for a negative value', () => {
+    expect(parsePositiveInt('-5')).toBeUndefined();
+  });
+
+  it('returns undefined for an empty string', () => {
+    expect(parsePositiveInt('')).toBeUndefined();
+  });
+
+  it('returns undefined for a non-numeric string', () => {
+    expect(parsePositiveInt('abc')).toBeUndefined();
+  });
+
+  it('returns undefined for undefined / null', () => {
+    expect(parsePositiveInt(undefined)).toBeUndefined();
+    expect(parsePositiveInt(null)).toBeUndefined();
+  });
+
+  it('accepts a real number too (not only ::text strings)', () => {
+    expect(parsePositiveInt(42)).toBe(42);
+  });
+});

apps/server/src/integrations/ai/ai-settings.service.ts

@@ -18,6 +18,18 @@ import {
   PROVIDER_SETTINGS_KEYS,
 } from './ai.types';
 
+/**
+ * Coerce a raw provider value (stored as `::text`, so it arrives as a string —
+ * see workspace.repo.ts) into a positive integer, or `undefined` when it is not
+ * a finite number greater than zero. Used for numeric `::text` settings such as
+ * `chatContextWindow`. Fractions are floored: `"1.9" → 1`, `"0"`/`"-5"`/`""`/
+ * `"abc"`/`undefined` → `undefined`.
+ */
+export function parsePositiveInt(raw: unknown): number | undefined {
+  const n = Number(raw);
+  return Number.isFinite(n) && n > 0 ? Math.floor(n) : undefined;
+}
+
 /**
  * Shape of the partial update accepted by `update`. Mirrors the validated
  * controller DTO. `apiKey` / `embeddingApiKey` are write-only: undefined =
@@ -26,6 +38,8 @@ import {
 export interface UpdateAiSettingsInput {
   driver?: AiDriver;
   chatModel?: string;
+  // Max context window in tokens for the chat header badge. 0/empty = no limit.
+  chatContextWindow?: number;
   chatApiStyle?: ChatApiStyle;
   embeddingModel?: string;
   baseUrl?: string;
@@ -160,6 +174,9 @@ export class AiSettingsService {
     const config: ResolvedAiConfig = {
       driver: provider.driver,
       chatModel: provider.chatModel,
+      // Max context window for the chat header badge denominator. Stored as
+      // ::text; 0/unset/invalid = no limit (undefined).
+      chatContextWindow: parsePositiveInt(provider.chatContextWindow),
       // Plain passthrough; getChatModel defaults unset to 'openai-compatible'.
       chatApiStyle: provider.chatApiStyle,
       // Cheap model id for the anonymous public-share assistant; reuses the chat
@@ -219,6 +236,10 @@ export class AiSettingsService {
   async getMasked(workspaceId: string): Promise<MaskedAiSettings> {
     const provider = await this.readProvider(workspaceId);
 
+    // Stored as ::text; coerce to a positive integer (or undefined) so the
+    // client receives a real number.
+    const chatContextWindow = parsePositiveInt(provider.chatContextWindow);
+
     let hasApiKey = false;
     let hasEmbeddingApiKey = false;
     let hasSttApiKey = false;
@@ -243,6 +264,7 @@ export class AiSettingsService {
     return {
       driver: provider.driver,
       chatModel: provider.chatModel,
+      chatContextWindow,
       chatApiStyle: provider.chatApiStyle,
       embeddingModel: provider.embeddingModel,
       baseUrl: provider.baseUrl,

apps/server/src/integrations/ai/ai.types.ts

@@ -32,6 +32,9 @@ export const CHAT_API_STYLES: ChatApiStyle[] = ['openai-compatible', 'openai'];
 export interface AiProviderSettings {
   driver: AiDriver;
   chatModel: string;
+  // Max context window in tokens; surfaced to the chat header badge as the
+  // denominator ("current / max"). 0/unset = no limit (badge shows no denominator).
+  chatContextWindow?: number;
   // Chat provider implementation for the `openai` driver. Unset → defaults to
   // 'openai-compatible' (so reasoning is surfaced by default). See ChatApiStyle.
   chatApiStyle?: ChatApiStyle;
@@ -72,6 +75,7 @@ export interface AiProviderSettings {
 export const PROVIDER_SETTINGS_KEYS = [
   'driver',
   'chatModel',
+  'chatContextWindow',
   'chatApiStyle',
   'embeddingModel',
   'baseUrl',
@@ -98,6 +102,9 @@ export const PROVIDER_SETTINGS_KEYS = [
 export interface ResolvedAiConfig extends Partial<AiProviderSettings> {
   driver?: AiDriver;
   chatModel?: string;
+  // Max context window in tokens; surfaced to the chat header badge as the
+  // "current / max" denominator. 0/unset = no limit.
+  chatContextWindow?: number;
   // Cheap model id for the public-share assistant; reuses the chat creds.
   publicShareChatModel?: string;
   // Agent-role id whose persona the public-share assistant adopts (empty/unset
@@ -116,6 +123,9 @@ export interface ResolvedAiConfig extends Partial<AiProviderSettings> {
 export interface MaskedAiSettings {
   driver?: AiDriver;
   chatModel?: string;
+  // Max context window in tokens; the chat header badge denominator. 0/unset =
+  // no limit.
+  chatContextWindow?: number;
   chatApiStyle?: ChatApiStyle;
   embeddingModel?: string;
   baseUrl?: string;

apps/server/src/integrations/ai/dto/update-ai-settings.dto.ts

@@ -1,4 +1,4 @@
-import { IsIn, IsOptional, IsString } from 'class-validator';
+import { IsIn, IsInt, IsOptional, IsString, Min } from 'class-validator';
 import {
   AI_DRIVERS,
   AiDriver,
@@ -25,6 +25,13 @@ export class UpdateAiSettingsDto {
   @IsString()
   chatModel?: string;
 
+  // Max context window in tokens shown in the chat header badge. 0/empty =
+  // clear the limit (no denominator shown).
+  @IsOptional()
+  @IsInt()
+  @Min(0)
+  chatContextWindow?: number;
+
   @IsOptional()
   @IsIn(CHAT_API_STYLES)
   chatApiStyle?: ChatApiStyle;