v1.12.4: complete inference.ts split into services/inference/

- sentinel-summaries.ts: runCapHitSummary, insertCapHitSentinel,
  runDoomLoopSummary, insertDoomLoopSentinel
- inference.ts → inference/turn.ts: residue is runAssistantTurn,
  runInference, createInferenceRunner orchestration only
- inference/index.ts: re-export shim preserves the public surface
  (createInferenceRunner, runInference, runAssistantTurn,
  detectDoomLoop, DOOM_LOOP_THRESHOLD, buildMessagesPayload, plus
  type-side InferenceContext/InferenceFrame/StreamResult/TurnArgs/
  FramePublisher)
- src/index.ts + auto_name.ts + the two vitest test files updated to
  import from ./services/inference/index.js explicitly (NodeNext ESM
  doesn't honor directory-index resolution)

Final tally: 11 files under services/inference/, the largest being
sentinel-summaries.ts at 523 LoC (two near-clone summary paths kept
side-by-side until a third sentinel justifies factoring out a shared
runWrapUpSummary). turn.ts is now 326 LoC, the next-largest is
stream-phase.ts at 380. Public import surface unchanged.

tool-phase.ts → turn.ts back-edge for runAssistantTurn remains
(cycle is safe; resolved at call time).

Prepares the file structure for v1.13 AI SDK migration — streamText
swap targets stream-phase.ts only.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

apps/server/src/index.ts

@@ -16,7 +16,7 @@ import { registerWebSocket } from './routes/ws.js';
 import { registerModelRoutes } from './routes/models.js';
 import { registerAgentRoutes } from './routes/agents.js';
 import { registerSkillsRoutes } from './routes/skills.js';
-import { createInferenceRunner } from './services/inference.js';
+import { createInferenceRunner } from './services/inference/index.js';
 import { createBroker } from './services/broker.js';
 import { listSkills } from './services/skills.js';
 import * as compaction from './services/compaction.js';

apps/server/src/routes/chats.ts

@@ -18,6 +18,12 @@ const ForkBody = z.object({
   name: z.string().min(1).max(200).optional(),
 });
 
+const DiscardStaleBody = z.object({
+  message_id: z.string().uuid(),
+});
+
+const STALE_MIN_AGE_SECONDS = 60;
+
 export function registerChatRoutes(
   app: FastifyInstance,
   sql: Sql,
@@ -320,6 +326,73 @@ export function registerChatRoutes(
     }
   );
 
+  // v1.12.3: explicit recovery from a stuck-streaming assistant row. The
+  // frontend gates this behind a 60s no-token-activity timer; the server
+  // re-checks the age and current status for safety. Non-streaming rows
+  // return 409 (frontend race; idempotent retry is fine).
+  app.post<{ Params: { id: string } }>(
+    '/api/chats/:id/discard_stale',
+    async (req, reply) => {
+      const parsed = DiscardStaleBody.safeParse(req.body ?? {});
+      if (!parsed.success) {
+        reply.code(400);
+        return { error: 'invalid body', details: parsed.error.flatten() };
+      }
+      const rows = await sql<{
+        id: string;
+        session_id: string;
+        chat_id: string;
+        status: string;
+        age_seconds: number;
+      }[]>`
+        SELECT id, session_id, chat_id, status,
+               EXTRACT(EPOCH FROM (clock_timestamp() - created_at))::int AS age_seconds
+        FROM messages
+        WHERE id = ${parsed.data.message_id} AND chat_id = ${req.params.id}
+      `;
+      if (rows.length === 0) {
+        reply.code(404);
+        return { error: 'message not found in chat' };
+      }
+      const msg = rows[0]!;
+      if (msg.status !== 'streaming') {
+        reply.code(409);
+        return { error: 'message is no longer streaming', current_status: msg.status };
+      }
+      if (msg.age_seconds < STALE_MIN_AGE_SECONDS) {
+        reply.code(409);
+        return { error: 'message is not stale yet', age_seconds: msg.age_seconds };
+      }
+      const updated = await sql<Message[]>`
+        UPDATE messages
+        SET status = 'failed',
+            content = COALESCE(content, ''),
+            finished_at = clock_timestamp()
+        WHERE id = ${msg.id} AND status = 'streaming'
+        RETURNING id, session_id, chat_id, role, content, kind, tool_calls, tool_results,
+                  status, last_seq, tokens_used, ctx_used, ctx_max, started_at, finished_at,
+                  created_at, metadata, summary, tail_start_id, compacted_at
+      `;
+      if (updated.length === 0) {
+        // Race: the row flipped out of 'streaming' between our SELECT and UPDATE.
+        reply.code(409);
+        return { error: 'message status changed mid-request' };
+      }
+      broker.publishUser('default', {
+        type: 'chat_status',
+        chat_id: msg.chat_id,
+        status: 'idle',
+        at: new Date().toISOString(),
+      });
+      broker.publish(msg.session_id, {
+        type: 'message_complete',
+        message_id: msg.id,
+        chat_id: msg.chat_id,
+      });
+      return updated[0];
+    }
+  );
+
   app.get<{ Params: { id: string } }>(
     '/api/chats/:id/messages',
     async (req, reply) => {

apps/server/src/services/__tests__/doom-loop.test.ts

@@ -1,5 +1,5 @@
 import { describe, it, expect } from 'vitest';
-import { DOOM_LOOP_THRESHOLD, detectDoomLoop } from '../inference.js';
+import { DOOM_LOOP_THRESHOLD, detectDoomLoop } from '../inference/index.js';
 import type { ToolCall } from '../../types/api.js';
 
 // ---- fixture ----------------------------------------------------------------

apps/server/src/services/__tests__/inference.test.ts

@@ -1,5 +1,5 @@
 import { describe, it, expect } from 'vitest';
-import { buildMessagesPayload } from '../inference.js';
+import { buildMessagesPayload } from '../inference/index.js';
 import type {
   Message,
   MessageRole,

apps/server/src/services/auto_name.ts

@@ -1,4 +1,4 @@
-import type { InferenceContext } from './inference.js';
+import type { InferenceContext } from './inference/index.js';
 
 const NAMING_SYSTEM_PROMPT =
   'You name chat sessions. Reply directly with no thinking, reasoning, or explanation. Output ONLY the title, 4 words max, no quotes, no punctuation, no prefix like "Title:".';

apps/server/src/services/inference/budget.ts

@@ -0,0 +1,20 @@
+import type { Agent } from '../../types/api.js';
+import { READ_ONLY_TOOL_NAMES } from '../tools.js';
+
+// v1.8.2: tool-call budget defaults. Resolved per-turn by resolveToolBudget.
+//   - Agent with explicit max_tool_calls: that value.
+//   - Agent with read-only-only tools:    BUDGET_READ_ONLY (30).
+//   - Agent with any non-read-only tool:  BUDGET_NON_READ_ONLY (10).
+//   - No agent (raw chat):                BUDGET_NO_AGENT (15).
+export const BUDGET_READ_ONLY = 30;
+export const BUDGET_NON_READ_ONLY = 10;
+export const BUDGET_NO_AGENT = 15;
+
+const READ_ONLY_SET: ReadonlySet<string> = new Set(READ_ONLY_TOOL_NAMES);
+
+export function resolveToolBudget(agent: Agent | null): number {
+  if (agent?.max_tool_calls != null) return agent.max_tool_calls;
+  if (!agent) return BUDGET_NO_AGENT;
+  const allReadOnly = agent.tools.every((t) => READ_ONLY_SET.has(t));
+  return allReadOnly ? BUDGET_READ_ONLY : BUDGET_NON_READ_ONLY;
+}

apps/server/src/services/inference/error-handler.ts

@@ -0,0 +1,148 @@
+import type { MessageMetadata, Session } from '../../types/api.js';
+import * as modelContext from '../model-context.js';
+import { maybeFlagForCompaction } from './payload.js';
+import type { InferenceContext, StreamResult, TurnArgs } from './turn.js';
+
+export async function handleAbortOrError(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  accumulated: string,
+  err: unknown
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId } = args;
+  const isAbort = err instanceof Error && err.name === 'AbortError';
+  const finalStatus = isAbort ? 'cancelled' : 'failed';
+  const errMsg = err instanceof Error ? err.message : String(err);
+  // v1.8.2: persist a structured error metadata blob on genuine failures so
+  // the bubble can render the reason on reload without re-deriving from the
+  // (one-shot) WS error frame. User-initiated abort skips this — there's no
+  // "reason" to surface for a stop the user already explicitly chose.
+  const errorMetadata: MessageMetadata | null = isAbort
+    ? null
+    : { kind: 'error', error_reason: 'llm_provider_error', error_text: errMsg };
+  if (errorMetadata) {
+    await ctx.sql`
+      UPDATE messages
+      SET status = ${finalStatus},
+          content = ${accumulated},
+          finished_at = clock_timestamp(),
+          metadata = ${ctx.sql.json(errorMetadata as never)}
+      WHERE id = ${assistantMessageId}
+    `;
+  } else {
+    await ctx.sql`
+      UPDATE messages
+      SET status = ${finalStatus},
+          content = ${accumulated},
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+    `;
+  }
+  const [failSessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({ type: 'session_updated', session_id: sessionId, project_id: failSessRow!.project_id, name: failSessRow!.name, updated_at: failSessRow!.updated_at });
+  // v1.8 mobile-tabs: cancellation is a user-initiated stop, treat as idle;
+  // genuine errors flip the dot red. v1.8.2: error path also carries a
+  // machine-readable `reason` so the UI can render specifics inline.
+  if (isAbort) {
+    // v1.12.1: defensive cancellation write. The status=${finalStatus} UPDATE
+    // above already sets 'cancelled' for the AbortError case, but a row can
+    // leak as 'streaming' when the abort fires between the post-tool-phase
+    // INSERT (executeToolPhase) and the next runAssistantTurn's stream setup,
+    // bypassing the try/catch around executeStreamPhase. The status guard
+    // makes this a no-op when the earlier write already landed.
+    await ctx.sql`
+      UPDATE messages
+      SET status = 'cancelled', content = ${accumulated}, finished_at = clock_timestamp()
+      WHERE id = ${args.assistantMessageId} AND status = 'streaming'
+    `;
+    ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+    });
+    ctx.log.info({ sessionId, chatId, assistantMessageId }, 'inference cancelled');
+  } else {
+    ctx.publishUser({
+      type: 'chat_status',
+      chat_id: chatId,
+      status: 'error',
+      at: new Date().toISOString(),
+      reason: 'llm_provider_error',
+    });
+    ctx.publish(sessionId, {
+      type: 'error',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      error: errMsg,
+      reason: 'llm_provider_error',
+    });
+    ctx.log.error({ err, sessionId, assistantMessageId }, 'inference failed');
+  }
+}
+
+export async function finalizeCompletion(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  result: StreamResult,
+  startedAt: string | null,
+  session: Session
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId } = args;
+  const { content, finishReason, promptTokens, completionTokens } = result;
+
+  // v1.11.3: see executeToolPhase for the rationale.
+  const mctx = await modelContext.getModelContext(session.model);
+  const nCtx = mctx?.n_ctx ?? null;
+
+  const [updated] = await ctx.sql<
+    { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null; finished_at: string | null }[]
+  >`
+    UPDATE messages
+    SET content = ${content},
+        status = 'complete',
+        tokens_used = ${completionTokens},
+        ctx_used = ${promptTokens},
+        ctx_max = ${nCtx},
+        finished_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING tokens_used, ctx_used, ctx_max, finished_at
+  `;
+  // v1.11: flag for compaction on the terminal turn too. Catches the common
+  // case of a turn that hit the limit without invoking tools.
+  await maybeFlagForCompaction(ctx, chatId, updated);
+  const [completeSessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({ type: 'session_updated', session_id: sessionId, project_id: completeSessRow!.project_id, name: completeSessRow!.name, updated_at: completeSessRow!.updated_at });
+  ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+  ctx.publish(sessionId, {
+    type: 'message_complete',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    tokens_used: updated?.tokens_used ?? null,
+    ctx_used: updated?.ctx_used ?? null,
+    ctx_max: updated?.ctx_max ?? null,
+    started_at: startedAt,
+    finished_at: updated?.finished_at ?? null,
+    model: session.model,
+  });
+  ctx.log.info(
+    {
+      sessionId,
+      chatId,
+      assistantMessageId,
+      finishReason,
+      chars: content.length,
+      tokens_used: updated?.tokens_used,
+      ctx_used: updated?.ctx_used,
+    },
+    'inference complete'
+  );
+}

apps/server/src/services/inference/index.ts

@@ -0,0 +1,20 @@
+// v1.12.4: re-export shim. Outside callers (apps/server/src/index.ts and the
+// vitest inference tests) import from './services/inference/index.js'. The
+// directory is now the public surface; turn.ts holds runAssistantTurn /
+// runInference / createInferenceRunner while the other inference/*.ts files
+// stay implementation-private.
+
+export {
+  createInferenceRunner,
+  runAssistantTurn,
+  runInference,
+} from './turn.js';
+export type {
+  FramePublisher,
+  InferenceContext,
+  InferenceFrame,
+  StreamResult,
+  TurnArgs,
+} from './turn.js';
+export { detectDoomLoop, DOOM_LOOP_THRESHOLD } from './sentinels.js';
+export { buildMessagesPayload } from './payload.js';

apps/server/src/services/inference/payload.ts

@@ -0,0 +1,155 @@
+import type { Sql } from '../../db.js';
+import type {
+  Agent,
+  Message,
+  Project,
+  Session,
+} from '../../types/api.js';
+import * as compaction from '../compaction.js';
+import { buildSystemPrompt } from '../system-prompt.js';
+import { isAnySentinel } from './sentinels.js';
+import type { InferenceContext } from './turn.js';
+
+export interface OpenAiMessage {
+  role: 'system' | 'user' | 'assistant' | 'tool';
+  content: string | null;
+  tool_calls?: Array<{
+    id: string;
+    type: 'function';
+    function: { name: string; arguments: string };
+  }>;
+  tool_call_id?: string;
+}
+
+// v1.12: buildSystemPrompt lives in services/system-prompt.ts. It awaits the
+// container-guidance loader, so this function is async too and every call
+// site in inference.ts awaits the result.
+export async function buildMessagesPayload(
+  session: Session,
+  project: Project,
+  history: Message[],
+  agent: Agent | null = null
+): Promise<OpenAiMessage[]> {
+  const out: OpenAiMessage[] = [];
+  const systemPrompt = await buildSystemPrompt(project, session, agent);
+  out.push({ role: 'system', content: systemPrompt });
+
+  // Find the latest compact marker — only send messages from that point onwards
+  let startIdx = 0;
+  for (let i = history.length - 1; i >= 0; i--) {
+    if (history[i]!.kind === 'compact') {
+      startIdx = i;
+      break;
+    }
+  }
+
+  for (let i = startIdx; i < history.length; i++) {
+    const m = history[i]!;
+    if (m.kind === 'compact') {
+      out.push({ role: 'system', content: m.content });
+      continue;
+    }
+    // v1.8.2 / v1.11.6: cap-hit and doom-loop sentinels are UI-only — never
+    // send them to the LLM. The synthetic instruction note lives only inside
+    // the summary call's messages array and is never persisted, so on a
+    // follow-up turn the model resumes with a clean context.
+    if (isAnySentinel(m)) continue;
+    if (m.role === 'assistant' && m.status === 'streaming') continue;
+    if (m.role === 'assistant' && m.status === 'cancelled') continue;
+    if (m.role === 'tool') {
+      const tr = m.tool_results;
+      if (!tr) continue;
+      const outputText = tr.error
+        ? `error: ${tr.error}`
+        : typeof tr.output === 'string'
+          ? tr.output
+          : JSON.stringify(tr.output);
+      out.push({
+        role: 'tool',
+        content: outputText,
+        tool_call_id: tr.tool_call_id,
+      });
+      continue;
+    }
+    if (m.role === 'assistant') {
+      const msg: OpenAiMessage = {
+        role: 'assistant',
+        content: m.content && m.content.length > 0 ? m.content : null,
+      };
+      if (m.tool_calls && m.tool_calls.length > 0) {
+        msg.tool_calls = m.tool_calls.map((tc) => ({
+          id: tc.id,
+          type: 'function' as const,
+          function: { name: tc.name, arguments: JSON.stringify(tc.args) },
+        }));
+      }
+      out.push(msg);
+      continue;
+    }
+    out.push({ role: 'user', content: m.content });
+  }
+  return out;
+}
+
+export async function loadContext(
+  sql: Sql,
+  sessionId: string,
+  chatId: string
+): Promise<{ session: Session; project: Project; history: Message[] } | null> {
+  const sessionRows = await sql<Session[]>`
+    SELECT id, project_id, name, model, system_prompt, status, created_at, updated_at,
+           agent_id, web_search_enabled
+    FROM sessions WHERE id = ${sessionId}
+  `;
+  if (sessionRows.length === 0) return null;
+  const session = sessionRows[0]!;
+
+  const projectRows = await sql<Project[]>`
+    SELECT id, name, path, added_at, last_session_id, status, gitea_remote,
+           default_system_prompt, default_web_search_enabled
+    FROM projects WHERE id = ${session.project_id}
+  `;
+  if (projectRows.length === 0) return null;
+  const project = projectRows[0]!;
+
+  // v1.11: filter compacted messages out of the inference assembly. The GET
+  // /api/sessions/:id/messages endpoint still returns everything (so the UI
+  // can show history with the summary card inline); only LLM payloads skip
+  // compacted rows. compacted_at IS NULL keeps the active summary + tail.
+  const history = await sql<Message[]>`
+    SELECT id, session_id, chat_id, role, content, kind, tool_calls, tool_results, status, last_seq,
+           tokens_used, ctx_used, ctx_max, started_at, finished_at, created_at, metadata
+    FROM messages
+    WHERE chat_id = ${chatId} AND compacted_at IS NULL
+    ORDER BY created_at ASC, id ASC
+  `;
+
+  return { session, project, history };
+}
+
+// v1.11: shared helper used after both finalizeCompletion and executeToolPhase
+// persist their token counts. Reads tokens off the just-UPDATEd row (which
+// the caller returns from RETURNING), runs compaction.isOverflow, and flips
+// chats.needs_compaction. The next runAssistantTurn invocation acts on it.
+// Silent on missing tokens — llama-swap occasionally omits usage on truncated
+// streams, and we'd rather miss one overflow than crash the inference path.
+export async function maybeFlagForCompaction(
+  ctx: InferenceContext,
+  chatId: string,
+  updated: { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null } | undefined,
+): Promise<void> {
+  if (!updated) return;
+  const promptTokens = updated.ctx_used;
+  const completionTokens = updated.tokens_used;
+  const contextLimit = updated.ctx_max;
+  if (typeof promptTokens !== 'number') return;
+  if (typeof completionTokens !== 'number') return;
+  if (typeof contextLimit !== 'number') return;
+  const overflow = compaction.isOverflow(
+    { prompt_tokens: promptTokens, completion_tokens: completionTokens },
+    contextLimit,
+  );
+  if (!overflow) return;
+  await ctx.sql`UPDATE chats SET needs_compaction = true WHERE id = ${chatId}`;
+  ctx.log.info({ chatId, promptTokens, completionTokens, contextLimit }, 'inference: flagged for compaction');
+}

apps/server/src/services/inference/sentinel-summaries.ts

@@ -0,0 +1,523 @@
+import type {
+  Agent,
+  Message,
+  MessageMetadata,
+  Project,
+  Session,
+} from '../../types/api.js';
+import * as modelContext from '../model-context.js';
+import { buildMessagesPayload } from './payload.js';
+import { DOOM_LOOP_THRESHOLD } from './sentinels.js';
+import { streamCompletion } from './stream-phase.js';
+import { DB_FLUSH_INTERVAL_MS } from './types.js';
+import type {
+  InferenceContext,
+  StreamResult,
+  TurnArgs,
+} from './turn.js';
+
+// Synthetic system note appended to the cap-hit summary call. Verbatim from
+// the v1.8.2 spec — do not paraphrase: the model is more reliable when the
+// instruction is short, declarative, and identical across calls.
+const CAP_HIT_SUMMARY_NOTE = (limit: number) =>
+  `You've reached the tool budget (${limit} calls). Produce the best answer you can with what you have. Do not call more tools.`;
+
+const DOOM_LOOP_NOTE = (name: string) =>
+  `You called ${name} with the same arguments ${DOOM_LOOP_THRESHOLD} times in a row. Stop calling it. Produce the best answer you can with what you have.`;
+
+export async function runCapHitSummary(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  session: Session,
+  project: Project,
+  history: Message[],
+  agent: Agent | null,
+  budget: number,
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId, signal } = args;
+
+  const messages = await buildMessagesPayload(session, project, history, agent);
+  messages.push({ role: 'system', content: CAP_HIT_SUMMARY_NOTE(budget) });
+
+  const startedRow = await ctx.sql<{ started_at: string }[]>`
+    UPDATE messages
+    SET started_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING started_at
+  `;
+  const startedAt = startedRow[0]?.started_at ?? null;
+
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    role: 'assistant',
+  });
+
+  let accumulated = '';
+  let pendingFlushTimer: NodeJS.Timeout | null = null;
+  let flushPromise: Promise<unknown> = Promise.resolve();
+  const flushNow = () => {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    const snapshot = accumulated;
+    flushPromise = flushPromise.then(() =>
+      ctx.sql`UPDATE messages SET content = ${snapshot} WHERE id = ${assistantMessageId}`
+    );
+  };
+  const scheduleFlush = () => {
+    if (pendingFlushTimer) return;
+    pendingFlushTimer = setTimeout(() => {
+      pendingFlushTimer = null;
+      flushNow();
+    }, DB_FLUSH_INTERVAL_MS);
+  };
+
+  let summaryOk = false;
+  let summarySoftCancelled = false;
+  let summaryError: string | null = null;
+  let result: StreamResult | null = null;
+  try {
+    result = await streamCompletion(
+      ctx,
+      session.model,
+      messages,
+      { tools: null, temperature: agent?.temperature },
+      (delta) => {
+        accumulated += delta;
+        ctx.publish(sessionId, {
+          type: 'delta',
+          message_id: assistantMessageId,
+          chat_id: chatId,
+          content: delta,
+        });
+        scheduleFlush();
+      },
+      undefined,
+      signal,
+    );
+    summaryOk = true;
+  } catch (err) {
+    if (err instanceof Error && err.name === 'AbortError') {
+      summarySoftCancelled = true;
+    } else {
+      summaryError = err instanceof Error ? err.message : String(err);
+    }
+  } finally {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    await flushPromise;
+  }
+
+  // Finalize the summary message based on the three outcomes. The sentinel
+  // is inserted regardless so the user always has the Continue affordance —
+  // even on a partial / failed summary the chat history shows where the
+  // budget was hit.
+  if (summaryOk && result) {
+    // v1.11.3: see executeToolPhase for the rationale.
+    const mctx = await modelContext.getModelContext(session.model);
+    const nCtx = mctx?.n_ctx ?? null;
+    const [updated] = await ctx.sql<
+      { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null; finished_at: string | null }[]
+    >`
+      UPDATE messages
+      SET content = ${result.content},
+          status = 'complete',
+          tokens_used = ${result.completionTokens},
+          ctx_used = ${result.promptTokens},
+          ctx_max = ${nCtx},
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+      RETURNING tokens_used, ctx_used, ctx_max, finished_at
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      tokens_used: updated?.tokens_used ?? null,
+      ctx_used: updated?.ctx_used ?? null,
+      ctx_max: updated?.ctx_max ?? null,
+      started_at: startedAt,
+      finished_at: updated?.finished_at ?? null,
+      model: session.model,
+    });
+  } else if (summarySoftCancelled) {
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'cancelled',
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+    });
+  } else {
+    const errMeta: MessageMetadata = {
+      kind: 'error',
+      error_reason: 'summary_after_cap_failed',
+      error_text: summaryError ?? 'summary failed',
+    };
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'failed',
+          finished_at = clock_timestamp(),
+          metadata = ${ctx.sql.json(errMeta as never)}
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'error',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      error: summaryError ?? 'summary failed',
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  // Bump session/chat updated_at exactly once for this turn.
+  const [sessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({
+    type: 'session_updated',
+    session_id: sessionId,
+    project_id: sessRow!.project_id,
+    name: sessRow!.name,
+    updated_at: sessRow!.updated_at,
+  });
+
+  await insertCapHitSentinel(ctx, sessionId, chatId, agent, budget);
+
+  // Status frame fires last so the dot color reflects the terminal state.
+  // Success → idle, abort → idle (user-driven stop), error → error+reason.
+  if (summaryOk) {
+    ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+  } else if (summarySoftCancelled) {
+    ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+  } else {
+    ctx.publishUser({
+      type: 'chat_status',
+      chat_id: chatId,
+      status: 'error',
+      at: new Date().toISOString(),
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  ctx.log.info(
+    { sessionId, chatId, assistantMessageId, budget, summaryOk, summaryCancelled: summarySoftCancelled },
+    'inference cap-hit summary finished',
+  );
+}
+
+async function insertCapHitSentinel(
+  ctx: InferenceContext,
+  sessionId: string,
+  chatId: string,
+  agent: Agent | null,
+  budget: number,
+): Promise<void> {
+  // Hard ceiling: count prior cap_hit sentinels in this chat. After two
+  // continues (sentinel count of 2), the next sentinel reports can_continue
+  // false and the UI disables the Continue button.
+  const priorRows = await ctx.sql<{ count: number }[]>`
+    SELECT COUNT(*)::int AS count
+    FROM messages
+    WHERE chat_id = ${chatId}
+      AND role = 'system'
+      AND metadata->>'kind' = 'cap_hit'
+  `;
+  const priorCount = priorRows[0]?.count ?? 0;
+  const canContinue = priorCount < 2;
+  const metadata: MessageMetadata = {
+    kind: 'cap_hit',
+    used: budget,
+    limit: budget,
+    agent_name: agent?.name ?? null,
+    can_continue: canContinue,
+  };
+  const content = `Reached tool budget (${budget}/${budget}). Continue to extend.`;
+
+  const [row] = await ctx.sql<{ id: string }[]>`
+    INSERT INTO messages (session_id, chat_id, role, content, status, created_at, metadata)
+    VALUES (${sessionId}, ${chatId}, 'system', ${content}, 'complete', clock_timestamp(), ${ctx.sql.json(metadata as never)})
+    RETURNING id
+  `;
+
+  // The sentinel content is static, but we still walk the standard frame
+  // sequence (started → delta → complete) so useSessionStream's reducer
+  // appends it via the same path it uses for streaming assistant messages.
+  // The delta carries the full text in one chunk.
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: row!.id,
+    chat_id: chatId,
+    role: 'system',
+  });
+  ctx.publish(sessionId, {
+    type: 'delta',
+    message_id: row!.id,
+    chat_id: chatId,
+    content,
+  });
+  ctx.publish(sessionId, {
+    type: 'message_complete',
+    message_id: row!.id,
+    chat_id: chatId,
+    metadata,
+  });
+}
+
+// v1.11.6: doom-loop wrap-up. Mirrors runCapHitSummary structurally — same
+// in-flight-slot reuse, same tools-disabled streaming-summary call, same
+// post-finalize sentinel insert + chat_status drop. Differences:
+//   - synthetic note text comes from DOOM_LOOP_NOTE (names the looping tool)
+//   - sentinel metadata is { kind: 'doom_loop', tool_name, args, threshold }
+//     and has no Continue affordance (manual retry would just re-loop)
+//   - chat_status error path uses reason: 'doom_loop_summary_failed'
+// Kept as a clone rather than refactored into a shared helper because the
+// two summary paths still differ in error reason + sentinel shape; a third
+// sentinel would justify factoring out runWrapUpSummary(opts).
+export async function runDoomLoopSummary(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  session: Session,
+  project: Project,
+  history: Message[],
+  agent: Agent | null,
+  loop: { name: string; args: Record<string, unknown> },
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId, signal } = args;
+
+  const messages = await buildMessagesPayload(session, project, history, agent);
+  messages.push({ role: 'system', content: DOOM_LOOP_NOTE(loop.name) });
+
+  const startedRow = await ctx.sql<{ started_at: string }[]>`
+    UPDATE messages
+    SET started_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING started_at
+  `;
+  const startedAt = startedRow[0]?.started_at ?? null;
+
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    role: 'assistant',
+  });
+
+  let accumulated = '';
+  let pendingFlushTimer: NodeJS.Timeout | null = null;
+  let flushPromise: Promise<unknown> = Promise.resolve();
+  const flushNow = () => {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    const snapshot = accumulated;
+    flushPromise = flushPromise.then(() =>
+      ctx.sql`UPDATE messages SET content = ${snapshot} WHERE id = ${assistantMessageId}`
+    );
+  };
+  const scheduleFlush = () => {
+    if (pendingFlushTimer) return;
+    pendingFlushTimer = setTimeout(() => {
+      pendingFlushTimer = null;
+      flushNow();
+    }, DB_FLUSH_INTERVAL_MS);
+  };
+
+  let summaryOk = false;
+  let summarySoftCancelled = false;
+  let summaryError: string | null = null;
+  let result: StreamResult | null = null;
+  try {
+    result = await streamCompletion(
+      ctx,
+      session.model,
+      messages,
+      { tools: null, temperature: agent?.temperature },
+      (delta) => {
+        accumulated += delta;
+        ctx.publish(sessionId, {
+          type: 'delta',
+          message_id: assistantMessageId,
+          chat_id: chatId,
+          content: delta,
+        });
+        scheduleFlush();
+      },
+      undefined,
+      signal,
+    );
+    summaryOk = true;
+  } catch (err) {
+    if (err instanceof Error && err.name === 'AbortError') {
+      summarySoftCancelled = true;
+    } else {
+      summaryError = err instanceof Error ? err.message : String(err);
+    }
+  } finally {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    await flushPromise;
+  }
+
+  if (summaryOk && result) {
+    const mctx = await modelContext.getModelContext(session.model);
+    const nCtx = mctx?.n_ctx ?? null;
+    const [updated] = await ctx.sql<
+      { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null; finished_at: string | null }[]
+    >`
+      UPDATE messages
+      SET content = ${result.content},
+          status = 'complete',
+          tokens_used = ${result.completionTokens},
+          ctx_used = ${result.promptTokens},
+          ctx_max = ${nCtx},
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+      RETURNING tokens_used, ctx_used, ctx_max, finished_at
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      tokens_used: updated?.tokens_used ?? null,
+      ctx_used: updated?.ctx_used ?? null,
+      ctx_max: updated?.ctx_max ?? null,
+      started_at: startedAt,
+      finished_at: updated?.finished_at ?? null,
+      model: session.model,
+    });
+  } else if (summarySoftCancelled) {
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'cancelled',
+          finished_at = clock_timestamp()
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'message_complete',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+    });
+  } else {
+    // Doom-loop summary failure reuses the existing summary_after_cap_failed
+    // error reason — the ErrorReason union is shared between sentinel paths
+    // and the UI surfaces a generic "summary failed" line for both. We don't
+    // add a new reason code because the user-visible failure mode is the
+    // same (model gave up mid-summary). Sentinel below still fires.
+    const errMeta: MessageMetadata = {
+      kind: 'error',
+      error_reason: 'summary_after_cap_failed',
+      error_text: summaryError ?? 'doom-loop summary failed',
+    };
+    await ctx.sql`
+      UPDATE messages
+      SET content = ${accumulated},
+          status = 'failed',
+          finished_at = clock_timestamp(),
+          metadata = ${ctx.sql.json(errMeta as never)}
+      WHERE id = ${assistantMessageId}
+    `;
+    ctx.publish(sessionId, {
+      type: 'error',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      error: summaryError ?? 'doom-loop summary failed',
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  const [sessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({
+    type: 'session_updated',
+    session_id: sessionId,
+    project_id: sessRow!.project_id,
+    name: sessRow!.name,
+    updated_at: sessRow!.updated_at,
+  });
+
+  await insertDoomLoopSentinel(ctx, sessionId, chatId, loop);
+
+  if (summaryOk || summarySoftCancelled) {
+    ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'idle', at: new Date().toISOString() });
+  } else {
+    ctx.publishUser({
+      type: 'chat_status',
+      chat_id: chatId,
+      status: 'error',
+      at: new Date().toISOString(),
+      reason: 'summary_after_cap_failed',
+    });
+  }
+
+  ctx.log.info(
+    { sessionId, chatId, assistantMessageId, loopedTool: loop.name, summaryOk, summaryCancelled: summarySoftCancelled },
+    'inference doom-loop summary finished',
+  );
+}
+
+async function insertDoomLoopSentinel(
+  ctx: InferenceContext,
+  sessionId: string,
+  chatId: string,
+  loop: { name: string; args: Record<string, unknown> },
+): Promise<void> {
+  // No hard-ceiling / can-continue logic here — doom-loop is a different
+  // failure mode from cap-hit. Continuing would re-trigger the loop with
+  // the same tools available; the user needs to restate their question
+  // or switch agents instead.
+  const metadata: MessageMetadata = {
+    kind: 'doom_loop',
+    tool_name: loop.name,
+    args: loop.args,
+    threshold: DOOM_LOOP_THRESHOLD,
+  };
+  const content = `Detected ${DOOM_LOOP_THRESHOLD} identical calls to ${loop.name}. Stopping the tool-call loop. Produce the best answer you can with what you have.`;
+
+  const [row] = await ctx.sql<{ id: string }[]>`
+    INSERT INTO messages (session_id, chat_id, role, content, status, created_at, metadata)
+    VALUES (${sessionId}, ${chatId}, 'system', ${content}, 'complete', clock_timestamp(), ${ctx.sql.json(metadata as never)})
+    RETURNING id
+  `;
+
+  // Standard frame sequence — same as cap-hit sentinel — so
+  // useSessionStream's reducer appends the row via the existing path.
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: row!.id,
+    chat_id: chatId,
+    role: 'system',
+  });
+  ctx.publish(sessionId, {
+    type: 'delta',
+    message_id: row!.id,
+    chat_id: chatId,
+    content,
+  });
+  ctx.publish(sessionId, {
+    type: 'message_complete',
+    message_id: row!.id,
+    chat_id: chatId,
+    metadata,
+  });
+}

apps/server/src/services/inference/sentinels.ts

@@ -0,0 +1,53 @@
+import type { Message, ToolCall } from '../../types/api.js';
+
+// v1.11.6: doom-loop guard. When the model calls the same tool with the
+// same arguments DOOM_LOOP_THRESHOLD times in a row within one user-message
+// turn, abort the recursion and run the same wrap-up summary path as the
+// cap-hit case. Ported from opencode (DOOM_LOOP_THRESHOLD in
+// session/processor.ts). Threshold of 3 is the smallest value that doesn't
+// false-positive on a model that retries once after a transient error.
+export const DOOM_LOOP_THRESHOLD = 3;
+
+// Returns the name + args of the looping tool when the LAST
+// DOOM_LOOP_THRESHOLD entries in `recentToolCalls` are identical (same name
+// AND deep-equal args via JSON.stringify). Returns null otherwise.
+// Pure; exported for unit-test access.
+export function detectDoomLoop(
+  recentToolCalls: ToolCall[],
+): { name: string; args: Record<string, unknown> } | null {
+  if (recentToolCalls.length < DOOM_LOOP_THRESHOLD) return null;
+  const last = recentToolCalls.slice(-DOOM_LOOP_THRESHOLD);
+  const ref = last[0]!;
+  const refArgs = JSON.stringify(ref.args);
+  for (let i = 1; i < last.length; i++) {
+    const tc = last[i]!;
+    if (tc.name !== ref.name) return null;
+    if (JSON.stringify(tc.args) !== refArgs) return null;
+  }
+  return { name: ref.name, args: ref.args };
+}
+
+export function isCapHitSentinel(m: Message): boolean {
+  return (
+    m.role === 'system' &&
+    m.metadata !== null &&
+    typeof m.metadata === 'object' &&
+    (m.metadata as { kind?: unknown }).kind === 'cap_hit'
+  );
+}
+
+// v1.11.6: parallel predicate. Same UI-only semantics as cap-hit sentinels —
+// never sent to the LLM (filtered by buildMessagesPayload through the
+// isAnySentinel check below).
+export function isDoomLoopSentinel(m: Message): boolean {
+  return (
+    m.role === 'system' &&
+    m.metadata !== null &&
+    typeof m.metadata === 'object' &&
+    (m.metadata as { kind?: unknown }).kind === 'doom_loop'
+  );
+}
+
+export function isAnySentinel(m: Message): boolean {
+  return isCapHitSentinel(m) || isDoomLoopSentinel(m);
+}

apps/server/src/services/inference/stream-phase.ts

@@ -0,0 +1,380 @@
+import type {
+  Agent,
+  Session,
+  ToolCall,
+} from '../../types/api.js';
+import * as modelContext from '../model-context.js';
+import { toolJsonSchemas, type ToolJsonSchema } from '../tools.js';
+import type { OpenAiMessage } from './payload.js';
+import {
+  XML_TOOL_CLOSE,
+  XML_TOOL_OPEN,
+  parseXmlToolCall,
+  partialXmlOpenerStart,
+} from './xml-parser.js';
+import { DB_FLUSH_INTERVAL_MS, type StreamPhaseState } from './types.js';
+import type {
+  InferenceContext,
+  StreamResult,
+  TurnArgs,
+} from './turn.js';
+
+interface ChatCompletionDelta {
+  role?: string;
+  content?: string | null;
+  tool_calls?: Array<{
+    index: number;
+    id?: string;
+    type?: 'function';
+    function?: { name?: string; arguments?: string };
+  }>;
+}
+
+interface ChatCompletionChunk {
+  choices?: Array<{
+    delta: ChatCompletionDelta;
+    finish_reason: string | null;
+  }>;
+  usage?: {
+    prompt_tokens?: number;
+    completion_tokens?: number;
+    total_tokens?: number;
+  };
+}
+
+interface StreamOptions {
+  // null = omit tools entirely (compact phase); [] = caller stripped all tools
+  // (rare; we still omit from the request body to avoid OpenAI 400).
+  tools: ToolJsonSchema[] | null;
+  temperature?: number;
+}
+
+async function* sseLines(stream: ReadableStream<Uint8Array>): AsyncGenerator<string> {
+  const reader = stream.getReader();
+  const decoder = new TextDecoder('utf-8');
+  let buffer = '';
+  try {
+    while (true) {
+      const { value, done } = await reader.read();
+      if (done) break;
+      buffer += decoder.decode(value, { stream: true });
+      let idx;
+      while ((idx = buffer.indexOf('\n')) >= 0) {
+        const line = buffer.slice(0, idx).replace(/\r$/, '');
+        buffer = buffer.slice(idx + 1);
+        if (line.length === 0) continue;
+        yield line;
+      }
+    }
+    if (buffer.length > 0) yield buffer;
+  } finally {
+    reader.releaseLock();
+  }
+}
+
+// v1.10.5 Qwen-coder XML fallback. Some local models (notably qwen3-coder via
+// llama-swap) emit tool calls as inline XML inside delta.content rather than
+// the structured delta.tool_calls field. The XML shape is:
+//   <tool_call>
+//   <function=NAME>
+//   <parameter=KEY>
+//   VALUE
+//   </parameter>
+//   ...more parameters...
+//   </function>
+//   </tool_call>
+// Multiple <tool_call> blocks may appear back-to-back; they never nest.
+// streamCompletion buffers delta.content, extracts complete blocks, parses
+// them via parseXmlToolCall, and pushes synthetic entries into the existing
+// toolCallsBuffer alongside any native JSON-format tool calls.
+export async function streamCompletion(
+  ctx: InferenceContext,
+  model: string,
+  messages: OpenAiMessage[],
+  opts: StreamOptions,
+  onDelta: (content: string) => void,
+  onUsage: ((prompt: number | null, completion: number | null) => void) | undefined,
+  signal?: AbortSignal
+): Promise<StreamResult> {
+  const body: Record<string, unknown> = {
+    model,
+    messages,
+    stream: true,
+    stream_options: { include_usage: true },
+  };
+  if (opts.tools && opts.tools.length > 0) {
+    body['tools'] = opts.tools;
+    body['tool_choice'] = 'auto';
+  }
+  if (typeof opts.temperature === 'number') {
+    body['temperature'] = opts.temperature;
+  }
+
+  const res = await fetch(`${ctx.config.LLAMA_SWAP_URL}/v1/chat/completions`, {
+    method: 'POST',
+    headers: { 'Content-Type': 'application/json' },
+    body: JSON.stringify(body),
+    signal,
+  });
+  if (!res.ok || !res.body) {
+    const text = await res.text().catch(() => '');
+    throw new Error(`llama-swap returned ${res.status}: ${text.slice(0, 200)}`);
+  }
+
+  let content = '';
+  // v1.10.5: holds delta.content bytes that may contain a partial XML tool
+  // call. Anything not part of a (possibly forming) <tool_call>…</tool_call>
+  // pair is flushed to content + onDelta as soon as we know it's safe.
+  let pendingBuffer = '';
+  let finishReason: string | null = null;
+  let promptTokens: number | null = null;
+  let completionTokens: number | null = null;
+  const toolCallsBuffer = new Map<number, { id: string; name: string; argsText: string }>();
+
+  for await (const line of sseLines(res.body)) {
+    if (!line.startsWith('data:')) continue;
+    const payload = line.slice(5).trim();
+    if (payload === '[DONE]') break;
+    let parsed: ChatCompletionChunk;
+    try {
+      parsed = JSON.parse(payload);
+    } catch {
+      continue;
+    }
+
+    if (parsed.usage) {
+      if (typeof parsed.usage.prompt_tokens === 'number') {
+        promptTokens = parsed.usage.prompt_tokens;
+      }
+      if (typeof parsed.usage.completion_tokens === 'number') {
+        completionTokens = parsed.usage.completion_tokens;
+      }
+      onUsage?.(promptTokens, completionTokens);
+    }
+    // v1.11.3: removed dead `parsed.timings.n_ctx` read. llama-server's
+    // streaming completion does NOT emit n_ctx in timings (verified
+    // empirically); the authoritative source is llama-swap's
+    // /upstream/<model>/props endpoint, fetched per-turn via
+    // model-context.getModelContext() at the finalization sites below.
+
+    const choice = parsed.choices?.[0];
+    if (!choice) continue;
+    const delta = choice.delta ?? {};
+    if (typeof delta.content === 'string' && delta.content.length > 0) {
+      // v1.10.5 XML fallback. Append, then extract any complete tool_call
+      // blocks before deciding what's safe to flush as visible content.
+      pendingBuffer += delta.content;
+      while (true) {
+        const startIdx = pendingBuffer.indexOf(XML_TOOL_OPEN);
+        if (startIdx === -1) break;
+        const closeIdx = pendingBuffer.indexOf(XML_TOOL_CLOSE, startIdx);
+        if (closeIdx === -1) break;
+        const blockEnd = closeIdx + XML_TOOL_CLOSE.length;
+        const block = pendingBuffer.slice(startIdx, blockEnd);
+        // Any text before the opener is plain content — flush it now.
+        if (startIdx > 0) {
+          const before = pendingBuffer.slice(0, startIdx);
+          content += before;
+          onDelta(before);
+        }
+        const parsedCall = parseXmlToolCall(block);
+        if (parsedCall) {
+          const synthIdx = toolCallsBuffer.size;
+          toolCallsBuffer.set(synthIdx, {
+            id: `xml_call_${synthIdx}`,
+            name: parsedCall.name,
+            argsText: JSON.stringify(parsedCall.args),
+          });
+        }
+        // If parsing failed we still drop the block — emitting unparseable
+        // XML to the chat would look worse than silently swallowing it.
+        pendingBuffer = pendingBuffer.slice(blockEnd);
+      }
+      // After all complete blocks are out, hold back any (partial or full)
+      // unclosed opener; flush the rest.
+      const partialIdx = partialXmlOpenerStart(pendingBuffer);
+      if (partialIdx >= 0) {
+        if (partialIdx > 0) {
+          const flush = pendingBuffer.slice(0, partialIdx);
+          content += flush;
+          onDelta(flush);
+        }
+        pendingBuffer = pendingBuffer.slice(partialIdx);
+      } else if (pendingBuffer.length > 0) {
+        content += pendingBuffer;
+        onDelta(pendingBuffer);
+        pendingBuffer = '';
+      }
+    }
+    if (Array.isArray(delta.tool_calls)) {
+      for (const tc of delta.tool_calls) {
+        const idx = tc.index;
+        const existing = toolCallsBuffer.get(idx) ?? { id: '', name: '', argsText: '' };
+        if (tc.id) existing.id = tc.id;
+        if (tc.function?.name) existing.name = tc.function.name;
+        if (typeof tc.function?.arguments === 'string') existing.argsText += tc.function.arguments;
+        toolCallsBuffer.set(idx, existing);
+      }
+    }
+    if (choice.finish_reason) finishReason = choice.finish_reason;
+  }
+
+  // v1.10.5: if the stream ended mid-XML (e.g. model truncated, no closer
+  // ever arrived), flush whatever was buffered as plain content so it isn't
+  // silently dropped. Better to show a stray `<tool_call>` than vanish text.
+  if (pendingBuffer.length > 0) {
+    content += pendingBuffer;
+    onDelta(pendingBuffer);
+    pendingBuffer = '';
+  }
+
+  const toolCalls: ToolCall[] = [];
+  for (const [, t] of [...toolCallsBuffer.entries()].sort(([a], [b]) => a - b)) {
+    let args: Record<string, unknown> = {};
+    if (t.argsText.length > 0) {
+      try {
+        args = JSON.parse(t.argsText);
+      } catch {
+        args = { _raw: t.argsText };
+      }
+    }
+    toolCalls.push({ id: t.id || `call_${toolCalls.length}`, name: t.name, args });
+  }
+
+  return { finishReason, content, toolCalls, promptTokens, completionTokens };
+}
+
+export async function executeStreamPhase(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  session: Session,
+  messages: OpenAiMessage[],
+  state: StreamPhaseState,
+  agent: Agent | null,
+  // v1.11.8: when false, web_search and web_fetch are stripped from the
+  // tool list sent to the LLM, so the model can't even attempt them.
+  webToolsEnabled: boolean,
+): Promise<StreamResult> {
+  const { sessionId, chatId, assistantMessageId, signal } = args;
+
+  const startedRow = await ctx.sql<{ started_at: string }[]>`
+    UPDATE messages
+    SET started_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING started_at
+  `;
+  state.startedAt = startedRow[0]?.started_at ?? null;
+
+  ctx.publish(sessionId, {
+    type: 'message_started',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    role: 'assistant',
+  });
+
+  let pendingFlushTimer: NodeJS.Timeout | null = null;
+  let flushPromise: Promise<unknown> = Promise.resolve();
+
+  const flushNow = () => {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    const snapshot = state.accumulated;
+    flushPromise = flushPromise.then(() =>
+      ctx.sql`UPDATE messages SET content = ${snapshot} WHERE id = ${assistantMessageId}`
+    );
+  };
+
+  const scheduleFlush = () => {
+    if (pendingFlushTimer) return;
+    pendingFlushTimer = setTimeout(() => {
+      pendingFlushTimer = null;
+      flushNow();
+    }, DB_FLUSH_INTERVAL_MS);
+  };
+
+  // Tool whitelist: if an agent is set, filter the global tool list to only the
+  // tool names it allows. Unknown names in agent.tools are dropped silently
+  // (handled here by intersection). When no agent: send all tools.
+  // v1.11.8: a second filter strips web_search + web_fetch unless the chat
+  // has them explicitly enabled. Counts as an opt-in security boundary: the
+  // model can't summon a tool that wasn't offered to it.
+  const WEB_TOOL_NAMES: ReadonlySet<string> = new Set(['web_search', 'web_fetch']);
+  const effectiveTools: ToolJsonSchema[] = (agent
+    ? toolJsonSchemas().filter((t) => agent.tools.includes(t.function.name))
+    : toolJsonSchemas()
+  ).filter((t) => webToolsEnabled || !WEB_TOOL_NAMES.has(t.function.name));
+  const effectiveTemperature = agent?.temperature;
+
+  // v1.12.2: ctx_max lookup is cached after the first hit per model, so this
+  // is a Map probe in steady state. We capture nCtx once at the top of the
+  // stream so the throttled usage publish doesn't refetch each tick.
+  const mctxForStream = await modelContext.getModelContext(session.model);
+  const nCtxForStream = mctxForStream?.n_ctx ?? null;
+
+  // v1.12.2: throttle live usage publishes to ~500ms. The model can land
+  // dozens of usage frames per second; without a throttle the WS turns into
+  // a firehose for a few KB savings on each render.
+  const USAGE_THROTTLE_MS = 500;
+  let lastUsageAt = 0;
+  let pendingUsage: { p: number | null; c: number | null } | null = null;
+  let usageTimer: NodeJS.Timeout | null = null;
+  const flushUsage = () => {
+    if (!pendingUsage) return;
+    const { p, c } = pendingUsage;
+    pendingUsage = null;
+    lastUsageAt = Date.now();
+    ctx.publish(sessionId, {
+      type: 'usage',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      completion_tokens: c,
+      ctx_used: p,
+      ctx_max: nCtxForStream,
+    });
+  };
+
+  try {
+    return await streamCompletion(
+      ctx,
+      session.model,
+      messages,
+      { tools: effectiveTools, temperature: effectiveTemperature },
+      (delta) => {
+        state.accumulated += delta;
+        ctx.publish(sessionId, {
+          type: 'delta',
+          message_id: assistantMessageId,
+          chat_id: chatId,
+          content: delta,
+        });
+        ctx.log.debug({ sessionId, delta }, 'inference delta');
+        scheduleFlush();
+      },
+      (prompt, completion) => {
+        pendingUsage = { p: prompt, c: completion };
+        const elapsed = Date.now() - lastUsageAt;
+        if (elapsed >= USAGE_THROTTLE_MS) {
+          flushUsage();
+        } else if (!usageTimer) {
+          usageTimer = setTimeout(() => {
+            usageTimer = null;
+            flushUsage();
+          }, USAGE_THROTTLE_MS - elapsed);
+        }
+      },
+      signal
+    );
+  } finally {
+    if (pendingFlushTimer) {
+      clearTimeout(pendingFlushTimer);
+      pendingFlushTimer = null;
+    }
+    if (usageTimer) {
+      clearTimeout(usageTimer);
+      usageTimer = null;
+    }
+    await flushPromise;
+  }
+}

apps/server/src/services/inference/tool-phase.ts

@@ -0,0 +1,213 @@
+import type { Session, ToolCall } from '../../types/api.js';
+import * as modelContext from '../model-context.js';
+import { PathScopeError } from '../path_guard.js';
+import { TOOLS_BY_NAME } from '../tools.js';
+import { maybeFlagForCompaction } from './payload.js';
+import type {
+  InferenceContext,
+  StreamResult,
+  TurnArgs,
+} from './turn.js';
+// v1.12.4: ESM value-import cycle. executeToolPhase recurses into
+// runAssistantTurn which lives in inference.ts. The cycle is safe because
+// the reference is read at call time (inside an async function body), not
+// at module top-level. Node + tsc resolve this cleanly.
+import { runAssistantTurn } from './turn.js';
+
+async function executeToolCall(
+  projectRoot: string,
+  toolCall: ToolCall
+): Promise<{ output: unknown; truncated: boolean; error?: string }> {
+  const tool = TOOLS_BY_NAME[toolCall.name];
+  if (!tool) {
+    return { output: null, truncated: false, error: `unknown tool: ${toolCall.name}` };
+  }
+  const parsed = tool.inputSchema.safeParse(toolCall.args);
+  if (!parsed.success) {
+    // v1.12 Track B.2: enrich the zod-reject path so the model sees a
+    // one-line, tool-named hint ("tool 'search_symbols' rejected — query:
+    // Required") instead of a JSON blob of flatten output. Higher recovery
+    // rate on the next turn; doom-loop guard still bounds infinite retries.
+    // The cast is because tool.inputSchema is ZodType<unknown>, so zod can't
+    // statically narrow flatten()'s fieldErrors key set — but the runtime
+    // shape is the standard { formErrors: string[]; fieldErrors: Record<...> }.
+    const flatten = parsed.error.flatten() as {
+      formErrors: string[];
+      fieldErrors: Record<string, string[] | undefined>;
+    };
+    const fieldErrors = Object.entries(flatten.fieldErrors)
+      .map(([field, errs]) => `${field}: ${errs?.[0] ?? 'invalid'}`)
+      .join('; ');
+    const formError = flatten.formErrors[0];
+    const hint = fieldErrors || formError || 'unknown validation error';
+    return {
+      output: null,
+      truncated: false,
+      error: `tool '${toolCall.name}' rejected — ${hint}`,
+    };
+  }
+  try {
+    const output = await tool.execute(parsed.data, projectRoot);
+    const truncated =
+      typeof output === 'object' && output !== null && 'truncated' in output
+        ? Boolean((output as { truncated: unknown }).truncated)
+        : false;
+    return { output, truncated };
+  } catch (err) {
+    if (err instanceof PathScopeError) {
+      return { output: null, truncated: false, error: err.message };
+    }
+    return {
+      output: null,
+      truncated: false,
+      error: err instanceof Error ? err.message : String(err),
+    };
+  }
+}
+
+export async function executeToolPhase(
+  ctx: InferenceContext,
+  args: TurnArgs,
+  result: StreamResult,
+  startedAt: string | null,
+  session: Session,
+  projectRoot: string
+): Promise<void> {
+  const { sessionId, chatId, assistantMessageId, toolsUsed, signal } = args;
+  const { content, toolCalls, promptTokens, completionTokens } = result;
+
+  // v1.11.3: ctx_max comes from llama-swap /upstream/<model>/props, not the
+  // streaming completion (which doesn't emit n_ctx). getModelContext caches
+  // the positive lookup for the process lifetime, so this is a single Map
+  // hit after the first invocation per model.
+  const mctx = await modelContext.getModelContext(session.model);
+  const nCtx = mctx?.n_ctx ?? null;
+
+  const [updated] = await ctx.sql<
+    { tokens_used: number | null; ctx_used: number | null; ctx_max: number | null; finished_at: string | null }[]
+  >`
+    UPDATE messages
+    SET content = ${content},
+        status = 'complete',
+        tool_calls = ${ctx.sql.json(toolCalls as never)},
+        tokens_used = ${completionTokens},
+        ctx_used = ${promptTokens},
+        ctx_max = ${nCtx},
+        finished_at = clock_timestamp()
+    WHERE id = ${assistantMessageId}
+    RETURNING tokens_used, ctx_used, ctx_max, finished_at
+  `;
+  // v1.11: flag for compaction if this turn pushed us over the usable budget.
+  // We never compact mid-loop (the recursive runAssistantTurn keeps tools
+  // flowing); the flag fires on the NEXT turn's pre-fetch hook above.
+  await maybeFlagForCompaction(ctx, chatId, updated);
+  const [toolSessRow] = await ctx.sql<{ project_id: string; name: string; updated_at: string }[]>`
+    UPDATE sessions SET updated_at = clock_timestamp()
+    WHERE id = ${sessionId}
+    RETURNING project_id, name, updated_at
+  `;
+  ctx.publishUser({ type: 'session_updated', session_id: sessionId, project_id: toolSessRow!.project_id, name: toolSessRow!.name, updated_at: toolSessRow!.updated_at });
+  for (const tc of toolCalls) {
+    ctx.publish(sessionId, {
+      type: 'tool_call',
+      message_id: assistantMessageId,
+      chat_id: chatId,
+      tool_call: tc,
+    });
+  }
+  ctx.publish(sessionId, {
+    type: 'message_complete',
+    message_id: assistantMessageId,
+    chat_id: chatId,
+    tokens_used: updated?.tokens_used ?? null,
+    ctx_used: updated?.ctx_used ?? null,
+    ctx_max: updated?.ctx_max ?? null,
+    started_at: startedAt,
+    finished_at: updated?.finished_at ?? null,
+    model: session.model,
+  });
+
+  // Batch 9.7: ask_user_input pauses the loop. The tool row is still inserted
+  // (the answer endpoint needs a target row to UPDATE), but tool_results is
+  // pre-stamped with output=null as a "pending" sentinel and no tool_result
+  // frame goes out — the card renders from the tool_call frame alone. Mixed
+  // batches still execute the other tools normally.
+  ctx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'tool_running', at: new Date().toISOString() });
+  let pausingForUserInput = false;
+  await Promise.all(
+    toolCalls.map(async (tc) => {
+      const [toolRow] = await ctx.sql<{ id: string }[]>`
+        INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+        VALUES (${sessionId}, ${chatId}, 'tool', '', 'complete', clock_timestamp())
+        RETURNING id
+      `;
+      const toolMessageId = toolRow!.id;
+      if (tc.name === 'ask_user_input') {
+        pausingForUserInput = true;
+        const sentinel = { tool_call_id: tc.id, output: null, truncated: false };
+        await ctx.sql`
+          UPDATE messages
+          SET tool_results = ${ctx.sql.json(sentinel as never)}
+          WHERE id = ${toolMessageId}
+        `;
+        return;
+      }
+      const tres = await executeToolCall(projectRoot, tc);
+      const stored = {
+        tool_call_id: tc.id,
+        output: tres.output,
+        truncated: tres.truncated,
+        ...(tres.error ? { error: tres.error } : {}),
+      };
+      await ctx.sql`
+        UPDATE messages
+        SET tool_results = ${ctx.sql.json(stored as never)}
+        WHERE id = ${toolMessageId}
+      `;
+      ctx.publish(sessionId, {
+        type: 'tool_result',
+        tool_message_id: toolMessageId,
+        chat_id: chatId,
+        tool_call_id: tc.id,
+        output: tres.output,
+        truncated: tres.truncated,
+        ...(tres.error ? { error: tres.error } : {}),
+      });
+    })
+  );
+
+  if (pausingForUserInput) {
+    ctx.publishUser({
+      type: 'chat_status',
+      chat_id: chatId,
+      status: 'waiting_for_input',
+      at: new Date().toISOString(),
+    });
+    ctx.log.info(
+      { sessionId, chatId, assistantMessageId },
+      'inference paused awaiting user input',
+    );
+    return;
+  }
+
+  const [nextAssistant] = await ctx.sql<{ id: string }[]>`
+    INSERT INTO messages (session_id, chat_id, role, content, status, created_at)
+    VALUES (${sessionId}, ${chatId}, 'assistant', '', 'streaming', clock_timestamp())
+    RETURNING id
+  `;
+  await runAssistantTurn(ctx, {
+    sessionId,
+    chatId,
+    assistantMessageId: nextAssistant!.id,
+    // v1.8.2: charge this turn's actual tool invocations against the budget.
+    // One assistant message can emit multiple tool_calls, so we add the run
+    // count, not 1. The next turn's budget check sees the cumulative total.
+    toolsUsed: toolsUsed + result.toolCalls.length,
+    // v1.11.6: append the just-executed tool calls to the per-turn history
+    // so the next runAssistantTurn's doom-loop check can see them. We don't
+    // cap the array length here — per-turn budgets keep it bounded
+    // (typically <30 entries), and slicing happens inside detectDoomLoop.
+    recentToolCalls: [...args.recentToolCalls, ...result.toolCalls],
+    signal,
+  });
+}

apps/server/src/services/inference/turn.ts

@@ -0,0 +1,326 @@
+import type { FastifyBaseLogger } from 'fastify';
+import type { Sql } from '../../db.js';
+import type { Config } from '../../config.js';
+import type {
+  Agent,
+  ErrorReason,
+  Message,
+  MessageMetadata,
+  Project,
+  Session,
+  ToolCall,
+  UserStreamFrame,
+} from '../../types/api.js';
+import { ALL_TOOLS } from '../tools.js';
+import { resolveProjectRoot } from '../path_guard.js';
+import { maybeAutoNameChat } from '../auto_name.js';
+import { getAgentById } from '../agents.js';
+import * as compaction from '../compaction.js';
+import * as modelContext from '../model-context.js';
+import type { Broker } from '../broker.js';
+import { resolveToolBudget } from './budget.js';
+import {
+  DOOM_LOOP_THRESHOLD,
+  detectDoomLoop,
+} from './sentinels.js';
+import {
+  buildMessagesPayload,
+  loadContext,
+} from './payload.js';
+import {
+  finalizeCompletion,
+  handleAbortOrError,
+} from './error-handler.js';
+import {
+  executeStreamPhase,
+  streamCompletion,
+} from './stream-phase.js';
+import { executeToolPhase } from './tool-phase.js';
+import { DB_FLUSH_INTERVAL_MS, type StreamPhaseState } from './types.js';
+import {
+  runCapHitSummary,
+  runDoomLoopSummary,
+} from './sentinel-summaries.js';
+
+// v1.12.4: re-exported so external callers (tests, future consumers) keep
+// importing from services/inference.js as the public surface.
+export { detectDoomLoop, DOOM_LOOP_THRESHOLD } from './sentinels.js';
+export { buildMessagesPayload } from './payload.js';
+
+export interface InferenceFrame {
+  type:
+    | 'message_started'
+    | 'delta'
+    | 'tool_call'
+    | 'tool_result'
+    | 'message_complete'
+    | 'usage'
+    | 'messages_deleted'
+    | 'session_renamed'
+    | 'chat_renamed'
+    | 'error';
+  message_id?: string;
+  message_ids?: string[];
+  chat_id?: string;
+  tool_message_id?: string;
+  tool_call_id?: string;
+  // v1.8.2: 'system' added so cap-hit sentinel messages can announce themselves
+  // through the normal message_started → delta → message_complete sequence.
+  role?: 'assistant' | 'tool' | 'user' | 'system';
+  content?: string;
+  tool_call?: ToolCall;
+  output?: unknown;
+  truncated?: boolean;
+  error?: string;
+  // v1.8.2: structured error reason. Set on `type: 'error'` so the UI can
+  // surface a specific message; `error` stays the human-readable text.
+  reason?: ErrorReason;
+  // v1.8.2: piggybacks on `message_complete` so static or terminally-resolved
+  // messages can carry their persisted metadata to the live stream without a
+  // refetch (sentinels carry { kind: 'cap_hit', ... }; failed messages carry
+  // { kind: 'error', ... }).
+  metadata?: MessageMetadata | null;
+  tokens_used?: number | null;
+  ctx_used?: number | null;
+  ctx_max?: number | null;
+  completion_tokens?: number | null;
+  started_at?: string | null;
+  finished_at?: string | null;
+  model?: string;
+  session_id?: string;
+  name?: string;
+}
+
+export type FramePublisher = (sessionId: string, frame: InferenceFrame) => void;
+
+export interface InferenceContext {
+  sql: Sql;
+  config: Config;
+  log: FastifyBaseLogger;
+  publish: FramePublisher;
+  publishUser: (frame: UserStreamFrame) => void;
+  // v1.11: passed through so compaction.process can publish 'compacted'
+  // frames on the same session WS channel useSessionStream subscribes to.
+  // Compaction is the only path that needs the raw broker handle (regular
+  // inference goes through `publish`); keeping a separate field avoids
+  // tempting other code paths into bypassing the session-id binding.
+  broker: Broker;
+}
+
+// v1.12.4: payload assembly extracted to ./inference/payload.ts (tests
+// import buildMessagesPayload from this module, so a re-export below
+// preserves the public surface). Stream + tool phases extracted to
+// ./inference/stream-phase.ts and ./inference/tool-phase.ts.
+
+export interface StreamResult {
+  finishReason: string | null;
+  content: string;
+  toolCalls: ToolCall[];
+  promptTokens: number | null;
+  completionTokens: number | null;
+}
+
+
+export interface TurnArgs {
+  sessionId: string;
+  chatId: string;
+  assistantMessageId: string;
+  // v1.8.2: cumulative tool calls executed this run. Compared against the
+  // resolved budget at the top of each turn. Replaces the older `depth`
+  // counter (which counted iterations, not invocations).
+  toolsUsed: number;
+  // v1.11.6: ordered tool calls executed in this user-message turn (across
+  // recursive runAssistantTurn invocations). Reset to [] at user-message
+  // boundaries by runInference, same as toolsUsed. Doom-loop check at the
+  // top of runAssistantTurn slices the last DOOM_LOOP_THRESHOLD entries.
+  recentToolCalls: ToolCall[];
+  signal: AbortSignal | undefined;
+}
+
+
+export async function runAssistantTurn(
+  ctx: InferenceContext,
+  args: TurnArgs,
+): Promise<void> {
+  const { sessionId, chatId } = args;
+
+  // v1.11: if the prior turn flagged this chat for compaction, run it first
+  // so loadContext below reads the post-compaction history. We swallow
+  // compaction failures (clearing the flag so we don't loop) and proceed
+  // with the un-compacted history — a slow turn that hits the model's
+  // hard limit is recoverable; a dead session is not.
+  const chatFlag = await ctx.sql<{ needs_compaction: boolean }[]>`
+    SELECT needs_compaction FROM chats WHERE id = ${chatId}
+  `;
+  if (chatFlag[0]?.needs_compaction) {
+    try {
+      await compaction.process({
+        sql: ctx.sql,
+        config: ctx.config,
+        log: ctx.log,
+        broker: ctx.broker,
+        chatId,
+      });
+    } catch (err) {
+      ctx.log.warn({ err, chatId }, 'auto-compaction failed; clearing flag and proceeding');
+      await ctx.sql`UPDATE chats SET needs_compaction = false WHERE id = ${chatId}`;
+    }
+  }
+
+  const loaded = await loadContext(ctx.sql, sessionId, chatId);
+  if (!loaded) {
+    ctx.log.warn({ sessionId }, 'inference: session or project missing');
+    return;
+  }
+  const { session, project, history } = loaded;
+  const projectRoot = await resolveProjectRoot(project.path);
+  // Agent resolution is per-turn so PATCH agent_id mid-conversation takes
+  // effect on the next message. Unknown agent_id returns null silently —
+  // session falls back to base prompt + all tools + default temperature.
+  const agent = session.agent_id
+    ? await getAgentById(project.path, session.agent_id)
+    : null;
+
+  // v1.8.2: cap-hit replaces the older "tool loop depth exceeded" failure.
+  // When we've already burned the budget *before* this turn even runs, we
+  // skip straight to the summary flow — the in-flight assistant message slot
+  // gets reused for the wrap-up reply instead of being marked failed.
+  const budget = resolveToolBudget(agent);
+  if (args.toolsUsed >= budget) {
+    await runCapHitSummary(ctx, args, session, project, history, agent, budget);
+    return;
+  }
+
+  // v1.11.6: doom-loop guard. Detected BEFORE the budget cap (the model can
+  // burn through 3 identical calls long before the 15-call budget fires).
+  // Same in-flight-slot-reuse pattern as runCapHitSummary — wrap-up reply
+  // lands in args.assistantMessageId, then a doom_loop sentinel is inserted
+  // to make the abort visible in the chat history.
+  const loop = detectDoomLoop(args.recentToolCalls);
+  if (loop) {
+    await runDoomLoopSummary(ctx, args, session, project, history, agent, loop);
+    return;
+  }
+
+  const messages = await buildMessagesPayload(session, project, history, agent);
+
+  // v1.11.8: resolve per-chat web-tools opt-in. Tri-state on the wire:
+  //   - session.web_search_enabled = null → inherit project default
+  //   - session.web_search_enabled = true/false → explicit
+  // Both web_search and web_fetch are gated by this single flag (the UI
+  // label is "Enable web search and fetch" — same store, both tools).
+  // Default is false unless explicitly opted in, matching the v1.9
+  // plumbing intent ("inert until Batch 8 ships the actual tools").
+  const webToolsEnabled =
+    session.web_search_enabled ?? project.default_web_search_enabled ?? false;
+
+  const state: StreamPhaseState = { accumulated: '', startedAt: null };
+  let result: StreamResult;
+  try {
+    result = await executeStreamPhase(ctx, args, session, messages, state, agent, webToolsEnabled);
+  } catch (err) {
+    await handleAbortOrError(ctx, args, state.accumulated, err);
+    return;
+  }
+
+  if (result.toolCalls.length > 0) {
+    await executeToolPhase(ctx, args, result, state.startedAt, session, projectRoot);
+    return;
+  }
+
+  await finalizeCompletion(ctx, args, result, state.startedAt, session);
+}
+
+export async function runInference(
+  ctx: InferenceContext,
+  sessionId: string,
+  chatId: string,
+  assistantMessageId: string,
+  signal?: AbortSignal
+): Promise<void> {
+  // v1.8.2: every fresh inference (initial send, regenerate, force_send,
+  // continue) starts with a clean budget. Tool-call accumulation across
+  // Continue invocations is what the hard ceiling guards against, not the
+  // per-call budget.
+  // v1.11.6: recentToolCalls also resets — doom-loop detection is scoped
+  // to a single user-message turn, so a Continue starts with no history.
+  return runAssistantTurn(ctx, {
+    sessionId,
+    chatId,
+    assistantMessageId,
+    toolsUsed: 0,
+    recentToolCalls: [],
+    signal,
+  });
+}
+
+// v1.8.2: cap-hit summary flow. Called instead of erroring when the loop
+// hits its budget. Reuses the in-flight assistant message slot to stream a
+// short wrap-up reply with the synthetic note prepended and tools disabled,
+// then always inserts a cap_hit sentinel afterward (regardless of summary
+// outcome) so the UI can show a Continue affordance.
+interface InferenceRegistration {
+  controller: AbortController;
+  completed: Promise<void>;
+}
+
+export function createInferenceRunner(
+  ctx: Omit<InferenceContext, 'publishUser'>,
+  publishUserFn: (user: string, frame: UserStreamFrame) => void
+) {
+  const registry = new Map<string, InferenceRegistration>();
+
+  return {
+    enqueue(sessionId: string, chatId: string, assistantMessageId: string, user: string) {
+      const callCtx: InferenceContext = {
+        ...ctx,
+        publishUser: (frame) => publishUserFn(user, frame),
+        // v1.11: broker comes in via ctx (set at registration time). Repeated
+        // here so the destructure carries it onto the per-call ctx without
+        // having to add it to every enqueue/cancel signature individually.
+        broker: ctx.broker,
+      };
+      // v1.8 mobile-tabs: announce working before the async loop starts so
+      // every device subscribed to the user channel sees the amber dot.
+      callCtx.publishUser({ type: 'chat_status', chat_id: chatId, status: 'streaming', at: new Date().toISOString() });
+      const controller = new AbortController();
+      let resolveCompleted!: () => void;
+      const completed = new Promise<void>((res) => { resolveCompleted = res; });
+      const registration: InferenceRegistration = { controller, completed };
+      registry.set(chatId, registration);
+      void (async () => {
+        try {
+          await runInference(callCtx, sessionId, chatId, assistantMessageId, controller.signal);
+          setImmediate(() => {
+            void maybeAutoNameChat(callCtx, chatId, sessionId).catch((err: Error) => {
+              callCtx.log.warn({ err, chatId }, 'auto-name failed');
+            });
+          });
+        } catch (err) {
+          callCtx.log.error({ err }, 'unhandled inference error');
+        } finally {
+          resolveCompleted();
+          // Only clear our own registration; a force-send may have replaced it.
+          if (registry.get(chatId) === registration) {
+            registry.delete(chatId);
+          }
+        }
+      })();
+    },
+
+    async cancel(_sessionId: string, chatId: string): Promise<boolean> {
+      const reg = registry.get(chatId);
+      if (!reg) return false;
+      reg.controller.abort();
+      // Swallow — we just need to wait for the catch/finally to persist state.
+      await reg.completed.catch(() => {});
+      return true;
+    },
+
+    hasActive(chatId: string): boolean {
+      return registry.has(chatId);
+    },
+  };
+}
+
+export const _toolNames = ALL_TOOLS.map((t) => t.name);

apps/server/src/services/inference/types.ts

@@ -0,0 +1,13 @@
+// v1.12.4: shared inter-phase types/constants for the extracted phase files.
+// Lives here so stream-phase, tool-phase, and the summary functions still in
+// inference.ts can all reference the same definitions without circular imports.
+
+export interface StreamPhaseState {
+  accumulated: string;
+  startedAt: string | null;
+}
+
+// 500ms keeps the DB UPDATE rate bounded under heavy streaming. Used by
+// executeStreamPhase, runCapHitSummary, and runDoomLoopSummary — every site
+// that does a debounced content flush during streaming.
+export const DB_FLUSH_INTERVAL_MS = 500;

apps/server/src/services/inference/xml-parser.ts

@@ -0,0 +1,53 @@
+// v1.10.5: XML-tag tool-call fallback. Some models emit
+// <tool_call><function=foo><parameter=key>value</parameter></function></tool_call>
+// in plain content instead of using the OpenAI tool_calls JSON channel.
+// The streaming loop in inference.ts extracts these blocks via these helpers.
+
+export const XML_TOOL_OPEN = '<tool_call>';
+export const XML_TOOL_CLOSE = '</tool_call>';
+
+export function parseXmlToolCall(
+  block: string,
+): { name: string; args: Record<string, unknown> } | null {
+  const nameMatch = block.match(/<function=([^>]+)>/);
+  if (!nameMatch || !nameMatch[1]) return null;
+  const name = nameMatch[1].trim();
+  if (!name) return null;
+  const args: Record<string, unknown> = {};
+  // Non-greedy body so each <parameter=…>…</parameter> pair is matched
+  // independently even when multiple appear in the same block.
+  const paramRe = /<parameter=([^>]+)>([\s\S]*?)<\/parameter>/g;
+  for (const m of block.matchAll(paramRe)) {
+    const key = (m[1] ?? '').trim();
+    if (!key) continue;
+    const raw = (m[2] ?? '').trim();
+    try {
+      args[key] = JSON.parse(raw);
+    } catch {
+      args[key] = raw;
+    }
+  }
+  return { name, args };
+}
+
+// Locate the first character that begins (or completely contains) an
+// unfinished <tool_call> opener in `s`. Returns -1 when `s` can be flushed
+// to the client in full without risking a partial tag leak.
+//   Case 1: a full `<tool_call>` opener with no matching closer — caller
+//           must keep everything from that index forward until the next
+//           chunk arrives with the closer.
+//   Case 2: `s` ends with a strict prefix of `<tool_call>` (e.g. `<tool_c`).
+//           Caller must keep just that suffix in the buffer.
+// Note: case 1 assumes the calling loop already extracted every complete
+// <tool_call>…</tool_call> pair before reaching this check.
+export function partialXmlOpenerStart(s: string): number {
+  const fullOpener = s.indexOf(XML_TOOL_OPEN);
+  if (fullOpener !== -1) return fullOpener;
+  const lastLt = s.lastIndexOf('<');
+  if (lastLt === -1) return -1;
+  const suffix = s.slice(lastLt);
+  if (XML_TOOL_OPEN.startsWith(suffix) && suffix.length < XML_TOOL_OPEN.length) {
+    return lastLt;
+  }
+  return -1;
+}

apps/web/src/api/client.ts

@@ -180,6 +180,11 @@ export const api = {
       request<{ ok: true }>(`/api/chats/${chatId}/compact`, { method: 'POST' }),
     stop: (chatId: string) =>
       request<{ stopped: boolean }>(`/api/chats/${chatId}/stop`, { method: 'POST' }),
+    discardStale: (chatId: string, messageId: string) =>
+      request<Message>(`/api/chats/${chatId}/discard_stale`, {
+        method: 'POST',
+        body: JSON.stringify({ message_id: messageId }),
+      }),
     forceSend: (chatId: string, content: string) =>
       request<{ user_message_id: string; assistant_message_id: string }>(
         `/api/chats/${chatId}/force_send`,

apps/web/src/components/StaleStreamBanner.tsx

@@ -0,0 +1,34 @@
+interface Props {
+  onRetry: () => void;
+  onDiscard: () => void;
+}
+
+// v1.12.3: shown when an assistant message has been 'streaming' for 60+
+// seconds without new tokens. Lives above ChatInput in ChatPane. Retry
+// discards the stuck row then resends the last user message; Discard just
+// clears the row and drops the dot to idle.
+export function StaleStreamBanner({ onRetry, onDiscard }: Props) {
+  return (
+    <div className="border border-amber-500/30 bg-amber-500/5 rounded-md p-3 mb-2 mx-4 flex items-center justify-between gap-2">
+      <span className="text-sm text-muted-foreground">
+        Previous response didn't complete.
+      </span>
+      <div className="flex gap-2">
+        <button
+          type="button"
+          onClick={onRetry}
+          className="text-xs px-2 py-1 rounded border border-border hover:bg-accent max-md:min-h-[44px] max-md:px-3"
+        >
+          Retry
+        </button>
+        <button
+          type="button"
+          onClick={onDiscard}
+          className="text-xs px-2 py-1 rounded border border-border hover:bg-accent max-md:min-h-[44px] max-md:px-3"
+        >
+          Discard
+        </button>
+      </div>
+    </div>
+  );
+}

apps/web/src/components/panes/ChatPane.tsx

@@ -5,6 +5,7 @@ import { api } from '@/api/client';
 import { useSessionStream } from '@/hooks/useSessionStream';
 import { MessageList } from '@/components/MessageList';
 import { ChatInput } from '@/components/ChatInput';
+import { StaleStreamBanner } from '@/components/StaleStreamBanner';
 import {
   DropdownMenu,
   DropdownMenuContent,
@@ -44,6 +45,38 @@ export function ChatPane({ sessionId, chatId, projectId, agentId, onAgentChange,
 
   const chatMessages = stream.messages.filter((m) => m.chat_id === chatId);
   const streaming = chatMessages.some((m) => m.status === 'streaming');
+
+  // v1.12.3: stale-stream detection. Watches the (at most one) streaming
+  // assistant row. If its content length doesn't grow for STALE_THRESHOLD_MS,
+  // assume the upstream call is dead and surface the recovery banner. We use
+  // content length as the activity signal because every token delta extends
+  // it; last_seq isn't currently bumped per delta.
+  const STALE_THRESHOLD_MS = 60_000;
+  const streamingMsg = chatMessages.find((m) => m.status === 'streaming' && m.role === 'assistant');
+  const streamingId = streamingMsg?.id ?? null;
+  const streamingLen = streamingMsg?.content.length ?? 0;
+  const lastActivityRef = useRef<{ id: string; len: number; at: number } | null>(null);
+  const [stale, setStale] = useState(false);
+  useEffect(() => {
+    if (!streamingId) {
+      lastActivityRef.current = null;
+      setStale(false);
+      return;
+    }
+    const prev = lastActivityRef.current;
+    if (!prev || prev.id !== streamingId || prev.len !== streamingLen) {
+      lastActivityRef.current = { id: streamingId, len: streamingLen, at: Date.now() };
+      setStale(false);
+    }
+    const interval = setInterval(() => {
+      const a = lastActivityRef.current;
+      if (!a) return;
+      if (Date.now() - a.at >= STALE_THRESHOLD_MS) {
+        setStale(true);
+      }
+    }, 5_000);
+    return () => clearInterval(interval);
+  }, [streamingId, streamingLen]);
   // v1.11.5: per-chat model context limit comes from chat.model_context_limit
   // populated by GET /api/sessions/:id/chats. Threaded into ChatInput so
   // ContextBar can render a zero-state before the first assistant message.
@@ -87,6 +120,45 @@ export function ChatPane({ sessionId, chatId, projectId, agentId, onAgentChange,
     }
   }
 
+  const handleDiscardStale = useCallback(async () => {
+    if (!streamingId) return;
+    try {
+      await api.chats.discardStale(chatId, streamingId);
+      setStale(false);
+      lastActivityRef.current = null;
+    } catch (err) {
+      // 409 (race) is benign — the row already terminated some other way.
+      const msg = err instanceof Error ? err.message : 'discard failed';
+      if (!msg.includes('409')) toast.error(msg);
+      setStale(false);
+    }
+  }, [chatId, streamingId]);
+
+  const handleRetryStale = useCallback(async () => {
+    if (!streamingId) return;
+    const lastUser = [...chatMessages].reverse().find((m) => m.role === 'user' && m.kind === 'message');
+    if (!lastUser) {
+      toast.error('no prior user message to retry');
+      return;
+    }
+    try {
+      await api.chats.discardStale(chatId, streamingId);
+    } catch (err) {
+      const msg = err instanceof Error ? err.message : 'discard failed';
+      if (!msg.includes('409')) {
+        toast.error(msg);
+        return;
+      }
+    }
+    setStale(false);
+    lastActivityRef.current = null;
+    try {
+      await api.messages.send(chatId, lastUser.content);
+    } catch (err) {
+      toast.error(err instanceof Error ? err.message : 'retry send failed');
+    }
+  }, [chatId, streamingId, chatMessages]);
+
   const handleForceSend = useCallback(async (content: string) => {
     const trimmed = content.trim();
     if (!trimmed) return;
@@ -187,6 +259,13 @@ export function ChatPane({ sessionId, chatId, projectId, agentId, onAgentChange,
         </div>
       )}
 
+      {stale && streamingId && (
+        <StaleStreamBanner
+          onRetry={() => void handleRetryStale()}
+          onDiscard={() => void handleDiscardStale()}
+        />
+      )}
+
       <ChatInput
         disabled={false}
         projectId={projectId}