feat(maf): conversation-driven MAF roundtable (R1/R2) + server-owned stage-3 dispatch

Backend:
- Conversation API at /api/maf-roundtable/conversations: create/list/detail,
  paginated messages, atomic send+run start, owner-only isolation,
  admin read-only diagnostic access
- SSE real-time event stream (GET /conversations/{id}/stream) backed by
  ConversationEventHub + conversation_events store; multi-worker deployment doc
- Deterministic server-owned stage-3 dispatch contracts (dispatch_contracts.py):
  routing resolved from trusted MAF deployment config instead of manager-chosen
  agent ids; config.yaml keeps stage-3 routing deployment-owned
  (default_manager_agent_id / seat ids / business_routes / max_seats_per_run)
- Conversation persistence (conversations/messages stores) + v2 migration script
- Tests: conversations, conversation stream, migration, phase3 conversation dispatch

Frontend:
- Vitest test infra for MAF (vitest.maf.config.ts + test:maf script) with
  component tests for ApprovalCard/RunHeader/SessionsPage and use-maf-events
- Session form lib, flow-graph dialog / run-detail / sessions page updates,
  routes and nav tweaks
- Docs: R0 legacy interaction baseline mapping, updated implementation plan

Co-authored-by: Cursor <cursoragent@cursor.com>
This commit is contained in:
董宏凯andCursor committed 2026-08-03 12:09:47 +08:00
1 parent df408831e5
commit 99642fca6c
46 files changed
+6358 -1306

No files matched your search

@@ -903,9 +903,10 @@ async def test_seat_matching_scopes_by_turn_index(svc, store):
@pytest.mark.asyncio
async def test_seat_matching_count_mismatch_fails(svc, store):
"""P1b: a pending tool call that does not match EXACTLY one waiting action
(by tool_call_id) fails the turn —the executor never guesses."""
async def test_seat_matching_zero_hits_warns_and_skips(svc, store):
"""A pending tool call that matches 0 waiting actions (e.g. unclassified
tool, hallucinated tool name, or WriteGuard let it through as read) should
log a warning and skip — NOT crash the entire workflow."""
from app.maf_roundtable.workflow_executors import SeatExecutor
session = await svc.create_session(user_id="u1", body={"task_text": "t"})
@@ -923,7 +924,39 @@ async def test_seat_matching_count_mismatch_fails(svc, store):
executor = SeatExecutor(_FakeSeatAdapter(store, response), state, "seat-1",
action_coordinator=ActionCoordinator(svc, user_id="u1"))
with pytest.raises(RuntimeError, match="expected exactly 1"):
# Should NOT raise — the unmatched tool call is logged and skipped.
await executor.handle("msg", _FakeSeatCtx())
# The unmatched call was not added to pending_approvals.
assert len(state.pending_approvals) == 0
@pytest.mark.asyncio
async def test_seat_matching_multiple_hits_raises(svc, store):
""">1 match for a single tool call is a genuine data-integrity error —
the executor refuses to guess and raises RuntimeError."""
from app.maf_roundtable.workflow_executors import SeatExecutor
session = await svc.create_session(user_id="u1", body={"task_text": "t"})
run = await svc.create_run(session_id=session["id"], user_id="u1")
# Two waiting actions with the same logical_action_id for turn 1.
await store.actions.create_action({
"session_id": session["id"], "run_id": run["id"], "agent_id": "seat-1",
"tool_name": "update_task", "turn_index": 1, "logical_action_id": "call-1",
"idempotency_key": "dup1", "risk_level": "R3", "status": "waiting_approval",
})
await store.actions.create_action({
"session_id": session["id"], "run_id": run["id"], "agent_id": "seat-1",
"tool_name": "update_task", "turn_index": 1, "logical_action_id": "call-1",
"idempotency_key": "dup2", "risk_level": "R3", "status": "waiting_approval",
})
state = _seat_state(session["id"], run["id"])
response = _FakeSeatResponse(tool_calls=[{"name": "update_task", "id": "call-1", "args": {}}])
executor = SeatExecutor(_FakeSeatAdapter(store, response), state, "seat-1",
action_coordinator=ActionCoordinator(svc, user_id="u1"))
with pytest.raises(RuntimeError, match="refusing to guess"):
await executor.handle("msg", _FakeSeatCtx())