@openchambery/web 2.1.8-beta.9 → 2.1.8

package/dist/index.html

@@ -552,7 +552,7 @@
         opacity: 0.72;
       }
     </style>
-    <script type="module" crossorigin src="/assets/main-BjL5V6ie.js"></script>
+    <script type="module" crossorigin src="/assets/main-MSG_QVt6.js"></script>
     <link rel="modulepreload" crossorigin href="/assets/rolldown-runtime-hePW80VL.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-auth-C8w0nlKy.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-switch-3ad3yzM7.js">
@@ -572,8 +572,8 @@
     <link rel="modulepreload" crossorigin href="/assets/terminalApi-kRMttlcP.js">
     <link rel="modulepreload" crossorigin href="/assets/types-BuiqMmh_.js">
     <link rel="modulepreload" crossorigin href="/assets/capgoAdapter-KNLHDj7G.js">
-    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-CRWwlAOU.js">
-    <link rel="stylesheet" crossorigin href="/assets/src-OjoeiMbd.css">
+    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-BHsWHIKP.js">
+    <link rel="stylesheet" crossorigin href="/assets/src-CdGit3gf.css">
   </head>
   <body class="h-full bg-background text-foreground">
     <!-- Keep the startup layer outside #root. React replaces every child of

package/dist/mini-chat.html

@@ -74,7 +74,7 @@
       }
     </style>
 
-    <script type="module" crossorigin src="/assets/miniChat-CHCwBkzu.js"></script>
+    <script type="module" crossorigin src="/assets/miniChat-zpQVvq46.js"></script>
     <link rel="modulepreload" crossorigin href="/assets/rolldown-runtime-hePW80VL.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-auth-C8w0nlKy.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-switch-3ad3yzM7.js">
@@ -94,8 +94,8 @@
     <link rel="modulepreload" crossorigin href="/assets/terminalApi-kRMttlcP.js">
     <link rel="modulepreload" crossorigin href="/assets/types-BuiqMmh_.js">
     <link rel="modulepreload" crossorigin href="/assets/capgoAdapter-KNLHDj7G.js">
-    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-CRWwlAOU.js">
-    <link rel="stylesheet" crossorigin href="/assets/src-OjoeiMbd.css">
+    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-BHsWHIKP.js">
+    <link rel="stylesheet" crossorigin href="/assets/src-CdGit3gf.css">
   </head>
   <body class="h-full bg-background text-foreground">
     <div id="root" class="h-full">

package/dist/mobile.html

@@ -63,7 +63,7 @@
       })();
     </script>
 
-    <script type="module" crossorigin src="/assets/mobile-CG3SgbAO.js"></script>
+    <script type="module" crossorigin src="/assets/mobile-EGZI5jqV.js"></script>
     <link rel="modulepreload" crossorigin href="/assets/rolldown-runtime-hePW80VL.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-auth-C8w0nlKy.js">
     <link rel="modulepreload" crossorigin href="/assets/runtime-switch-3ad3yzM7.js">
@@ -83,8 +83,8 @@
     <link rel="modulepreload" crossorigin href="/assets/terminalApi-kRMttlcP.js">
     <link rel="modulepreload" crossorigin href="/assets/types-BuiqMmh_.js">
     <link rel="modulepreload" crossorigin href="/assets/capgoAdapter-KNLHDj7G.js">
-    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-CRWwlAOU.js">
-    <link rel="stylesheet" crossorigin href="/assets/src-OjoeiMbd.css">
+    <link rel="modulepreload" crossorigin href="/assets/runtimeConfig-BHsWHIKP.js">
+    <link rel="stylesheet" crossorigin href="/assets/src-CdGit3gf.css">
   </head>
   <body class="h-full bg-background text-foreground">
     <div id="root" class="h-full"></div>

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@openchambery/web",
-  "version": "2.1.8-beta.9",
+  "version": "2.1.8",
   "private": false,
   "type": "module",
   "main": "./server/index.js",

package/server/lib/assistants/DOCUMENTATION.md

@@ -1,5 +1,16 @@
 # Assistants Module
 
+## Transient contact errors
+
+Contact-turn failures are delivered through `openchamber:contact-turn-end` only,
+never inserted into the durable contact transcript. The store rejects `status:error`
+writes and atomically deletes legacy error messages and their parts during startup;
+a cleanup failure rolls back and fails startup instead of exposing partial history.
+User messages and already-published valid replies/cards remain intact. Live clients
+show temporary error feedback until a newer message arrives; reconnect/restart does
+not recover errors. Errors do not become list previews, unread counts, or LLM memory.
+Worker-card execution status remains part of the card's authoritative state.
+
 ## Avatar identity
 
 `avatarSeed` is an optional nonempty string (maximum 256 characters) on create
@@ -141,7 +152,7 @@ and keep their existing error result handling.
 - The model is prompted to announce its next action and report verified milestones during execution, then give a short final outcome. Public messages share the existing SSE, durable contact rows, and turn-global indices across web, desktop, VS Code, hosted mobile, and Capacitor. No timer narration, tool logs, progress widgets, or synthetic chat messages are added. Timing of model-authored milestones remains model policy; the transport guarantees immediate delivery once an explicit public record arrives.
 - Completion notifications use the latest spoken message, keeping earlier progress out of the final notification body. Transcript persistence reconciles by stable bubble IDs and adds only remaining messages.
 - Tool results remain model context for every tool. The outcome projector uses assistant-authored text only. Terminal operations retain the mutation stop gate, followed by a tools-disabled completion on the same provider/model to communicate the actual result. A failed/empty summary surfaces a turn error; a completed side effect stays committed and is never replayed by that summary pass.
-- Reset summaries receive only the successful reset tool result, excluding prior messages and attachments. The service retains one model-generated confirmation and discards pre-reset output for that turn. Existing structured failure rows retain diagnostics for recovery; the UI renders those as separate error feedback, outside assistant speech bubbles.
+- Reset summaries receive only the successful reset tool result, excluding prior messages and attachments. The service retains one model-generated confirmation and discards pre-reset output for that turn. Failures emit temporary live error feedback outside assistant speech bubbles, without a persisted failure row.
 
 - Protocol correction never replays rejected assistant prose into model history: simulated User/Assistant turns and tool-result records are rejected as fabricated evidence. Known rejected coding calls cannot be skipped with a subsequent `complete` declaration; the model must issue actual calls or report a blocker. Corrections remain bounded and exhaustion is an error, not a completion.
 
@@ -178,7 +189,7 @@ and keep their existing error result handling.
 - Worker sessions stamp `metadata.openchamber.assigned` (`from: 'contact'`, `assistantID`, `name`). They must **not** use `openchamber.assistant.assistantID` — that marker hides archived Assistant bindings from Chat.
 - `AssistantDTO.assignedSessionIDs` lists in-flight (`busy`/`question`) watches for assigned worker sessions. Session-card busy stays on the card; it does **not** drive the list green dot.
 - **Contact active turn (server authoritative):** process-local metadata tracks admission → `queued` → `running` → settled (removed). `AssistantDTO.working` is true only while that assistant has an unsettled contact turn. `AssistantDTO.activeContactTurn` is `{ turnID, messageID, status: 'queued'|'running', admittedAt }` or `null` so APP restart / reconnect can rehydrate list green dots and the contact 3-dot row from the snapshot HTTP GET. Spoken bubbles and cards persist to SQLite as soon as they publish (`done:true` / `onCard`) so another device or APP restart can GET them while the turn is still running. Token-level SSE deltas remain live paint only. A server process restart clears in-memory activity (no permanent green; tools are not auto-rerun) but already-published bubbles/cards stay in the transcript. Per-assistant lanes isolate turns across assistants/runtimes. Stale `contact-turn-end` for an older `turnID` must not clear a newer active turn.
-- **List latest-message preview (server authoritative):** every `AssistantDTO` path (`snapshot` / `create` / `update` / other `output` rows) carries `latestMessagePreview: { messageID, ordinal, createdAt, role: 'user'|'assistant', text, fallbackKind: 'image'|'file'|'session'|'assistant'|'schedule'|null } | null`. Source is the contact transcript only (not `defaultPrompt` / `updatedAt`). Order is the same keyset as contact pages: newest `(ordinal DESC, message_id DESC)`. Visible roles are `user`, `assistant`, and `peer` (peer maps to `role: 'user'` because inbound DMs are user-visible). Internal assigned-session settle markers (`oc.settle.*`) are excluded in the SQL candidate probe (and again in `buildContactMessagePreview`). Durable error bubbles keep their real body text. Plain text is sanitized and capped at `CONTACT_PREVIEW_MAX_CHARS` (500) with no data-URL/base64 bodies. File-only rows set `fallbackKind` `image`/`file`; pure cards set `session`/`assistant`/`schedule`. Empty transcript → `null`. **Snapshot cost:** one prepared statement reused per assistant (`assistant_id` + indexed `ORDER BY ordinal DESC, message_id DESC LIMIT candidateLimit`, catalog ≤100) collecting only those bounded candidates, then **one** parts `IN` query — not a window/`ROW_NUMBER` over full multi-assistant history and not HTTP N+1. SQL already drops pure settle tails so they cannot exhaust the LIMIT and hide an older real bubble; remaining non-preview empty rows may still consume a candidate slot within that same LIMIT. Admission, assistant completion, peer DM, card insert, wipe, and clear-memory already bump domain revision / `openchamber:assistants-changed`; clients re-GET snapshot for the new preview. Clear-memory keeps transcript rows so the preview still reflects the last bubble; clear-history / reset remove rows so the preview becomes the wipe confirm (or `null` after a full unbounded wipe with no remaining user).
+- **List latest-message preview (server authoritative):** every `AssistantDTO` path (`snapshot` / `create` / `update` / other `output` rows) carries `latestMessagePreview: { messageID, ordinal, createdAt, role: 'user'|'assistant', text, fallbackKind: 'image'|'file'|'session'|'assistant'|'schedule'|null } | null`. Source is the contact transcript only (not `defaultPrompt` / `updatedAt`). Order is the same keyset as contact pages: newest `(ordinal DESC, message_id DESC)`. Visible roles are `user`, `assistant`, and `peer` (peer maps to `role: 'user'` because inbound DMs are user-visible). Internal assigned-session settle markers (`oc.settle.*`) are excluded in the SQL candidate probe (and again in `buildContactMessagePreview`). Transient errors are never stored and cannot replace the preview. Plain text is sanitized and capped at `CONTACT_PREVIEW_MAX_CHARS` (500) with no data-URL/base64 bodies. File-only rows set `fallbackKind` `image`/`file`; pure cards set `session`/`assistant`/`schedule`. Empty transcript → `null`. **Snapshot cost:** one prepared statement reused per assistant (`assistant_id` + indexed `ORDER BY ordinal DESC, message_id DESC LIMIT candidateLimit`, catalog ≤100) collecting only those bounded candidates, then **one** parts `IN` query — not a window/`ROW_NUMBER` over full multi-assistant history and not HTTP N+1. SQL already drops pure settle tails so they cannot exhaust the LIMIT and hide an older real bubble; remaining non-preview empty rows may still consume a candidate slot within that same LIMIT. Admission, assistant completion, peer DM, card insert, wipe, and clear-memory already bump domain revision / `openchamber:assistants-changed`; clients re-GET snapshot for the new preview. Clear-memory keeps transcript rows so the preview still reflects the last bubble; clear-history / reset remove rows so the preview becomes the wipe confirm (or `null` after a full unbounded wipe with no remaining user).
 - No registered project → clear `project_required` (ask the user to configure a project). Never fall back to managed assistant-workspaces.
 - Assign must not go through the peer DM channel.
 - `assignWork` (service) resolves the worker model via the connected catalog (`capabilities.input.image` preferred), forwards tool `variant` (cross-model drops contact variant), passes AbortSignal into create/prompt/delete/catalog, and supplies `lookupMessage` over real SDK exact lookup (`v2.session.message` → `session.message` → bounded `message.list`). Ambiguous prompt admission returns `prompt_ambiguous` with `sessionID`/`messageID` and does **not** delete the worker; the contact-turn assign gate retains that identity so a same-args retry reuses the session and cannot blind-create another worker this turn.

package/server/lib/assistants/contact-attachments.http.test.js

@@ -256,7 +256,7 @@ describe('contact attachments real HTTP (core-routes + assistants routes)', () =
 });
 
 describe('contact attachment materialize failure terminal settle', () => {
-  it('deleting bytes after admit causes durable error, working=false, next turn ok', async () => {
+  it('deleting bytes after admit causes transient error, working=false, next turn ok', async () => {
     const directory = mkRoot();
     closers.push(async () => fsp.rm(directory, { recursive: true, force: true }).catch(() => {}));
 
@@ -339,7 +339,7 @@ describe('contact attachment materialize failure terminal settle', () => {
     expect(snap.activeContactTurn).toBeNull();
 
     const page = service.contactMessages(assistant.id, { limit: 20 });
-    expect(page.messages.some((m) => m.status === 'error' || (m.role === 'assistant' && m.text))).toBe(true);
+    expect(page.messages.some((m) => m.status === 'error' || (m.role === 'assistant' && m.text))).toBe(false);
     const pageJson = JSON.stringify(page);
     expect(pageJson).not.toMatch(/data:image\/png;base64,[A-Za-z0-9+/]{40,}/);
 

package/server/lib/assistants/contact-store.js

@@ -107,6 +107,15 @@ export function ensureContactSchema(db) {
   // only counts assistant tails that belong to this delivery.
   if (!watchColumns.has('message_id')) db.exec('ALTER TABLE assistant_contact_watch ADD COLUMN message_id TEXT');
   // Attachment table ownership lives in contact-attachments.ensureContactAttachmentSchema.
+  // Errors are live feedback, not history. Remove legacy rows and their bodies atomically.
+  db.transaction(() => {
+    db.exec(`
+      DELETE FROM assistant_contact_part WHERE message_id IN (
+        SELECT message_id FROM assistant_contact_message WHERE status = 'error'
+      );
+      DELETE FROM assistant_contact_message WHERE status = 'error';
+    `);
+  })();
 }
 
 /** Current contact transcript generation for one assistant (default 0). */
@@ -719,7 +728,7 @@ export function sanitizeContactPreviewText(text, maxChars = CONTACT_PREVIEW_MAX_
  * Build AssistantDTO.latestMessagePreview from one hydrated contact message.
  * Returns null when the row is not list-visible (internal settle-only, empty,
  * unsupported role). Peer inbox rows map to role `user` (user-visible DM).
- * Visible error bubbles keep real body text. Cards/files set fallbackKind.
+ * Cards/files set fallbackKind. Error feedback never enters persisted previews.
  */
 export function buildContactMessagePreview(message) {
   if (!message || typeof message !== 'object') return null;
@@ -1003,6 +1012,7 @@ export function insertContactMessage(db, {
   fromAssistantID = null,
   fromAssistantName = null,
 }) {
+  if (status === 'error') return;
   db.prepare(
     'INSERT INTO assistant_contact_message(message_id,assistant_id,role,turn_id,bubble_index,created_at,ordinal,status,from_assistant_id,from_assistant_name) VALUES (?,?,?,?,?,?,?,?,?,?)',
   ).run(messageID, assistantID, role, turnID, bubbleIndex, createdAt, ordinal, status, fromAssistantID, fromAssistantName);

package/server/lib/assistants/contact-store.test.js

@@ -120,7 +120,7 @@ describe('contact LLM history trim', () => {
       { role: 'assistant', content: 'keep-assistant' },
       { role: 'assistant', content: expect.stringContaining('\"sessionID\":\"ses_1\"') },
     ]);
-    expect(listContactMessages(db, assistantID, { limit: 50 }).messages).toHaveLength(5);
+    expect(listContactMessages(db, assistantID, { limit: 50 }).messages).toHaveLength(4);
     db.close();
   });
 
@@ -549,21 +549,40 @@ describe('contact latest message preview', () => {
     db.close();
   });
 
-  it('keeps durable error body text as the preview', () => {
+  it('never stores error rows or replaces the latest message preview with an error', () => {
     const db = openDb();
     const assistantID = 'asst_preview_err';
-    insert(db, assistantID, { role: 'user', text: 'hi' });
+    const user = insert(db, assistantID, { role: 'user', text: 'hi' });
     const err = insert(db, assistantID, {
       role: 'assistant',
       text: 'No connected model',
       status: 'error',
     });
     expect(getLatestContactMessagePreview(db, assistantID)).toMatchObject({
-      messageID: err.messageID,
-      role: 'assistant',
-      text: 'No connected model',
+      messageID: user.messageID,
+      role: 'user',
+      text: 'hi',
       fallbackKind: null,
     });
+    expect(db.prepare('SELECT * FROM assistant_contact_message WHERE message_id=?').get(err.messageID)).toBeUndefined();
+    expect(db.prepare('SELECT * FROM assistant_contact_part WHERE message_id=?').all(err.messageID)).toEqual([]);
+    db.close();
+  });
+
+  it('atomically purges legacy error messages and parts on startup', () => {
+    const db = openDb();
+    const user = insert(db, 'legacy', { role: 'user', text: 'keep' });
+    const legacy = insert(db, 'legacy', { role: 'assistant', text: 'old failure' });
+    db.prepare("UPDATE assistant_contact_message SET status='error' WHERE message_id=?").run(legacy.messageID);
+    db.exec(`CREATE TRIGGER fail_error_cleanup BEFORE DELETE ON assistant_contact_message
+      WHEN OLD.status = 'error' BEGIN SELECT RAISE(ABORT, 'cleanup failed'); END;`);
+    expect(() => ensureContactSchema(db)).toThrow('cleanup failed');
+    expect(db.prepare('SELECT * FROM assistant_contact_part WHERE message_id=?').all(legacy.messageID)).toHaveLength(1);
+    db.exec('DROP TRIGGER fail_error_cleanup');
+    ensureContactSchema(db);
+    ensureContactSchema(db);
+    expect(listContactMessages(db, 'legacy').messages.map((message) => message.messageID)).toEqual([user.messageID]);
+    expect(db.prepare('SELECT * FROM assistant_contact_part WHERE message_id=?').all(legacy.messageID)).toEqual([]);
     db.close();
   });
 
@@ -718,7 +737,7 @@ describe('contact latest message preview', () => {
 });
 
 describe('contact read watermark / unread', () => {
-  it('counts only complete/error assistant and peer replies; ignores user and settle-only', () => {
+  it('counts complete assistant and peer replies; ignores errors, user and settle-only', () => {
     expect(isContactVisiblePart({ type: 'text', text: 'hi' })).toBe(true);
     expect(isContactVisiblePart({ type: 'text', text: '  ' })).toBe(false);
     expect(isContactVisiblePart({ type: 'text', text: CONTACT_SETTLE_TEXT.complete })).toBe(false);
@@ -793,7 +812,7 @@ describe('contact read watermark / unread', () => {
     insert(db, assistantID, { role: 'peer', text: 'p1', fromAssistantID: 'other', fromAssistantName: 'Peer' });
     insert(db, assistantID, { role: 'assistant', text: 'failed', status: 'error' });
     // No watermark → all countable unread.
-    expect(countContactUnread(db, assistantID)).toBe(3);
+    expect(countContactUnread(db, assistantID)).toBe(2);
     db.close();
   });
 

package/server/lib/assistants/service.js

@@ -1423,31 +1423,6 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
       throw error;
     }
   };
-  /** Durable failure bubble so APP restart can recover error without SSE. */
-  const persistContactTurnFailure = (assistantID, { userMessageID, turnID, error }) => {
-    const detail = typeof error === 'string' && error.trim() ? error.trim() : 'upstream_error';
-    const failureID = `${userMessageID}:error`;
-    if (getContactMessage(db, failureID)) return;
-    db.exec('BEGIN IMMEDIATE');
-    try {
-      insertContactMessage(db, {
-        messageID: failureID,
-        assistantID,
-        role: 'assistant',
-        turnID,
-        bubbleIndex: 0,
-        createdAt: now(),
-        ordinal: nextContactOrdinal(db, assistantID),
-        status: 'error',
-        parts: [{ type: 'text', text: detail }],
-      });
-      bump();
-      db.exec('COMMIT');
-    } catch (writeError) {
-      db.exec('ROLLBACK');
-      throw writeError;
-    }
-  };
   const contactBubbleMessageID = (userMessageID, bubbleIndex) => `${userMessageID}:bubble:${bubbleIndex + 1}`;
   const contactCardMessageID = (userMessageID, index) => `${userMessageID}:card:${index + 1}`;
 
@@ -2695,7 +2670,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
     const messageID = scheduledRun ? `scheduled_${assistantID}_${scheduledRun.id}` : assignedSessionResumeMessageID(assistantID, sessionID, status, updatedAt);
     const capturedAssistant = scheduledRun ? output(row) : null;
     const turnID = messageID;
-    // Idempotent: same watch status already resumed (bubbles or durable error).
+    // Idempotent: same watch status already resumed with published output.
     if (
       getContactMessage(db, `${messageID}:bubble:1`)
       || getContactMessage(db, `${messageID}:error`)
@@ -2966,16 +2941,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
           ].filter((card, index, list) => list.findIndex((item) => contactCardIdentity(item) === contactCardIdentity(card)) === index);
         if (bubbles.length === 0 && cards.length === 0) {
           const detail = 'Assistant returned no text';
-          try {
-            persistContactTurnFailure(assistantID, {
-              userMessageID: messageID,
-              turnID,
-              error: detail,
-            });
-            settleActiveContactTurn(assistantID, turnID);
-          } catch {
-            settleActiveContactTurn(assistantID, turnID, { bumpRevision: true });
-          }
+          settleActiveContactTurn(assistantID, turnID, { bumpRevision: true });
           emitContactTurnEvent('openchamber:contact-turn-end', {
             assistantID,
             turnID,
@@ -3021,16 +2987,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
           ? error.message.trim()
           : (error?.code || 'upstream_error');
         const statusCode = error?.code === 'no_provider' ? 'no_provider' : (error?.code || 'upstream_error');
-        try {
-          persistContactTurnFailure(assistantID, {
-            userMessageID: messageID,
-            turnID,
-            error: detail,
-          });
-          settleActiveContactTurn(assistantID, turnID);
-        } catch {
-          settleActiveContactTurn(assistantID, turnID, { bumpRevision: true });
-        }
+        settleActiveContactTurn(assistantID, turnID, { bumpRevision: true });
         emitContactTurnEvent('openchamber:contact-turn-end', {
           assistantID,
           turnID,
@@ -3182,7 +3139,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
         };
         const modelContext = await loadContactModelContext(controller.signal);
         try {
-          // Materialize inside turn try so failures durable-error + settle working.
+          // Materialize inside turn try so failures emit live feedback and settle working.
           // DB/UI keep descriptors; harness + assign get execution-time data URLs under budget.
           const executionHistory = await materializeContactHistory(
             db,
@@ -3340,16 +3297,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
             ].filter((card, index, list) => list.findIndex((item) => contactCardIdentity(item) === contactCardIdentity(card)) === index);
           if (bubbles.length === 0 && cards.length === 0) {
             const detail = 'Assistant returned no text';
-            try {
-              persistContactTurnFailure(row.assistant_id, {
-                userMessageID: messageID,
-                turnID,
-                error: detail,
-              });
-              settleActiveContactTurn(row.assistant_id, turnID);
-            } catch {
-              settleActiveContactTurn(row.assistant_id, turnID, { bumpRevision: true });
-            }
+            settleActiveContactTurn(row.assistant_id, turnID, { bumpRevision: true });
             emitContactTurnEvent('openchamber:contact-turn-end', {
               assistantID: row.assistant_id,
               turnID,
@@ -3409,16 +3357,7 @@ export const createAssistantsService = ({ dbPath, dataDir, buildOpenCodeUrl, get
             ? error.message.trim()
             : (error?.code || 'upstream_error');
           const statusCode = error?.code === 'no_provider' ? 'no_provider' : (error?.code || 'upstream_error');
-          try {
-            persistContactTurnFailure(row.assistant_id, {
-              userMessageID: messageID,
-              turnID,
-              error: detail,
-            });
-            settleActiveContactTurn(row.assistant_id, turnID);
-          } catch {
-            settleActiveContactTurn(row.assistant_id, turnID, { bumpRevision: true });
-          }
+          settleActiveContactTurn(row.assistant_id, turnID, { bumpRevision: true });
           emitContactTurnEvent('openchamber:contact-turn-end', {
             assistantID: row.assistant_id,
             turnID,

package/server/lib/assistants/service.test.js

@@ -4230,8 +4230,11 @@ describe('assistants service', () => {
     service.close();
   });
 
-  it('persists a durable contact error bubble so query recovery works without SSE', async () => {
-    const service = setup(root(), {}, {
+  it('delivers contact errors without persisting a transcript message', async () => {
+    const directory = root();
+    const events = [];
+    const service = setup(directory, {}, {
+      onContactTurnEvent: (event) => events.push(event),
       runContactTurn: async () => {
         const error = new Error('No connected model');
         error.code = 'no_provider';
@@ -4247,10 +4250,23 @@ describe('assistants service', () => {
     const page = service.contactMessages(assistant.id);
     expect(page.messages.map((message) => ({ role: message.role, status: message.status, text: message.text }))).toEqual([
       { role: 'user', status: 'complete', text: 'hi' },
-      { role: 'assistant', status: 'error', text: 'No connected model' },
     ]);
     expect(service.snapshot().assistants[0]).toMatchObject({ working: false, activeContactTurn: null });
+    expect(events).toContainEqual(expect.objectContaining({
+      type: 'openchamber:contact-turn-end',
+      properties: expect.objectContaining({ status: 'error', error: 'No connected model' }),
+    }));
     service.close();
+    const restarted = setup(directory, {}, {
+      runContactTurn: async () => ({ text: 'ok', bubbles: ['ok'] }),
+    });
+    expect(restarted.contactMessages(assistant.id).messages.map((message) => message.text)).toEqual(['hi']);
+    await restarted.send(assistant.id, {
+      messageID: 'after_failure', parts: [{ type: 'text', text: 'again' }],
+    });
+    await restarted.whenContactTurnSettled('after_failure');
+    expect(restarted.contactMessages(assistant.id).messages.map((message) => message.text)).toEqual(['hi', 'again', 'ok']);
+    restarted.close();
   });
 
   it('bumps revision when queued contact turn becomes running', async () => {
@@ -4395,10 +4411,9 @@ describe('assistants service', () => {
       && event.properties?.status === 'error'
       && event.properties?.turnID === 'client_no_provider'
     ))).toBe(true);
-    // User row stays; durable error bubble enables query recovery without SSE.
+    // User row stays; error feedback is delivered only through the live event.
     expect(service.contactMessages(assistant.id).messages.map((message) => ({ role: message.role, status: message.status }))).toEqual([
       { role: 'user', status: 'complete' },
-      { role: 'assistant', status: 'error' },
     ]);
     service.close();
   });

package/server/lib/session-index/DOCUMENTATION.md

@@ -100,6 +100,9 @@ consume further pages within a small page budget so the newest-20 **active**
 window can still fill; exhausting the budget before the window is full is an
 explicit partial result (`hasMore` stays true). `replaceDirectory` /
 event-ingest upsert keep active-only rows (archived clears the summary and pin).
+An envelope without a next cursor is exhausted even when the final page is full.
+After exhaustion, `hasMore` is true only when visible root sessions overflow the
+20-row index window; child, archived, and system rows do not create a Show-more hint.
 It publishes an in-memory revision after every externally observable index or
 synchronization-state change. A successful empty directory refresh remains in
 the snapshot so another client can use it as a worktree-topology recovery hint.

package/server/lib/session-index/sync-runtime.js

@@ -225,8 +225,9 @@ export const createSessionIndexSyncRuntime = ({
           continue;
         }
 
-        // No next token: bare-array / terminal page.
-        if (!pageFull) {
+        // An envelope without a next token is authoritative exhaustion, even
+        // when its terminal page is full. Bare arrays still use the page size.
+        if (isEnvelope || !pageFull) {
           upstreamHasMore = false;
           break;
         }
@@ -254,9 +255,13 @@ export const createSessionIndexSyncRuntime = ({
 
       // Root + system filters before filling the active-20 capacity.
       const active = [];
+      let hasOverflowSessions = false;
       for (const session of projected) {
         if (!isRootActiveCandidate(session)) continue;
-        if (active.length >= SESSION_LIMIT) break;
+        if (active.length >= SESSION_LIMIT) {
+          hasOverflowSessions = true;
+          break;
+        }
         active.push(session);
       }
 
@@ -268,7 +273,7 @@ export const createSessionIndexSyncRuntime = ({
         directory: task.directory,
         sessions: active,
         cursor: cursorNumber,
-        hasMore: upstreamHasMore || hitPageBudget || active.length === SESSION_LIMIT,
+        hasMore: upstreamHasMore || hitPageBudget || hasOverflowSessions,
         fullSync: true,
         now: now(),
       });

package/server/lib/session-index/sync-runtime.test.js

@@ -41,6 +41,33 @@ const isFullyIdle = (runtime) => (
 );
 
 describe('session index background sync runtime', () => {
+  it.each([1, 20, 21])('only advertises more when terminal pages contain over 20 visible roots (%i roots)', async (rootCount) => {
+    const service = createService();
+    const rows = Array.from({ length: Math.max(20, rootCount) }, (_, i) => ({
+      ...session(`ses_${i}`, 100 - i),
+      ...(i >= rootCount ? { parentID: 'ses_0' } : {}),
+    }));
+    const runtime = createSessionIndexSyncRuntime({
+      sessionIndexService: service,
+      buildOpenCodeUrl: (route) => `http://opencode.test${route}`,
+      getOpenCodeAuthHeaders: () => ({}),
+      waitForOpenCodeReady: async () => true,
+      fetchFn: async (url) => {
+        if (url.pathname !== '/session' || url.searchParams.has('parentID')) return new Response(JSON.stringify([]));
+        const secondPage = url.searchParams.has('cursor');
+        return new Response(JSON.stringify({
+          data: secondPage ? rows.slice(20) : rows.slice(0, 20),
+          cursor: { next: !secondPage && rows.length > 20 ? 'next-page' : null },
+        }));
+      },
+    });
+    runtime.enqueue(['/repo']);
+    await waitUntil(() => isFullyIdle(runtime));
+    expect(service.snapshot().directories[0].sessions).toHaveLength(Math.min(rootCount, 20));
+    expect(service.snapshot().directories[0].hasMore).toBe(rootCount > 20);
+    runtime.stop();
+  });
+
   it('skips a full background-tagged page without consuming the active session budget', async () => {
     const service = createService();
     const cursors = [];