Add task brief research agent: Stage 1 deep-research + canvas draft emission

- run_task_brief_research() runner with brief-specific tool set and max_steps=12
- New agents: client_agent (list_clients, get_client) and relations_agent (query_relations)
- search_associative tool wrapping MemoryMiddleware semantic search
- BRIEF_RESEARCH_TOOLS constant: read-only task/project/note/timeline + memory + client/relations
- canvas block extraction in output_formatter (splits visible text from <canvas> draft)
- device_ws.py: task_brief_research request type; emits canvas_draft mutation on stream_end
- Stage 2 briefMode: briefing_context injected into floating system prompt when present
- briefingContext kwarg wired through compile_prompt call chain

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

app/api/routes/device_ws.py

@@ -43,7 +43,8 @@ from app.api.routes.agent_setup import handle_journey_message, handle_journey_st
 from app.config.settings import settings
 from app.core.agent_runner import trigger_pending_runs
 from app.core.brief_agent import run_home_brief, run_project_brief
-from app.core.deep_agent import run_floating_stream, run_home_stream
+from app.core.deep_agent import run_floating_stream, run_home_stream, run_task_brief_research_stream
+from app.core.output_formatter import extract_canvas_block
 from app.core.device_manager import device_manager
 from app.core.memory_middleware import MemoryMiddleware
 from app.core.output_formatter import StreamFormatter
@@ -164,6 +165,11 @@ async def _message_loop(websocket: WebSocket, user_id: str) -> None:
                 _handle_brief_request(websocket, user_id, frame)
             )
 
+        elif frame_type == WsFrameType.task_brief_request:
+            asyncio.create_task(
+                _handle_task_brief_request(websocket, user_id, frame)
+            )
+
         elif frame_type == WsFrameType.journey_start:
             asyncio.create_task(
                 _handle_journey_start(websocket, user_id, frame)
@@ -415,6 +421,97 @@ async def _handle_brief_request(
     )
 
 
+# ── v6 Task Brief Handler ────────────────────────────────────────────
+
+
+async def _handle_task_brief_request(
+    websocket: WebSocket,
+    user_id: str,
+    frame: dict,
+) -> None:
+    """Handle a task_brief_request frame — Stage-1 executive assistant deep research.
+
+    Streams the briefing markdown back to the client.
+    On stream_end, emits a ``canvas_draft`` mutation if the agent produced one.
+    """
+    request_id = frame.get("request_id") or str(uuid4())
+    session_id = frame.get("session_id") or str(uuid4())
+    task_id: str = frame.get("task_id") or frame.get("taskId") or ""
+
+    logger.info(
+        "device_ws: task_brief_request_start user=%s req=%s task=%s [cache_miss]",
+        user_id, request_id, task_id,
+    )
+
+    if not task_id:
+        await websocket.send_text(
+            WsStreamEnd(request_id=request_id, error="task_id is required").model_dump_json()
+        )
+        return
+
+    async with async_session() as db:
+        memory = MemoryMiddleware(db)
+        memory_context = await memory.enrich_context(
+            user_id,
+            f"task brief: {task_id}",
+            trace_id=request_id,
+            session_id=session_id,
+        )
+
+    context: dict = {
+        "_debug": {"request_id": request_id, "session_id": session_id, "user_id": user_id},
+        "format_prefs": frame.get("format_prefs"),
+        **memory_context,
+    }
+
+    executor = await _make_ws_executor(websocket, user_id)
+    set_client_executor(executor)
+    response_chunks: list[str] = []
+
+    try:
+        event_stream = run_task_brief_research_stream(user_id, task_id, context)
+        formatter = StreamFormatter(request_id=request_id)
+        async for ws_frame in formatter.format(event_stream):
+            if ws_frame.type == "stream_text":  # type: ignore[union-attr]
+                response_chunks.append(ws_frame.chunk)  # type: ignore[union-attr]
+                await websocket.send_text(ws_frame.model_dump_json())
+            elif ws_frame.type == "stream_start":
+                await websocket.send_text(ws_frame.model_dump_json())
+            # stream_end is emitted below with mutations — skip formatter's version
+    except Exception as exc:
+        logger.error(
+            "device_ws: task_brief_request failed user=%s req=%s task=%s: %s",
+            user_id, request_id, task_id, exc,
+        )
+        await websocket.send_text(
+            WsStreamEnd(request_id=request_id, error=str(exc)).model_dump_json()
+        )
+        return
+    finally:
+        clear_client_executor()
+
+    # Extract canvas block then emit stream_end with optional mutations.
+    full_response = "".join(response_chunks)
+    _visible, canvas_content, canvas_kind = extract_canvas_block(full_response)
+
+    mutations: list[dict] = []
+    if canvas_content:
+        mutations.append({
+            "type": "canvas_draft",
+            "content": canvas_content,
+            "kind": canvas_kind,
+        })
+
+    await websocket.send_text(
+        WsStreamEnd(request_id=request_id, mutations=mutations or None).model_dump_json()
+    )
+
+    logger.info(
+        "device_ws: task_brief_request_end user=%s req=%s task=%s response_chars=%d canvas=%s",
+        user_id, request_id, task_id, len(full_response), canvas_kind or "none",
+    )
+
+
 # ── v4 Journey Handlers ─────────────────────────────────────────────