[chore] logical-gaps: release approve async dispatch (202) — kill the 40min synchronous HTTP 504

The approve route ran the whole fail-closed execute inline: clone(600s) +
gate(1800s) + CI poll(2400s) + publish(300s) ≈ up to 85min worst case. nginx
(the single :3000 entry point, ~60s read timeout) 504'd long before it
finished, so the CEO's approve always appeared to fail even when the release
succeeded server-side — the structured ReleaseResult was unreachable over the
wire. dispatch_approve spawns the execute in a background task with a fresh
session (built from the request session's engine) and the route returns 202
'accepted' immediately; _INFLIGHT_APPROVES tracks the dispatched task for
observability (self-cleans via done-callback; the Redis mutex still refuses a
double-execute on a second click). The panel already polls GET /proposal every
30s, so it observes the final status (COMPLETED on published/already_published,
else the proposal stays open for retry); the card's approve toast now treats
'accepted' as an info 'dispatched, running in the background' instead of the
old 'Release halted' warning.

TDD: 2 route tests red→green (approve returns 202 'accepted' + the proposal
transitions to COMPLETED / stays PENDING once the background faked execute
completes; the dispatched task is awaited while the executor patch is live).
83 release tests green; ruff/mypy clean; panel typecheck+lint+format+test
green.

panel/src/components/dashboard/release-proposal-card.tsx

@@ -65,6 +65,14 @@ export function ReleaseProposalCard({ className }: { className?: string }) {
           `Published v${result.version}` +
             (result.release_url ? "" : " (no release URL returned)"),
         );
+      } else if (result.status === "accepted") {
+        // The execute runs in the background (a synchronous request would 504
+        // at nginx before the ~40min fail-closed gate/CI/publish finished).
+        // This card polls GET /proposal every 30s and reflects the final
+        // outcome (COMPLETED on a publish, else the proposal stays open).
+        toast.info(
+          "Release execute dispatched — running in the background. This card updates as it progresses.",
+        );
       } else {
         toast.warning(`Release halted (${result.status}): ${result.detail}`);
       }

roboco/api/routes/release.py

@@ -8,6 +8,7 @@ keeps the proposal held. Nothing here publishes without the CEO's explicit POST.
 from typing import TYPE_CHECKING, cast
 
 from fastapi import APIRouter, HTTPException, status
+from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 
 from roboco.api.deps import CurrentAgentContext, DbSession, require_ceo_role
 from roboco.api.schemas.release import (
@@ -18,7 +19,10 @@ from roboco.api.schemas.release import (
     ReleaseReportModel,
 )
 from roboco.foundation.policy.content import markers
-from roboco.services.release_proposal import get_release_proposal_service
+from roboco.services.release_proposal import (
+    dispatch_approve,
+    get_release_proposal_service,
+)
 
 if TYPE_CHECKING:
     from uuid import UUID
@@ -71,11 +75,25 @@ async def get_release_proposal(
     return _to_response(task)
 
 
-@router.post("/proposal/approve", response_model=ReleaseExecuteResponse)
+@router.post(
+    "/proposal/approve",
+    response_model=ReleaseExecuteResponse,
+    status_code=status.HTTP_202_ACCEPTED,
+)
 async def approve_release_proposal(
     db: DbSession, agent: CurrentAgentContext
 ) -> ReleaseExecuteResponse:
-    """Approve the held proposal → run the fail-closed executor."""
+    """Approve the held proposal → dispatch the fail-closed executor async.
+
+    The execute is a ~40min clone→gate→CI→publish pipeline; running it inline
+    would 504 at nginx (the single :3000 entry point, ~60s read timeout) before
+    it finished, so the CEO's approve always appeared to fail even when the
+    release succeeded server-side. The route dispatches the execute in a
+    background task with a fresh session and returns 202 immediately; the panel
+    polls ``GET /proposal`` to observe the final status (COMPLETED on
+    published/already_published, else the proposal stays open for retry). A
+    second click is refused by the Redis mutex (``already_in_progress``).
+    """
     _require_ceo(agent)
     svc = get_release_proposal_service(db)
     task = await svc.open_proposal()
@@ -83,20 +101,24 @@ async def approve_release_proposal(
         raise HTTPException(
             status_code=status.HTTP_404_NOT_FOUND, detail="No open release proposal"
         )
-    result = await svc.approve(cast("UUID", task.id))
-    if result is None:
-        raise HTTPException(
-            status_code=status.HTTP_400_BAD_REQUEST,
-            detail="Proposal has no stored readiness report",
-        )
+    # Materialize the proposal for the background session (a no-op in prod,
+    # where the release-manager engine already committed it; tests seed it only
+    # flushed into the request session).
     await db.commit()
+    factory = async_sessionmaker(
+        bind=db.bind, class_=AsyncSession, expire_on_commit=False, autoflush=False
+    )
+    dispatch_approve(cast("UUID", task.id), factory)
     return ReleaseExecuteResponse(
-        status=result.status,
-        version=result.version,
-        files_changed=result.files_changed,
-        commit_sha=result.commit_sha,
-        release_url=result.release_url,
-        detail=result.detail,
+        status="accepted",
+        version="",
+        files_changed=[],
+        commit_sha=None,
+        release_url=None,
+        detail=(
+            "Release execute dispatched in the background; poll"
+            " /api/release/proposal for the final status."
+        ),
     )
 
 

roboco/services/release_proposal.py

@@ -29,7 +29,7 @@ from roboco.services.task import RELEASE_MANAGER_SOURCE, get_task_service
 if TYPE_CHECKING:
     from uuid import UUID
 
-    from sqlalchemy.ext.asyncio import AsyncSession
+    from sqlalchemy.ext.asyncio import AsyncSession, async_sessionmaker
 
     from roboco.db.tables import TaskTable
 
@@ -290,3 +290,47 @@ class ReleaseProposalService(BaseService):
 def get_release_proposal_service(session: AsyncSession) -> ReleaseProposalService:
     """Construct a ReleaseProposalService bound to ``session``."""
     return ReleaseProposalService(session)
+
+
+# In-flight background approves keyed by proposal task id. The HTTP approve
+# route dispatches the ~40min execute asynchronously (a synchronous request
+# would 504 at any proxy before the fail-closed gate/CI/publish finished) and
+# returns 202 immediately; the panel polls GET /proposal for the final status.
+# This registry lets a status endpoint / tests await the dispatched execute;
+# it self-cleans via a done-callback and the Redis mutex still prevents a
+# double-execute on a second click.
+_INFLIGHT_APPROVES: dict[UUID, asyncio.Task[None]] = {}
+
+
+async def _run_approve_background(
+    task_id: UUID, session_factory: async_sessionmaker[AsyncSession]
+) -> None:
+    """Run ``approve`` in a background task with a fresh session (the request
+    session closes when the 202 returns). Commits the outcome; a failure logs
+    and rolls back — the proposal stays open for the CEO to retry."""
+    async with session_factory() as bg_db:
+        try:
+            result = await get_release_proposal_service(bg_db).approve(task_id)
+            logger.info(
+                "release approve completed task_id=%s status=%s",
+                task_id,
+                result.status if result is not None else "no_report",
+            )
+            await bg_db.commit()
+        except Exception:
+            logger.exception(
+                "release approve background task failed task_id=%s", task_id
+            )
+            await bg_db.rollback()
+
+
+def dispatch_approve(
+    task_id: UUID, session_factory: async_sessionmaker[AsyncSession]
+) -> asyncio.Task[None]:
+    """Spawn the long release execute in a background task so the HTTP approve
+    route can return 202 immediately. Registered in ``_INFLIGHT_APPROVES`` for
+    observability (done-callback removes the entry)."""
+    bg_task = asyncio.create_task(_run_approve_background(task_id, session_factory))
+    _INFLIGHT_APPROVES[task_id] = bg_task
+    bg_task.add_done_callback(lambda _t: _INFLIGHT_APPROVES.pop(task_id, None))
+    return bg_task

tests/integration/test_release_routes.py

@@ -12,18 +12,22 @@ import pytest_asyncio
 from fastapi import FastAPI
 from httpx import ASGITransport, AsyncClient
 from roboco.api.deps import get_agent_context, get_db
+from roboco.api.routes import release as release_route
 from roboco.api.routes.release import router as release_router
 from roboco.db.tables import AgentTable, ProjectTable, TaskTable
 from roboco.models import AgentRole, AgentStatus, Team
 from roboco.models.base import TaskNature, TaskStatus, TaskType
 from roboco.models.permissions import AgentContext
 from roboco.services.release_executor import ReleaseResult
+from roboco.services.release_proposal import ReleaseProposalService
 from roboco.services.release_readiness import ReleaseReadinessReport, report_to_dict
 from roboco.services.task import RELEASE_MANAGER_SOURCE
 from sqlalchemy import delete
 
 if TYPE_CHECKING:
+    import asyncio
     from collections.abc import AsyncIterator
+    from typing import Any
 
     from sqlalchemy.ext.asyncio import AsyncSession
 
@@ -148,9 +152,15 @@ async def test_get_proposal_404_when_none(ceo_client: AsyncClient) -> None:
 
 
 @pytest.mark.asyncio
-async def test_approve_runs_executor_and_completes(
+async def test_approve_dispatches_async_and_completes(
     db_session: AsyncSession, ceo_client: AsyncClient
 ) -> None:
+    """#324: the approve route dispatches the ~40min execute in a background
+    task and returns 202 immediately (a synchronous request would 504 at nginx
+    before the fail-closed gate/CI/publish finished). The panel polls
+    GET /proposal for the final status; here we await the dispatched task and
+    assert the proposal transitions to COMPLETED once the (faked) publish
+    succeeds."""
     task = await _seed_proposal(db_session)
     published = ReleaseResult(
         status="published",
@@ -162,23 +172,55 @@ async def test_approve_runs_executor_and_completes(
     )
     fake_executor = AsyncMock()
     fake_executor.execute = AsyncMock(return_value=published)
-    with patch(
-        "roboco.services.release_proposal.get_release_executor",
-        AsyncMock(return_value=fake_executor),
+    captured: dict[str, asyncio.Task[None]] = {}
+    real_dispatch = release_route.dispatch_approve
+
+    def _capturing_dispatch(task_id: UUID, factory: Any) -> asyncio.Task[None]:
+        bg = real_dispatch(task_id, factory)
+        captured["task"] = bg
+        return bg
+
+    with (
+        patch(
+            "roboco.services.release_proposal.get_release_executor",
+            AsyncMock(return_value=fake_executor),
+        ),
+        patch(
+            "roboco.api.routes.release.dispatch_approve",
+            side_effect=_capturing_dispatch,
+        ),
+        patch.object(
+            ReleaseProposalService, "_acquire_release_lock", AsyncMock(return_value="t")
+        ),
+        patch.object(
+            ReleaseProposalService,
+            "_release_release_lock",
+            AsyncMock(return_value=None),
+        ),
+        patch.object(
+            ReleaseProposalService,
+            "_heartbeat_release_lock",
+            AsyncMock(return_value=True),
+        ),
     ):
         resp = await ceo_client.post("/api/release/proposal/approve")
-    assert resp.status_code == HTTPStatus.OK
-    assert resp.json()["status"] == "published"
+        assert resp.status_code == HTTPStatus.ACCEPTED
+        assert resp.json()["status"] == "accepted"
+        # Await the background execute WHILE the executor patch is still active
+        # (the dispatched task runs the faked publish).
+        bg = captured["task"]
+        await bg
     fake_executor.execute.assert_awaited_once()
-    refreshed = await db_session.get(TaskTable, task.id)
-    assert refreshed is not None
-    assert refreshed.status == TaskStatus.COMPLETED
+    await db_session.refresh(task)
+    assert task.status == TaskStatus.COMPLETED
 
 
 @pytest.mark.asyncio
-async def test_approve_gate_failure_keeps_proposal_open(
+async def test_approve_gate_failure_keeps_proposal_open_async(
     db_session: AsyncSession, ceo_client: AsyncClient
 ) -> None:
+    """A gate failure in the background execute leaves the proposal open (the
+    CEO retries after the cause is fixed); the route still returned 202."""
     task = await _seed_proposal(db_session)
     failed = ReleaseResult(
         status="gate_failed",
@@ -190,16 +232,43 @@ async def test_approve_gate_failure_keeps_proposal_open(
     )
     fake_executor = AsyncMock()
     fake_executor.execute = AsyncMock(return_value=failed)
-    with patch(
-        "roboco.services.release_proposal.get_release_executor",
-        AsyncMock(return_value=fake_executor),
+    captured: dict[str, asyncio.Task[None]] = {}
+    real_dispatch = release_route.dispatch_approve
+
+    def _capturing_dispatch(task_id: UUID, factory: Any) -> asyncio.Task[None]:
+        bg = real_dispatch(task_id, factory)
+        captured["task"] = bg
+        return bg
+
+    with (
+        patch(
+            "roboco.services.release_proposal.get_release_executor",
+            AsyncMock(return_value=fake_executor),
+        ),
+        patch(
+            "roboco.api.routes.release.dispatch_approve",
+            side_effect=_capturing_dispatch,
+        ),
+        patch.object(
+            ReleaseProposalService, "_acquire_release_lock", AsyncMock(return_value="t")
+        ),
+        patch.object(
+            ReleaseProposalService,
+            "_release_release_lock",
+            AsyncMock(return_value=None),
+        ),
+        patch.object(
+            ReleaseProposalService,
+            "_heartbeat_release_lock",
+            AsyncMock(return_value=True),
+        ),
     ):
         resp = await ceo_client.post("/api/release/proposal/approve")
-    assert resp.status_code == HTTPStatus.OK
-    assert resp.json()["status"] == "gate_failed"
-    refreshed = await db_session.get(TaskTable, task.id)
-    assert refreshed is not None
-    assert refreshed.status == TaskStatus.PENDING  # still held
+        assert resp.status_code == HTTPStatus.ACCEPTED
+        assert resp.json()["status"] == "accepted"
+        await captured["task"]
+    await db_session.refresh(task)
+    assert task.status == TaskStatus.PENDING  # still held for retry
 
 
 @pytest.mark.asyncio