Files
PaperClipAI/server/src/services/execution-projection.ts
T
DottaandPaperclip 5842185e4f fix: surface native cleanup quarantine and add chat qualification evals (#13775)
## Thinking Path

> - Paperclip lets people manage AI agents and their work.
> - Agent Chat needs reliable native execution before native runners
become the onboarding default.
> - Existing stories covered idle reassignment and controller restart,
but not an executing worker handoff or worker process loss.
> - Status answer tests also need to reject stale claims and invented
facts.
> - This pull request adds six opt-in full-stack cells with independent
state assertions and retained evidence.
> - The probes exposed a misleading Retry across server projection and
recovery-banner paths; the fix reports the blocked recovery honestly.
> - The tests preserve failures without changing recovery policy,
production prompts, or onboarding defaults.

## Linked Issues or Issue Description

Refs: #13762. Related: #13765 (Retry targets the latest failed attempt),
#13753 (task context ownership), #13746 (native recovery work).

## What Changed

- Add active reassignment with saved draft and plan preservation,
old-worker cancellation, and successor completion checks.
- Preserve recovery-needed projection when native cleanup fails before
its coordinator exists, refuse a generic retry that would immediately
fail again, and replace the recovery banner's misleading Retry with
Inspect run.
- Add verified local worker process loss with a required successful
continuation; retain a failing qualification result when recovery is
unavailable, while independently verifying the UI/API refuse doomed
retries.
- Add two-turn factual answer checks for current blockers, stale claims,
inactive backlog work, and unknown facts. Retain prose for separate
semantic review.
- Add positive and negative oracle calibration and document fault
isolation, cleanup, billing, and qualification limits.

## Verification

- Eval TypeScript check passes.
- All 442 eval support tests pass locally. The 89 focused server tests
and server typecheck pass. Six recovery-banner UI tests and token gates
pass.
- Initial new-cell campaign:
https://github.com/paperclipai/paperclip/actions/runs/35657128077. All
six results are retained; four failed on fixture-contract issues and two
exposed real worker cleanup quarantine.
- All 26 existing native onboarding cells:
https://github.com/paperclipai/paperclip/actions/runs/35656761484 (26/26
passed on master 846336e5a, all cleanup passed).
- Intermediate handoff/fault campaign:
https://github.com/paperclipai/paperclip/actions/runs/35657945095 (four
retained failures: two overly strict draft oracles, two real crash
quarantines).
- Final active handoff:
https://github.com/paperclipai/paperclip/actions/runs/35659014397 (2/2
passed on cf6d4ae3a; both cleanup passed).
- Clarified answer-quality fixtures:
https://github.com/paperclipai/paperclip/actions/runs/35658262695 (2/2
passed on 4a26f10be; both cleanup passed; all four answers semantically
reviewed).
- Quarantine guard regression campaign:
https://github.com/paperclipai/paperclip/actions/runs/35659580100 (both
API requests correctly refused with 409/no second run, but exposed a
separate misleading Retry in the recovery banner and a fixture wait on a
non-admitted run).
- Final quarantine guard verification:
https://github.com/paperclipai/paperclip/actions/runs/35661067305
(147e42f7e: both providers verify Inspect run/no Retry, HTTP 409, one
retained run, unchanged saved plan, and successful disposable cleanup.
Both evals intentionally remain red with
`worker_crash_recovery_unqualified`; no successful continuation exists).
The preceding campaign 35658772755 never ran provider cases because
GitHub artifact finalization returned HTTP 403.
- Full repository CI passes on 147e42f7e: typecheck, tests, build, and
browser gates. One unchanged local-service-supervisor readiness test
failed initially; its six-test file passed in isolation and the failed
shard passed on its single rerun. Latest-head rollup: 54 successful, 2
intentionally skipped, no failed or pending checks. Greptile is 5/5 with
zero unresolved findings.
- See tests/runner-e2e/QUALIFICATION-2026-09-21.md for retained attempts
and semantic review. Published reports:
[onboarding](https://d1p6rlowie26tp.cloudfront.net/runner-e2e/campaigns/gha-35656761484-1/),
[handoff](https://d1p6rlowie26tp.cloudfront.net/runner-e2e/campaigns/gha-35659014397-1/),
[grounded
answers](https://d1p6rlowie26tp.cloudfront.net/runner-e2e/campaigns/gha-35658262695-1/),
[crash guards and unqualified
recovery](https://d1p6rlowie26tp.cloudfront.net/runner-e2e/campaigns/gha-35661067305-1/).

## Risks

- Paid cells are explicit-only and local-only. The fault fixture signals
only the exact native run PID after checking its identity.
- Live worker-loss probes currently fail on cleanup quarantine for both
providers. The eval must remain red until there is a usable recovery,
even when preservation and refusal checks pass. Verified cleanup with a
fresh attempt versus exact-session resume remains a product decision.
- Structured facts alone do not qualify prose quality; semantic review
remains separate.
- Onboarding uses the existing runtime switch after the real wizard and
before provider execution. Native UI selection and public defaults
remain unchanged.

## Model Used

OpenAI GPT-6 through Codex, with reasoning, tool use, and code
execution. The exact served snapshot and context-window size are not
exposed in this session.

## Checklist


- [x] I have included a thinking path that traces from project context
to this change
- [x] I have specified the model used (with version and capability
details)
- [x] I have checked ROADMAP.md and confirmed this PR does not duplicate
planned core work
- [x] I have searched GitHub for duplicate or related PRs and linked
them above
- [x] I have either (a) linked existing issues with `Fixes: #` / `Closes
#` / `Refs #` OR (b) described the issue in-PR following the relevant
issue template
- [x] I have not referenced internal/instance-local Paperclip issues or
links (only public GitHub `#NNN` / `github.com/paperclipai/paperclip`
URLs)
- [x] My branch name describes the change (e.g. `docs/...`, `fix/...`)
and contains no internal Paperclip ticket id or instance-derived details
- [x] I have run tests locally and they pass
- [x] I have added or updated tests where applicable
- [x] I have updated relevant documentation to reflect my changes
- [x] I have considered and documented any risks above
- [x] All Paperclip CI gates are green
- [x] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

---------

Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-09-21 18:13:01 -05:00

320 lines
12 KiB
TypeScript

import { and, desc, eq, inArray, sql } from "drizzle-orm";
import {
heartbeatRuns,
issueRecoveryActions,
issueThreadInteractions,
nativeRunFinalizations,
type Db,
} from "@paperclipai/db";
import type { ExecutionProjection } from "@paperclipai/shared";
import { EXECUTION_CONTROL_DEADLINE_MS } from "./execution-control-deadline.js";
import { executionFailureRetryCount } from "./execution-recovery-attempt.js";
const text = (v: unknown) => (typeof v === "string" ? v : null);
const executionRunColumns = {
id: heartbeatRuns.id,
errorCode: heartbeatRuns.errorCode,
executionControlDeadlineAt: heartbeatRuns.executionControlDeadlineAt,
finishedAt: heartbeatRuns.finishedAt,
lastOutputAt: heartbeatRuns.lastOutputAt,
lastUsefulActionAt: heartbeatRuns.lastUsefulActionAt,
nativeIssueId: heartbeatRuns.nativeIssueId,
nextAction: heartbeatRuns.nextAction,
processPid: heartbeatRuns.processPid,
retryOfRunId: heartbeatRuns.retryOfRunId,
runtimeMode: heartbeatRuns.runtimeMode,
scheduledRetryAt: heartbeatRuns.scheduledRetryAt,
scheduledRetryAttempt: heartbeatRuns.scheduledRetryAttempt,
scheduledRetryReason: heartbeatRuns.scheduledRetryReason,
startedAt: heartbeatRuns.startedAt,
status: heartbeatRuns.status,
contextSnapshot: sql<Record<string, unknown>>`jsonb_build_object(
'issueId', ${heartbeatRuns.contextSnapshot}->'issueId',
'failureRetriesBeforeAiConnectionWait', ${heartbeatRuns.contextSnapshot}->'failureRetriesBeforeAiConnectionWait',
'failureRetriesBeforeWorkspaceWait', ${heartbeatRuns.contextSnapshot}->'failureRetriesBeforeWorkspaceWait')`,
};
type Run = Pick<typeof heartbeatRuns.$inferSelect, keyof typeof executionRunColumns>;
type Coordinator = typeof nativeRunFinalizations.$inferSelect;
type Recovery = Pick<
typeof issueRecoveryActions.$inferSelect,
"cause" | "nextAction"
> &
Partial<
Pick<typeof issueRecoveryActions.$inferSelect, "status" | "evidence">
>;
/** Batched reads; list consumers do not perform per-task polling. */
export async function executionProjectionsForRuns(
db: Db,
companyId: string,
runIds: string[],
now = new Date(),
) {
const projections = new Map<string, ExecutionProjection>();
if (!runIds.length) return projections;
const runs = await db
.select(executionRunColumns)
.from(heartbeatRuns)
.where(
and(
eq(heartbeatRuns.companyId, companyId),
inArray(heartbeatRuns.id, runIds),
),
);
const coordinators = await db
.select()
.from(nativeRunFinalizations)
.where(
and(
eq(nativeRunFinalizations.companyId, companyId),
inArray(nativeRunFinalizations.runId, runIds),
),
);
const issueIds = [
...new Set(
runs
.map((run) => run.nativeIssueId ?? text(run.contextSnapshot?.issueId))
.filter((id): id is string => !!id),
),
];
const pending = issueIds.length
? await db
.select({
issueId: issueThreadInteractions.issueId,
kind: issueThreadInteractions.kind,
})
.from(issueThreadInteractions)
.where(
and(
eq(issueThreadInteractions.companyId, companyId),
inArray(issueThreadInteractions.issueId, issueIds),
eq(issueThreadInteractions.status, "pending"),
),
)
: [];
const recovery = issueIds.length
? await db
.select({
issueId: issueRecoveryActions.sourceIssueId,
cause: issueRecoveryActions.cause,
nextAction: issueRecoveryActions.nextAction,
evidence: issueRecoveryActions.evidence,
status: issueRecoveryActions.status,
})
.from(issueRecoveryActions)
.where(
and(
eq(issueRecoveryActions.companyId, companyId),
inArray(issueRecoveryActions.sourceIssueId, issueIds),
inArray(issueRecoveryActions.status, [
"active",
"escalated",
"resolved",
]),
),
)
.orderBy(desc(issueRecoveryActions.updatedAt))
: [];
const coordinatorByRun = new Map(coordinators.map((row) => [row.runId, row]));
for (const run of runs) {
const issueId = run.nativeIssueId ?? text(run.contextSnapshot?.issueId);
const matching = recovery.filter(
(row) =>
row.issueId === issueId &&
(row.status !== "resolved" || row.evidence.executionReconciliation || row.evidence.automaticRecovery) &&
(row.evidence.runId === run.id || row.evidence.sourceRunId === run.id),
);
const action =
matching.find((row) => row.status !== "resolved") ?? matching[0];
projections.set(
run.id,
projectExecution(
run,
coordinatorByRun.get(run.id),
pending.filter((row) => row.issueId === issueId),
action,
now,
),
);
}
return projections;
}
export async function executionProjectionForRun(
db: Db,
companyId: string,
runId: string,
now = new Date(),
): Promise<ExecutionProjection | null> {
return (
(await executionProjectionsForRuns(db, companyId, [runId], now)).get(
runId,
) ?? null
);
}
export function projectExecution(
run: Run,
coordinator: Coordinator | undefined,
pending: Array<{ kind: string }>,
recoveryAction: Recovery | undefined,
now = new Date(),
): ExecutionProjection {
const detail = coordinator?.failureDetail ?? {};
const successorRunId = text(detail.successorRunId);
const projection: ExecutionProjection = {
phase: "queued",
label: "Queued",
cause: run.errorCode ?? coordinator?.failureCode ?? null,
lastConfirmedActivityAt:
(
run.lastUsefulActionAt ??
run.lastOutputAt ??
run.startedAt
)?.toISOString() ?? null,
retryAt:
(coordinator?.nextAttemptAt ?? run.scheduledRetryAt)?.toISOString() ??
null,
attempt: coordinator?.attempt ?? executionFailureRetryCount(run) + 1,
maxAttempts: 3,
recoveryOwner: null,
nextAction: text(detail.nextAction) ?? run.nextAction,
permittedActions: ["inspect_run"],
predecessorRunId: run.retryOfRunId,
successorRunId,
};
const set = (phase: ExecutionProjection["phase"], label: string) =>
Object.assign(projection, { phase, label });
if (recoveryAction?.status === "resolved" && recoveryAction.evidence?.automaticRecovery) {
projection.cause = recoveryAction.cause;
projection.nextAction = recoveryAction.nextAction;
const continuation = recoveryAction.evidence.explicitUserContinuation as
{ previousRunId?: unknown; runId?: unknown } | undefined;
const explicitSuccessor = text(continuation?.runId);
if (continuation?.previousRunId === run.id && explicitSuccessor && explicitSuccessor !== run.id) {
// The admission transaction already recorded the user's successor. Keep
// the old failure diagnostic without making it hold the newer attempt.
projection.successorRunId = explicitSuccessor;
projection.nextAction = null;
return set("completed", "Continued in another run");
}
// Diagnostic projection only: no user decision or replay affordance.
return set("recovery_needed", "Stopped");
}
if (
recoveryAction?.status === "resolved" &&
recoveryAction.evidence?.executionReconciliation
) {
if (recoveryAction.evidence.continuationDelivery === "invalidated") {
projection.cause = "continuation_ownership_changed";
projection.nextAction =
"The task was closed or reassigned before delivery. Inspect the current task owner and status.";
return set("failed", "Continuation cancelled");
}
const deliveredRunId = text(recoveryAction.evidence.continuationRunId);
projection.successorRunId = deliveredRunId;
projection.cause = null;
projection.nextAction = deliveredRunId
? null
: "The previous execution is reconciled. Its continuation is queued for delivery.";
return set(
deliveredRunId ? "completed" : "queued",
deliveredRunId ? "Continued in another run" : "Continuation queued",
);
}
if (
(run.status === "running" &&
(run.executionControlDeadlineAt || coordinator?.controlDeadlineAt)) ||
[
"result_persisted",
"ready_for_assessment",
"arbitrating",
"workspace_pending",
].includes(coordinator?.phase ?? "")
)
return set("finishing", "Finishing");
if (successorRunId) return set("completed", "Continued in another run");
if (run.status === "scheduled_retry" && run.scheduledRetryReason === "ai_connection_busy") {
projection.nextAction = "Waiting for the AI subscription's current execution to finish; the scheduled check will revalidate access.";
return set("retry_scheduled", "Waiting for AI subscription");
}
if (run.status === "scheduled_retry" && run.scheduledRetryReason === "workspace_busy") {
projection.nextAction = "Waiting for the live workspace holder to finish; the scheduled check will revalidate ownership.";
return set("retry_scheduled", "Waiting for workspace");
}
// Cleanup can fail before a finalization coordinator exists. The missing
// row must not turn a quarantined native session into an ordinary Retry.
const cleanupQuarantined = run.runtimeMode === "native" &&
["failed", "timed_out"].includes(run.status) &&
run.errorCode === "native_session_cleanup_quarantined";
if (!cleanupQuarantined && (
coordinator?.phase === "retryable_failure" ||
run.status === "scheduled_retry"
)) {
projection.recoveryOwner = "agent";
return set(
projection.retryAt && new Date(projection.retryAt) > now
? "retry_scheduled"
: "reconnecting",
projection.retryAt && new Date(projection.retryAt) > now
? "Retry scheduled"
: "Reconnecting",
);
}
if (coordinator?.phase === "terminal_failure" || recoveryAction || cleanupQuarantined) {
if (
!cleanupQuarantined &&
coordinator?.failureCode === "native_provider_terminal_failed" &&
!detail.replacementDenied &&
run.finishedAt &&
now.getTime() - run.finishedAt.getTime() < EXECUTION_CONTROL_DEADLINE_MS
) {
projection.recoveryOwner = "agent";
projection.nextAction =
"Checking that the previous provider stopped and its action outcomes are known before continuing.";
return set("reconnecting", "Checking recovery");
}
projection.recoveryOwner = "board";
projection.cause =
recoveryAction?.cause ??
text(detail.replacementDenied) ??
projection.cause;
projection.nextAction = recoveryAction?.nextAction ?? projection.nextAction;
if (cleanupQuarantined && !projection.nextAction) {
projection.nextAction = "Verify the stopped session and its saved work before starting a new attempt.";
}
projection.permittedActions.push("inspect_recovery");
return set("recovery_needed", "Recovery needed");
}
if (run.status === "succeeded") {
if (pending.length)
return set(
pending.some((row) => row.kind === "connection_intent")
? "waiting_for_access"
: "waiting_for_answer",
pending.some((row) => row.kind === "connection_intent")
? "Waiting for access"
: "Waiting for answer",
);
return set("completed", "Completed");
}
if (["failed", "cancelled", "timed_out", "interrupted"].includes(run.status))
return set("failed", run.status === "cancelled" ? "Cancelled" : "Failed");
if (run.status === "running") {
const leaseConfirmed =
coordinator?.phase === "observed" &&
coordinator.leaseExpiresAt &&
coordinator.leaseExpiresAt > now;
let processConfirmed = false;
if (run.runtimeMode === "legacy" && run.processPid) {
try {
process.kill(run.processPid, 0);
processConfirmed = true;
} catch {
/* No execution confirmation. */
}
}
return leaseConfirmed || processConfirmed
? set("working", "Working")
: set("reconnecting", "Confirming execution");
}
return projection;
}