feat(chat): add multi-agent mode routing scaffold and telemetry.

docs/multi-agent-architecture-plan.md

@@ -0,0 +1,211 @@
+# Multi-Agent Architecture Plan (Phased)
+
+This document defines a phased migration from the current `single_agent` flow to a multi-agent architecture while keeping rollback simple and immediate.
+
+## Naming
+
+- `single_agent`: current architecture (default at start)
+- `shadow_multi_agent_v1`: run multi-agent path in background, return `single_agent` output
+- `multi_agent_v1`: multi-agent architecture is the user-facing path
+
+---
+
+## Phase 1 - Parallel Safety Layer
+
+**Goal:** Add safe routing controls with zero behavior change.
+
+### Todo
+
+- [ ] Add mode selector with values: `single_agent`, `shadow_multi_agent_v1`, `multi_agent_v1`
+- [ ] Add global kill switch: force all traffic to `single_agent`
+- [ ] Add mode resolution priority:
+  1. kill switch
+  2. request override
+  3. system default
+- [ ] Keep `single_agent` as default mode
+- [ ] Keep frontend stream/output contract unchanged
+- [ ] Add telemetry tags:
+  - `architecture_mode`
+  - `worker_count`
+  - `retry_count`
+  - `latency_ms`
+  - `token_total`
+- [ ] Write short rollback runbook
+
+### Exit Criteria
+
+- [ ] Can switch modes in staging
+- [ ] Kill switch verified
+- [ ] No frontend contract regressions
+
+---
+
+## Phase 2 - Orchestrator Core and Contracts
+
+**Goal:** Build multi-agent control-plane only (planner/router/merge), with strict schemas.
+
+### Todo
+
+- [ ] Implement orchestrator responsibilities:
+  - intent detection
+  - routing
+  - delegation
+  - fan-in merge
+- [ ] Add budget controls:
+  - max workers per turn
+  - max parallel workers
+  - max turn duration
+- [ ] Add loop/stall guard:
+  - repeated task signature detection
+  - no-progress threshold
+- [ ] Define `WorkerTask` schema:
+  - `domain`, `goal`, `constraints`, `budget`
+- [ ] Define `WorkerResult` schema:
+  - `status`, `summary`, `evidence[]`, `artifacts[]`, `needs_human`, `error_class`
+- [ ] Add schema validation on send/receive boundaries
+- [ ] Add controlled fallback on invalid worker results
+
+### Exit Criteria
+
+- [ ] Orchestrator works end-to-end with mock workers
+- [ ] Invalid worker payloads are blocked cleanly
+
+---
+
+## Phase 3 - Pilot Workers (Gmail and Calendar)
+
+**Goal:** Validate multi-agent architecture with two real domains only.
+
+### Todo
+
+- [ ] Create Gmail worker
+  - [ ] domain-scoped prompt
+  - [ ] domain-only tool loadout
+  - [ ] local query rewrite
+  - [ ] normalized `WorkerResult`
+- [ ] Create Calendar worker
+  - [ ] domain-scoped prompt
+  - [ ] domain-only tool loadout
+  - [ ] local query rewrite/time normalization
+  - [ ] normalized `WorkerResult`
+- [ ] Enforce no cross-domain tool access
+- [ ] Preserve HITL for write actions
+- [ ] Add retry policy by `error_class`
+- [ ] Add tests for routing, loadout isolation, HITL behavior
+
+### Exit Criteria
+
+- [ ] Gmail and Calendar tasks complete in `multi_agent_v1`
+- [ ] No cross-domain tool leakage
+- [ ] HITL still enforced for sensitive writes
+
+---
+
+## Phase 4 - Knowledge Base and Evidence Normalization
+
+**Goal:** Isolate KB retrieval and make evidence citation-ready.
+
+### Todo
+
+- [ ] Move KB retrieval behind dedicated worker/stage
+- [ ] Reuse current KB retrieval logic, but return compact structured evidence only
+- [ ] Define `EvidenceItem` fields:
+  - `claim`, `source_type`, `source_ref`, `confidence`, `snippet`
+- [ ] Add top-k and output-size controls
+- [ ] Add quote-first extraction mode for long contexts
+- [ ] Add tests for traceability and bounded payloads
+
+### Exit Criteria
+
+- [ ] Orchestrator consumes compact evidence (no raw KB dumps)
+- [ ] Citation refs remain valid and traceable
+
+---
+
+## Phase 5 - Verifier and Citation Gate
+
+**Goal:** Prevent unsupported factual claims in final responses.
+
+### Todo
+
+- [ ] Add verifier stage before final synthesis
+- [ ] Enforce claim-to-evidence checks
+- [ ] Add conflict handling policy:
+  - consistent evidence -> accept
+  - conflicting evidence -> label uncertainty or retry
+- [ ] Add unsupported-claim policy:
+  - remove claim or mark uncertain
+- [ ] Add verifier telemetry:
+  - supported claims
+  - unsupported claims
+  - conflicts
+- [ ] Support strict gate and warning modes
+
+### Exit Criteria
+
+- [ ] Unsupported factual claims are blocked or clearly annotated
+- [ ] Citation precision improves on evaluation set
+
+---
+
+## Phase 6 - Shadow Evaluation and Canary
+
+**Goal:** Ship based on data, not intuition.
+
+### Todo
+
+- [ ] Enable `shadow_multi_agent_v1` for selected traffic
+- [ ] Compare metrics vs `single_agent`:
+  - success rate
+  - citation precision
+  - tool-selection accuracy
+  - p95 latency
+  - tokens/request
+  - cost per successful task
+- [ ] Define rollout gates and auto-stop thresholds
+- [ ] Start canary rollout for `multi_agent_v1`
+- [ ] Ramp traffic only if quality and reliability gates pass
+- [ ] Keep kill switch live for entire rollout
+- [ ] Record go/no-go decision with evidence
+
+### Exit Criteria
+
+- [ ] Clear decision based on measured outcomes
+- [ ] Rollback tested successfully during canary
+
+---
+
+## Phase 7 - Domain Expansion and Heavy Tool Reassignment
+
+**Goal:** Scale multi-agent architecture safely across more domains.
+
+### Todo
+
+- [ ] Add domains incrementally (`notion`, `slack`, `jira`, ...)
+- [ ] For each new domain enforce:
+  - scoped tool loadout
+  - local query rewrite
+  - contract validation
+  - eval plus canary gate
+- [ ] Move heavy tools to specialist workers:
+  - podcast generation
+  - artifact/report generation
+  - video presentation
+- [ ] Keep orchestrator toolbelt minimal and control-plane focused
+- [ ] Regularly prune prompts and tool descriptions
+
+### Exit Criteria
+
+- [ ] New domains onboard without reliability regressions
+- [ ] Orchestrator remains lean and stable
+- [ ] Cost per successful task stays controlled
+
+---
+
+## Always-On Checklist
+
+- [ ] Keep `single_agent` path healthy until rollout completion
+- [ ] Keep one-click rollback available at all times
+- [ ] Update observability dashboards every phase
+- [ ] Track failure taxonomy and review weekly
+- [ ] Validate prompt/tool changes via eval before broad rollout

docs/multi-agent-phase1-runbook.md

@@ -0,0 +1,70 @@
+# Multi-Agent Architecture Phase 1 Runbook
+
+## Scope
+
+This runbook covers mode selection and emergency rollback for:
+
+- `single_agent`
+- `shadow_multi_agent_v1`
+- `multi_agent_v1`
+
+Phase 1 keeps execution behavior on the current single-agent path while mode wiring and telemetry are introduced.
+
+## Resolution Priority
+
+Mode resolution follows this fixed order:
+
+1. Global kill switch (`FORCE_SINGLE_AGENT`)
+2. Request override (`architecture_mode` in chat payload)
+3. System default (`AGENT_ARCHITECTURE_MODE`)
+4. Safe fallback (`single_agent`)
+
+## Configuration
+
+Set environment values in backend runtime:
+
+- `AGENT_ARCHITECTURE_MODE=single_agent` (default)
+- `FORCE_SINGLE_AGENT=FALSE` (default)
+
+Changes require backend restart because config is loaded at process startup.
+
+## Mode Switching
+
+### System default switch
+
+1. Set `AGENT_ARCHITECTURE_MODE` to desired value.
+2. Keep `FORCE_SINGLE_AGENT=FALSE`.
+3. Restart backend.
+4. Verify logs include `[architecture_telemetry]` with expected `architecture_mode`.
+
+### Per-request override
+
+Send optional `architecture_mode` in chat request payload:
+
+- `"single_agent"`
+- `"shadow_multi_agent_v1"`
+- `"multi_agent_v1"`
+
+If `FORCE_SINGLE_AGENT=TRUE`, request override is ignored by design.
+
+## Emergency Rollback
+
+Use the kill switch:
+
+1. Set `FORCE_SINGLE_AGENT=TRUE`.
+2. Restart backend.
+3. Verify new requests log `architecture_mode=single_agent`.
+4. Keep this state until incident is resolved.
+
+## Verification Checklist
+
+- Mode resolves according to the priority order.
+- Kill switch overrides all request/default values.
+- Streaming response schema remains unchanged.
+- Architecture telemetry is emitted with:
+  - `architecture_mode`
+  - `orchestrator_used`
+  - `worker_count`
+  - `retry_count`
+  - `latency_ms`
+  - `token_total`

surfsense_backend/app/agents/multi_agent_v1/__init__.py

@@ -0,0 +1,17 @@
+"""Multi-agent v1 architecture package."""
+
+from app.agents.multi_agent_v1.contracts import (
+    GroundingEvidence,
+    SubagentResult,
+    SubagentTaskPlan,
+    WorkerBudget,
+)
+from app.agents.multi_agent_v1.entrypoint import MultiAgentEntrypoint
+
+__all__ = [
+    "GroundingEvidence",
+    "MultiAgentEntrypoint",
+    "SubagentResult",
+    "SubagentTaskPlan",
+    "WorkerBudget",
+]

surfsense_backend/app/agents/multi_agent_v1/contracts.py

@@ -0,0 +1,36 @@
+"""Contracts for multi_agent_v1 orchestrator and subagent communication."""
+
+from __future__ import annotations
+
+from typing import Literal
+
+from pydantic import BaseModel, Field
+
+
+class WorkerBudget(BaseModel):
+    max_steps: int = Field(default=1, ge=1)
+    max_duration_ms: int = Field(default=15_000, ge=100)
+
+
+class SubagentTaskPlan(BaseModel):
+    domain: str = Field(..., min_length=1)
+    goal: str = Field(..., min_length=1)
+    constraints: list[str] = Field(default_factory=list)
+    budget: WorkerBudget = Field(default_factory=WorkerBudget)
+
+
+class GroundingEvidence(BaseModel):
+    claim: str = Field(..., min_length=1)
+    source_type: str = Field(..., min_length=1)
+    source_ref: str = Field(..., min_length=1)
+    confidence: float = Field(default=0.0, ge=0.0, le=1.0)
+    snippet: str = ""
+
+
+class SubagentResult(BaseModel):
+    status: Literal["success", "partial", "error"]
+    summary: str = ""
+    evidence: list[GroundingEvidence] = Field(default_factory=list)
+    artifacts: list[str] = Field(default_factory=list)
+    needs_human: bool = False
+    error_class: str | None = None

surfsense_backend/app/agents/multi_agent_v1/entrypoint.py

@@ -0,0 +1,24 @@
+"""Multi-agent v1 entrypoint scaffold with safe fallback behavior."""
+
+from __future__ import annotations
+
+from collections.abc import AsyncGenerator, Callable
+from typing import Any
+
+
+class MultiAgentEntrypoint:
+    def stream_new_chat(
+        self,
+        *,
+        fallback_streamer: Callable[..., AsyncGenerator[str, None]],
+        fallback_kwargs: dict[str, Any],
+    ) -> AsyncGenerator[str, None]:
+        return fallback_streamer(**fallback_kwargs)
+
+    def stream_resume_chat(
+        self,
+        *,
+        fallback_streamer: Callable[..., AsyncGenerator[str, None]],
+        fallback_kwargs: dict[str, Any],
+    ) -> AsyncGenerator[str, None]:
+        return fallback_streamer(**fallback_kwargs)

surfsense_backend/app/agents/new_chat/architecture_mode.py

@@ -0,0 +1,40 @@
+"""Architecture mode contracts and resolution helpers for chat sessions."""
+
+from __future__ import annotations
+
+from enum import StrEnum
+
+from app.config import config
+
+
+class ArchitectureMode(StrEnum):
+    SINGLE_AGENT = "single_agent"
+    SHADOW_MULTI_AGENT_V1 = "shadow_multi_agent_v1"
+    MULTI_AGENT_V1 = "multi_agent_v1"
+
+
+def parse_architecture_mode(value: str | None) -> ArchitectureMode | None:
+    if not value:
+        return None
+    normalized = value.strip().lower()
+    if not normalized:
+        return None
+    try:
+        return ArchitectureMode(normalized)
+    except ValueError:
+        return None
+
+
+def resolve_architecture_mode(request_override: str | None = None) -> ArchitectureMode:
+    if config.FORCE_SINGLE_AGENT:
+        return ArchitectureMode.SINGLE_AGENT
+
+    override_mode = parse_architecture_mode(request_override)
+    if override_mode is not None:
+        return override_mode
+
+    default_mode = parse_architecture_mode(config.AGENT_ARCHITECTURE_MODE)
+    if default_mode is not None:
+        return default_mode
+
+    return ArchitectureMode.SINGLE_AGENT

surfsense_backend/app/agents/new_chat/telemetry.py

@@ -0,0 +1,43 @@
+"""Architecture telemetry logging for chat execution modes."""
+
+from __future__ import annotations
+
+import json
+from typing import Any
+
+from app.utils.perf import get_perf_logger
+
+_perf_log = get_perf_logger()
+
+
+def log_architecture_telemetry(
+    *,
+    phase: str,
+    architecture_mode: str,
+    orchestrator_used: bool,
+    worker_count: int,
+    retry_count: int,
+    latency_ms: float,
+    token_total: int,
+    request_id: str | None = None,
+    turn_id: str | None = None,
+    status: str = "ok",
+    source: str = "new_chat",
+    extra: dict[str, Any] | None = None,
+) -> None:
+    payload: dict[str, Any] = {
+        "phase": phase,
+        "source": source,
+        "status": status,
+        "architecture_mode": architecture_mode,
+        "orchestrator_used": orchestrator_used,
+        "worker_count": worker_count,
+        "retry_count": retry_count,
+        "latency_ms": round(latency_ms, 2),
+        "token_total": token_total,
+        "request_id": request_id or "unknown",
+        "turn_id": turn_id or "unknown",
+    }
+    if extra:
+        payload.update(extra)
+    _perf_log.info("[architecture_telemetry] %s", json.dumps(payload, ensure_ascii=False))

surfsense_backend/app/config/__init__.py

@@ -342,6 +342,8 @@ class Config:
     ENABLE_DESKTOP_LOCAL_FILESYSTEM = (
         os.getenv("ENABLE_DESKTOP_LOCAL_FILESYSTEM", "FALSE").upper() == "TRUE"
     )
+    AGENT_ARCHITECTURE_MODE = os.getenv("AGENT_ARCHITECTURE_MODE", "single_agent")
+    FORCE_SINGLE_AGENT = os.getenv("FORCE_SINGLE_AGENT", "FALSE").upper() == "TRUE"
 
     @classmethod
     def is_self_hosted(cls) -> bool:

surfsense_backend/app/routes/anonymous_chat_routes.py

@@ -4,6 +4,7 @@ from __future__ import annotations
 
 import logging
 import secrets
+import time
 import uuid
 from pathlib import PurePosixPath
 from typing import Any
@@ -12,6 +13,11 @@ from fastapi import APIRouter, HTTPException, Request, Response, UploadFile, sta
 from fastapi.responses import StreamingResponse
 from pydantic import BaseModel, Field
 
+from app.agents.new_chat.architecture_mode import (
+    ArchitectureMode,
+    resolve_architecture_mode,
+)
+from app.agents.new_chat.telemetry import log_architecture_telemetry
 from app.config import config
 from app.etl_pipeline.file_classifier import (
     DIRECT_CONVERT_EXTENSIONS,
@@ -84,6 +90,7 @@ class AnonChatRequest(BaseModel):
     messages: list[dict[str, Any]] = Field(..., min_length=1)
     disabled_tools: list[str] | None = None
     turnstile_token: str | None = None
+    architecture_mode: ArchitectureMode | None = None
 
 
 class AnonQuotaResponse(BaseModel):
@@ -361,6 +368,22 @@ async def stream_anonymous_chat(
 
         accumulator = start_turn()
         streaming_service = VercelStreamingService()
+        architecture_mode = resolve_architecture_mode(body.architecture_mode)
+        started_at = time.perf_counter()
+        turn_id = f"anon:{session_id}:{request_id}"
+        log_architecture_telemetry(
+            phase="turn_start",
+            source="anon_chat",
+            status="started",
+            architecture_mode=architecture_mode.value,
+            orchestrator_used=False,
+            worker_count=0,
+            retry_count=0,
+            latency_ms=0.0,
+            token_total=0,
+            request_id=request_id,
+            turn_id=turn_id,
+        )
 
         try:
             async with shielded_async_session() as session:
@@ -400,7 +423,10 @@ async def stream_anonymous_chat(
                 }
 
                 langgraph_config = {
-                    "configurable": {"thread_id": anon_thread_id},
+                    "configurable": {
+                        "thread_id": anon_thread_id,
+                        "architecture_mode": architecture_mode.value,
+                    },
                     "recursion_limit": 40,
                 }
 
@@ -468,6 +494,19 @@ async def stream_anonymous_chat(
                         "total_tokens": accumulator.grand_total,
                     },
                 )
+            log_architecture_telemetry(
+                phase="turn_end",
+                source="anon_chat",
+                status="completed",
+                architecture_mode=architecture_mode.value,
+                orchestrator_used=False,
+                worker_count=0,
+                retry_count=0,
+                latency_ms=(time.perf_counter() - started_at) * 1000.0,
+                token_total=accumulator.grand_total,
+                request_id=request_id,
+                turn_id=turn_id,
+            )
 
             yield streaming_service.format_finish_step()
             yield streaming_service.format_finish()
@@ -475,6 +514,20 @@ async def stream_anonymous_chat(
 
         except Exception as e:
             logger.exception("Anonymous chat stream error")
+            log_architecture_telemetry(
+                phase="turn_end",
+                source="anon_chat",
+                status="error",
+                architecture_mode=architecture_mode.value,
+                orchestrator_used=False,
+                worker_count=0,
+                retry_count=0,
+                latency_ms=(time.perf_counter() - started_at) * 1000.0,
+                token_total=accumulator.grand_total,
+                request_id=request_id,
+                turn_id=turn_id,
+                extra={"error_type": type(e).__name__},
+            )
             await TokenQuotaService.anon_release(session_key, ip_key, request_id)
             yield streaming_service.format_error(f"Error during chat: {e!s}")
             yield streaming_service.format_done()

surfsense_backend/app/routes/new_chat_routes.py

@@ -22,6 +22,7 @@ from sqlalchemy.ext.asyncio import AsyncSession
 from sqlalchemy.future import select
 from sqlalchemy.orm import selectinload
 
+from app.agents.new_chat.architecture_mode import resolve_architecture_mode
 from app.agents.new_chat.filesystem_selection import (
     ClientPlatform,
     FilesystemMode,
@@ -61,7 +62,10 @@ from app.schemas.new_chat import (
     TokenUsageSummary,
 )
 from app.services.token_tracking_service import record_token_usage
-from app.tasks.chat.stream_new_chat import stream_new_chat, stream_resume_chat
+from app.tasks.chat.stream_dispatch import (
+    dispatch_new_chat_stream,
+    dispatch_resume_chat_stream,
+)
 from app.users import current_active_user
 from app.utils.rbac import check_permission
 from app.utils.user_message_multimodal import (
@@ -1244,23 +1248,28 @@ async def handle_new_chat(
         image_urls = (
             [p.as_data_url() for p in request.user_images] if request.user_images else None
         )
+        architecture_mode = resolve_architecture_mode(request.architecture_mode)
 
         return StreamingResponse(
-            stream_new_chat(
-                user_query=request.user_query,
-                search_space_id=request.search_space_id,
-                chat_id=request.chat_id,
-                user_id=str(user.id),
-                llm_config_id=llm_config_id,
-                mentioned_document_ids=request.mentioned_document_ids,
-                mentioned_surfsense_doc_ids=request.mentioned_surfsense_doc_ids,
-                needs_history_bootstrap=thread.needs_history_bootstrap,
-                thread_visibility=thread.visibility,
-                current_user_display_name=user.display_name or "A team member",
-                disabled_tools=request.disabled_tools,
-                filesystem_selection=filesystem_selection,
-                request_id=getattr(http_request.state, "request_id", "unknown"),
-                user_image_data_urls=image_urls,
+            dispatch_new_chat_stream(
+                architecture_mode=architecture_mode.value,
+                stream_kwargs={
+                    "user_query": request.user_query,
+                    "search_space_id": request.search_space_id,
+                    "chat_id": request.chat_id,
+                    "user_id": str(user.id),
+                    "llm_config_id": llm_config_id,
+                    "mentioned_document_ids": request.mentioned_document_ids,
+                    "mentioned_surfsense_doc_ids": request.mentioned_surfsense_doc_ids,
+                    "needs_history_bootstrap": thread.needs_history_bootstrap,
+                    "thread_visibility": thread.visibility,
+                    "current_user_display_name": user.display_name or "A team member",
+                    "disabled_tools": request.disabled_tools,
+                    "filesystem_selection": filesystem_selection,
+                    "request_id": getattr(http_request.state, "request_id", "unknown"),
+                    "user_image_data_urls": image_urls,
+                    "architecture_mode": architecture_mode.value,
+                },
             ),
             media_type="text/event-stream",
             headers={
@@ -1458,6 +1467,7 @@ async def regenerate_response(
 
         if request.user_images is not None:
             regenerate_image_urls = [p.as_data_url() for p in request.user_images]
+        architecture_mode = resolve_architecture_mode(request.architecture_mode)
 
         if user_query_to_use is None:
             raise HTTPException(
@@ -1506,23 +1516,28 @@ async def regenerate_response(
         async def stream_with_cleanup():
             streaming_completed = False
             try:
-                async for chunk in stream_new_chat(
-                    user_query=str(user_query_to_use),
-                    search_space_id=request.search_space_id,
-                    chat_id=thread_id,
-                    user_id=str(user.id),
-                    llm_config_id=llm_config_id,
-                    mentioned_document_ids=request.mentioned_document_ids,
-                    mentioned_surfsense_doc_ids=request.mentioned_surfsense_doc_ids,
-                    checkpoint_id=target_checkpoint_id,
-                    needs_history_bootstrap=thread.needs_history_bootstrap,
-                    thread_visibility=thread.visibility,
-                    current_user_display_name=user.display_name or "A team member",
-                    disabled_tools=request.disabled_tools,
-                    filesystem_selection=filesystem_selection,
-                    request_id=getattr(http_request.state, "request_id", "unknown"),
-                    user_image_data_urls=regenerate_image_urls or None,
-                ):
+                stream = dispatch_new_chat_stream(
+                    architecture_mode=architecture_mode.value,
+                    stream_kwargs={
+                        "user_query": str(user_query_to_use),
+                        "search_space_id": request.search_space_id,
+                        "chat_id": thread_id,
+                        "user_id": str(user.id),
+                        "llm_config_id": llm_config_id,
+                        "mentioned_document_ids": request.mentioned_document_ids,
+                        "mentioned_surfsense_doc_ids": request.mentioned_surfsense_doc_ids,
+                        "checkpoint_id": target_checkpoint_id,
+                        "needs_history_bootstrap": thread.needs_history_bootstrap,
+                        "thread_visibility": thread.visibility,
+                        "current_user_display_name": user.display_name or "A team member",
+                        "disabled_tools": request.disabled_tools,
+                        "filesystem_selection": filesystem_selection,
+                        "request_id": getattr(http_request.state, "request_id", "unknown"),
+                        "user_image_data_urls": regenerate_image_urls or None,
+                        "architecture_mode": architecture_mode.value,
+                    },
+                )
+                async for chunk in stream:
                     yield chunk
                 streaming_completed = True
             finally:
@@ -1628,6 +1643,7 @@ async def resume_chat(
         )
 
         decisions = [d.model_dump() for d in request.decisions]
+        architecture_mode = resolve_architecture_mode(request.architecture_mode)
 
         # Release the read-transaction so we don't hold ACCESS SHARE locks
         # on searchspaces/documents for the entire duration of the stream.
@@ -1635,15 +1651,19 @@ async def resume_chat(
         await session.close()
 
         return StreamingResponse(
-            stream_resume_chat(
-                chat_id=thread_id,
-                search_space_id=request.search_space_id,
-                decisions=decisions,
-                user_id=str(user.id),
-                llm_config_id=llm_config_id,
-                thread_visibility=thread.visibility,
-                filesystem_selection=filesystem_selection,
-                request_id=getattr(http_request.state, "request_id", "unknown"),
+            dispatch_resume_chat_stream(
+                architecture_mode=architecture_mode.value,
+                stream_kwargs={
+                    "chat_id": thread_id,
+                    "search_space_id": request.search_space_id,
+                    "decisions": decisions,
+                    "user_id": str(user.id),
+                    "llm_config_id": llm_config_id,
+                    "thread_visibility": thread.visibility,
+                    "filesystem_selection": filesystem_selection,
+                    "request_id": getattr(http_request.state, "request_id", "unknown"),
+                    "architecture_mode": architecture_mode.value,
+                },
             ),
             media_type="text/event-stream",
             headers={

surfsense_backend/app/schemas/new_chat.py

@@ -176,6 +176,11 @@ class LocalFilesystemMountPayload(BaseModel):
 
 MAX_NEW_CHAT_IMAGE_BYTES = 8 * 1024 * 1024
 MAX_NEW_CHAT_IMAGES = 4
+ArchitectureModeLiteral = Literal[
+    "single_agent",
+    "shadow_multi_agent_v1",
+    "multi_agent_v1",
+]
 
 
 class NewChatUserImagePart(BaseModel):
@@ -210,6 +215,7 @@ class NewChatRequest(BaseModel):
     disabled_tools: list[str] | None = (
         None  # Optional list of tool names the user has disabled from the UI
     )
+    architecture_mode: ArchitectureModeLiteral | None = None
     filesystem_mode: Literal["cloud", "desktop_local_folder"] = "cloud"
     client_platform: Literal["web", "desktop"] = "web"
     local_filesystem_mounts: list[LocalFilesystemMountPayload] | None = None
@@ -250,6 +256,7 @@ class RegenerateRequest(BaseModel):
     mentioned_document_ids: list[int] | None = None
     mentioned_surfsense_doc_ids: list[int] | None = None
     disabled_tools: list[str] | None = None
+    architecture_mode: ArchitectureModeLiteral | None = None
     filesystem_mode: Literal["cloud", "desktop_local_folder"] = "cloud"
     client_platform: Literal["web", "desktop"] = "web"
     local_filesystem_mounts: list[LocalFilesystemMountPayload] | None = None
@@ -286,6 +293,7 @@ class ResumeDecision(BaseModel):
 class ResumeRequest(BaseModel):
     search_space_id: int
     decisions: list[ResumeDecision]
+    architecture_mode: ArchitectureModeLiteral | None = None
     filesystem_mode: Literal["cloud", "desktop_local_folder"] = "cloud"
     client_platform: Literal["web", "desktop"] = "web"
     local_filesystem_mounts: list[LocalFilesystemMountPayload] | None = None

surfsense_backend/app/tasks/chat/stream_dispatch.py

@@ -0,0 +1,47 @@
+"""Thin architecture dispatch seam for chat streaming entrypoints."""
+
+from __future__ import annotations
+
+from collections.abc import AsyncGenerator
+from typing import Any
+
+from app.agents.multi_agent_v1.entrypoint import MultiAgentEntrypoint
+from app.agents.new_chat.architecture_mode import (
+    ArchitectureMode,
+    parse_architecture_mode,
+)
+from app.tasks.chat.stream_new_chat import stream_new_chat, stream_resume_chat
+
+
+def _resolve_mode(mode_value: str) -> ArchitectureMode:
+    return parse_architecture_mode(mode_value) or ArchitectureMode.SINGLE_AGENT
+
+
+def dispatch_new_chat_stream(
+    *,
+    architecture_mode: str,
+    stream_kwargs: dict[str, Any],
+) -> AsyncGenerator[str, None]:
+    mode = _resolve_mode(architecture_mode)
+    if mode == ArchitectureMode.SINGLE_AGENT:
+        return stream_new_chat(**stream_kwargs)
+    entrypoint = MultiAgentEntrypoint()
+    return entrypoint.stream_new_chat(
+        fallback_streamer=stream_new_chat,
+        fallback_kwargs=stream_kwargs,
+    )
+
+
+def dispatch_resume_chat_stream(
+    *,
+    architecture_mode: str,
+    stream_kwargs: dict[str, Any],
+) -> AsyncGenerator[str, None]:
+    mode = _resolve_mode(architecture_mode)
+    if mode == ArchitectureMode.SINGLE_AGENT:
+        return stream_resume_chat(**stream_kwargs)
+    entrypoint = MultiAgentEntrypoint()
+    return entrypoint.stream_resume_chat(
+        fallback_streamer=stream_resume_chat,
+        fallback_kwargs=stream_kwargs,
+    )

surfsense_backend/app/tasks/chat/stream_new_chat.py

@@ -42,6 +42,7 @@ from app.agents.new_chat.memory_extraction import (
     extract_and_save_memory,
     extract_and_save_team_memory,
 )
+from app.agents.new_chat.telemetry import log_architecture_telemetry
 from app.db import (
     ChatVisibility,
     NewChatMessage,
@@ -149,6 +150,7 @@ class StreamResult:
     agent_called_update_memory: bool = False
     request_id: str | None = None
     turn_id: str = ""
+    architecture_mode: str = "single_agent"
     filesystem_mode: str = "cloud"
     client_platform: str = "web"
     intent_detected: str = "chat_only"
@@ -182,9 +184,7 @@ def _tool_output_has_error(tool_output: Any) -> bool:
         if tool_output.get("error"):
             return True
         result = tool_output.get("result")
-        if isinstance(result, str) and result.strip().lower().startswith("error:"):
-            return True
-        return False
+        return isinstance(result, str) and result.strip().lower().startswith("error:")
     if isinstance(tool_output, str):
         return tool_output.strip().lower().startswith("error:")
     return False
@@ -231,6 +231,7 @@ def _log_file_contract(stage: str, result: StreamResult, **extra: Any) -> None:
         "request_id": result.request_id or "unknown",
         "turn_id": result.turn_id or "unknown",
         "chat_id": result.turn_id.split(":", 1)[0] if ":" in result.turn_id else "unknown",
+        "architecture_mode": result.architecture_mode,
         "filesystem_mode": result.filesystem_mode,
         "client_platform": result.client_platform,
         "intent_detected": result.intent_detected,
@@ -1308,18 +1309,17 @@ async def _stream_agent_events(
         result.commit_gate_passed, result.commit_gate_reason = (
             _evaluate_file_contract_outcome(result)
         )
-        if not result.commit_gate_passed:
-            if _contract_enforcement_active(result):
-                gate_notice = (
-                    "I could not complete the requested file write because no successful "
-                    "write_file/edit_file operation was confirmed."
-                )
-                gate_text_id = streaming_service.generate_text_id()
-                yield streaming_service.format_text_start(gate_text_id)
-                yield streaming_service.format_text_delta(gate_text_id, gate_notice)
-                yield streaming_service.format_text_end(gate_text_id)
-                yield streaming_service.format_terminal_info(gate_notice, "error")
-                accumulated_text = gate_notice
+        if not result.commit_gate_passed and _contract_enforcement_active(result):
+            gate_notice = (
+                "I could not complete the requested file write because no successful "
+                "write_file/edit_file operation was confirmed."
+            )
+            gate_text_id = streaming_service.generate_text_id()
+            yield streaming_service.format_text_start(gate_text_id)
+            yield streaming_service.format_text_delta(gate_text_id, gate_notice)
+            yield streaming_service.format_text_end(gate_text_id)
+            yield streaming_service.format_terminal_info(gate_notice, "error")
+            accumulated_text = gate_notice
     else:
         result.commit_gate_passed = True
         result.commit_gate_reason = ""
@@ -1351,6 +1351,7 @@ async def stream_new_chat(
     filesystem_selection: FilesystemSelection | None = None,
     request_id: str | None = None,
     user_image_data_urls: list[str] | None = None,
+    architecture_mode: str = "single_agent",
 ) -> AsyncGenerator[str, None]:
     """
     Stream chat responses from the new SurfSense deep agent.
@@ -1384,8 +1385,22 @@ async def stream_new_chat(
     )
     stream_result.request_id = request_id
     stream_result.turn_id = f"{chat_id}:{int(time.time() * 1000)}"
+    stream_result.architecture_mode = architecture_mode
     stream_result.filesystem_mode = fs_mode
     stream_result.client_platform = fs_platform
+    log_architecture_telemetry(
+        phase="turn_start",
+        source="new_chat",
+        status="started",
+        architecture_mode=architecture_mode,
+        orchestrator_used=False,
+        worker_count=0,
+        retry_count=0,
+        latency_ms=0.0,
+        token_total=0,
+        request_id=request_id,
+        turn_id=stream_result.turn_id,
+    )
     _log_file_contract("turn_start", stream_result)
     _perf_log.info(
         "[stream_new_chat] filesystem_mode=%s client_platform=%s",
@@ -1638,6 +1653,7 @@ async def stream_new_chat(
             "search_space_id": search_space_id,
             "request_id": request_id or "unknown",
             "turn_id": stream_result.turn_id,
+            "architecture_mode": architecture_mode,
         }
 
         _perf_log.info(
@@ -1669,6 +1685,7 @@ async def stream_new_chat(
         configurable = {"thread_id": str(chat_id)}
         configurable["request_id"] = request_id or "unknown"
         configurable["turn_id"] = stream_result.turn_id
+        configurable["architecture_mode"] = architecture_mode
         if checkpoint_id:
             configurable["checkpoint_id"] = checkpoint_id
 
@@ -1884,6 +1901,19 @@ async def stream_new_chat(
                         "call_details": accumulator.serialized_calls(),
                     },
                 )
+            log_architecture_telemetry(
+                phase="turn_end",
+                source="new_chat",
+                status="interrupted",
+                architecture_mode=stream_result.architecture_mode,
+                orchestrator_used=False,
+                worker_count=0,
+                retry_count=0,
+                latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+                token_total=accumulator.grand_total,
+                request_id=request_id,
+                turn_id=stream_result.turn_id,
+            )
 
             yield streaming_service.format_finish_step()
             yield streaming_service.format_finish()
@@ -1956,6 +1986,19 @@ async def stream_new_chat(
                     "call_details": accumulator.serialized_calls(),
                 },
             )
+        log_architecture_telemetry(
+            phase="turn_end",
+            source="new_chat",
+            status="completed",
+            architecture_mode=stream_result.architecture_mode,
+            orchestrator_used=False,
+            worker_count=0,
+            retry_count=0,
+            latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+            token_total=accumulator.grand_total,
+            request_id=request_id,
+            turn_id=stream_result.turn_id,
+        )
 
         # Fire background memory extraction if the agent didn't handle it.
         # Shared threads write to team memory; private threads write to user memory.
@@ -2000,6 +2043,20 @@ async def stream_new_chat(
         print(f"[stream_new_chat] {error_message}")
         print(f"[stream_new_chat] Exception type: {type(e).__name__}")
         print(f"[stream_new_chat] Traceback:\n{traceback.format_exc()}")
+        log_architecture_telemetry(
+            phase="turn_end",
+            source="new_chat",
+            status="error",
+            architecture_mode=stream_result.architecture_mode,
+            orchestrator_used=False,
+            worker_count=0,
+            retry_count=0,
+            latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+            token_total=accumulator.grand_total,
+            request_id=request_id,
+            turn_id=stream_result.turn_id,
+            extra={"error_type": type(e).__name__},
+        )
 
         yield streaming_service.format_error(error_message)
         yield streaming_service.format_finish_step()
@@ -2093,6 +2150,7 @@ async def stream_resume_chat(
     thread_visibility: ChatVisibility | None = None,
     filesystem_selection: FilesystemSelection | None = None,
     request_id: str | None = None,
+    architecture_mode: str = "single_agent",
 ) -> AsyncGenerator[str, None]:
     streaming_service = VercelStreamingService()
     stream_result = StreamResult()
@@ -2103,8 +2161,22 @@ async def stream_resume_chat(
     )
     stream_result.request_id = request_id
     stream_result.turn_id = f"{chat_id}:{int(time.time() * 1000)}"
+    stream_result.architecture_mode = architecture_mode
     stream_result.filesystem_mode = fs_mode
     stream_result.client_platform = fs_platform
+    log_architecture_telemetry(
+        phase="turn_start",
+        source="resume_chat",
+        status="started",
+        architecture_mode=architecture_mode,
+        orchestrator_used=False,
+        worker_count=0,
+        retry_count=0,
+        latency_ms=0.0,
+        token_total=0,
+        request_id=request_id,
+        turn_id=stream_result.turn_id,
+    )
     _log_file_contract("turn_start", stream_result)
     _perf_log.info(
         "[stream_resume] filesystem_mode=%s client_platform=%s",
@@ -2250,6 +2322,7 @@ async def stream_resume_chat(
                 "thread_id": str(chat_id),
                 "request_id": request_id or "unknown",
                 "turn_id": stream_result.turn_id,
+                "architecture_mode": architecture_mode,
             },
             "recursion_limit": 80,
         }
@@ -2300,6 +2373,19 @@ async def stream_resume_chat(
                         "call_details": accumulator.serialized_calls(),
                     },
                 )
+            log_architecture_telemetry(
+                phase="turn_end",
+                source="resume_chat",
+                status="interrupted",
+                architecture_mode=stream_result.architecture_mode,
+                orchestrator_used=False,
+                worker_count=0,
+                retry_count=0,
+                latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+                token_total=accumulator.grand_total,
+                request_id=request_id,
+                turn_id=stream_result.turn_id,
+            )
 
             yield streaming_service.format_finish_step()
             yield streaming_service.format_finish()
@@ -2353,6 +2439,19 @@ async def stream_resume_chat(
                     "call_details": accumulator.serialized_calls(),
                 },
             )
+        log_architecture_telemetry(
+            phase="turn_end",
+            source="resume_chat",
+            status="completed",
+            architecture_mode=stream_result.architecture_mode,
+            orchestrator_used=False,
+            worker_count=0,
+            retry_count=0,
+            latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+            token_total=accumulator.grand_total,
+            request_id=request_id,
+            turn_id=stream_result.turn_id,
+        )
 
         yield streaming_service.format_finish_step()
         yield streaming_service.format_finish()
@@ -2364,6 +2463,20 @@ async def stream_resume_chat(
         error_message = f"Error during resume: {e!s}"
         print(f"[stream_resume_chat] {error_message}")
         print(f"[stream_resume_chat] Traceback:\n{traceback.format_exc()}")
+        log_architecture_telemetry(
+            phase="turn_end",
+            source="resume_chat",
+            status="error",
+            architecture_mode=stream_result.architecture_mode,
+            orchestrator_used=False,
+            worker_count=0,
+            retry_count=0,
+            latency_ms=(time.perf_counter() - _t_total) * 1000.0,
+            token_total=accumulator.grand_total,
+            request_id=request_id,
+            turn_id=stream_result.turn_id,
+            extra={"error_type": type(e).__name__},
+        )
         yield streaming_service.format_error(error_message)
         yield streaming_service.format_finish_step()
         yield streaming_service.format_finish()