Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/runtime-infrastructure.test.ts
T
DottaandClaude Opus 5.5 2eda54162a fix(evals): grade recovery on fresh continuation, accept in-turn queued-comment reads, verify the saved service model (#21210)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat eval catalog checks that agents in chat do the right thing,
graded on saved product state and rendered links
> - Full cohort 8 (OpenCode,
`openrouter/deepseek/deepseek-v4.1-flash-1017`, master 7747487ee7)
failed six cases after two deliberate product changes and one harness
race
> - #21135 continues an unfinished run in a fresh run from saved work
instead of leaving an exhausted, board-owned recovery action. Three
recovery cases still waited for that action.
> - #21175 counts a `get_task_history` read of a queued comment's
current revision as delivery. Two steering cases still required a
continuation run.
> - `companies.prompt-injection-guard` read the settings back before its
own save finished
> - This pull request grades the recovery and steering cases on the new
contracts and makes the settings harness verify the saved model before
it continues
> - The benefit is that the evals accept the intended product behavior
and still fail a broken product

## Linked Issues or Issue Description

Refs #21135
Refs #21175

**What happened?**
Cohort 8 failed `recovery.explain-stall`, `recovery.retry-exhausted` and
`recovery.source-action`: the harness waited 30 s for an "exhausted
source recovery action with no automatic wake". #21135 now logs
"Unfinished run continues in a fresh run from saved work" (`trigger:
recovery_refused`, `delivery: recorded`) instead.
`steering.edit-queued-message` and `agent-chat.multiple-handoffs` failed
their run counts: the running worker turn read the queued comment
(current revision) and applied it, so no continuation run was queued.
`companies.prompt-injection-guard` read `openai/gpt-oss-safeguard-20b`:
the server log shows the harness GET finished before the settings PUT.

**Expected behavior**
Infrastructure loss continues automatically in one fresh run, the chat
explains it, and the continuation finishes the task. A queued comment is
consumed once, in the running turn or in one continuation. The settings
harness tests only the model it saved.

**Steps to reproduce**
Run the six `chat-features.runner-opencode.local.*` cases on master
after #21135 and #21175.

## What Changed

- `runtime-infrastructure.ts` (`recovery.explain-stall`,
`recovery.source-action`, oracle `infrastructure-recovery.v6`): after
three real Runner losses, the setup waits for one system-owned
fresh-continuation record of the failed run (`?inspectSourceRunId`),
resolved as `restored`, `replay: "none"`, the failed run marked
superseded, and one run with `retryOfRunId` = the failed run that waits
for its brief in its own workspace. The chat question is asked while it
runs. The harness then releases the brief, and the grade requires the
continuation to succeed, save the brief's reference as the latest
revision and complete the original task. No person-owned hold and no
blocked task at the reply. Three runs: failed 1, succeeded 2. The reply
must still come from a run that read the recovery record (now also the
continuation run) and link the original task. Setup timing is saved to
`runtime-fresh-continuation.json`.
- `runtime-infrastructure-retry.ts` (`recovery.retry-exhausted`, oracle
`retry-exhausted.v6`): an infrastructure loss no longer offers Retry.
The case uses a failure only the person can resolve
(`native_provider_model_rejected`, a `PERSON_OWNED_FAILURE_CODES`
member) and keeps the Retry-under-budget-hold journey unchanged. The
setup proves the task waits blocked with no fresh continuation and no
successor run for 45 s.
- `runtime-busy-worker.ts`: the Retry listener helpers move here; a
listener counts only when its `.waiting` file is newer than the run's
start.
- `runtime-queue.ts`: `historyReadRevisions`, `readCurrentRevision`,
`runHistoryReads` read `task_history.read` events; `settledQueue` needs
a 30 s quiet settled state.
- `runtime-queue-controls.ts` (`steering.edit-queued-message`, oracle
`structural.v5`) and `runtime-multiple-handoffs.ts` (oracle `v5`):
accept in-turn consumption only with a worker `task_history.read` of
every queued comment's current revision, the change saved as the latest
revision (blue; "Launch sooner"), an empty queue, a done task and zero
continuations. The continuation path stays: exactly one continuation
with the unread comments in order, when the turn did not read the
current revision. Run bounds 2–3 and 4–5. The worker brief no longer
tells the agent to leave follow-ups for a continuation.
- `identity-company-models.ts` (`prompt-injection-guard` oracle
`prompt-injection-guard.v2`, `fast-responses` `v4`): before saving,
check the model field holds the requested model; wait for the PUT;
require 200, the sent model, the saved model, then a matching read-back.
- `recovery.yaml`: context and expected outcome for the three recovery
cases. Recorded in `chat-full-catalog-testops-manifest.json`
`changedFields` and pinned in its test. Descriptors and the inventory
digest are updated. `maps/feature-map/recovery.md` exhaustion line
corrected.

## Verification

- `npx vitest run --config tests/runner-e2e/vitest.config.ts
tests/runner-e2e/chat-catalog tests/runner-e2e/chat-catalog-domains`:
283 files, 3895 tests pass. New calibration tests cover valid,
reworded-valid and broken cases: no continuation, a person-owned hold,
two continuation records or runs, a replayed turn, an unfinished
continuation, a missing saved result; an old-revision-only read, a
missing outcome, a double application, two continuations; a model field
that lost the model, a different sent model, a wrong saved model.
- `pnpm test:e2e:runner:typecheck`: pass.
- `node --test
tests/runner-e2e/chat-full-catalog-testops-manifest.test.mjs`: 10/10
pass.
- `node tests/runner-e2e/source-sdk-public-inputs.mjs <head sha>
--check`: `changed: false`.
- `node scripts/check-maps.mjs`: aligned.
- Paid run: Runner Full-Stack E2E 43321773136 at 9de2bd5a6f, six cases:
prompt-injection-guard passed. The two fresh-continuation cases found a
product failure (the continuation never started), fixed in #21217.
retry-exhausted showed that a model-rejected failure offers no Retry
control. multiple-handoffs failed correctly: the turn read the
instruction and did not apply it. edit-queued-message did not reach the
new grade (6-minute CEO turn). Details are in a PR comment.

## Risks

- The recovery cases depend on the fresh continuation starting within 5
minutes. The paid run showed it did not start; #21217 fixes that. Until
#21217 merges, those cases fail on master.
- `recovery.retry-exhausted` expects a Retry control that the product
does not show for a model-rejected failure. It fails until a product
decision is made.
- `recovery.retry-exhausted` no longer covers infrastructure exhaustion;
the two other recovery cases cover the new automatic path.

> For core feature work, check [`ROADMAP.md`](ROADMAP.md) first and
discuss it in `#dev` before opening the PR. Feature PRs that overlap
with planned core work may need to be redirected — check the roadmap
first. See `CONTRIBUTING.md`.

## Model Used

- Claude Opus 5.5 (`claude-opus-5-5`), Anthropic, used through Claude
Code with tool use and code execution.

## 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
- [ ] All Paperclip CI gates are green
- [ ] Greptile is 5/5 with no open P2s, recommendations, or follow-ups
- [x] I will address all Greptile and reviewer comments before
requesting merge

🤖 Generated with [Claude Code](https://claude.com/claude-code)

---------

Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
2026-11-12 17:26:50 -06:00

119 lines
13 KiB
TypeScript

import { expect, it, vi } from "vitest";
import { freshContinuationRecord, gradeInfrastructureRecovery, nativeInfrastructureLoss, readSourceRecoveryRecord } from "./runtime-infrastructure.js";
vi.mock("./runtime-busy-worker.js", async importOriginal => ({ ...await importOriginal<typeof import("./runtime-busy-worker.js")>(),
busyRuntimeWorker: async () => ({ worker: { id: "morgan" }, task: { id: "website" }, gate: {}, workerRunId: "run" }) }));
vi.mock("./runtime-native-fault.js", async importOriginal => ({ ...await importOriginal<typeof import("./runtime-native-fault.js")>(),
faultOwnedNativeProcess: async () => ({ id: "run", processPid: 7, errorCode: "native_runner_process_exited", status: "running", nativePhase: "running" }) }));
import { savedCall } from "./api-read-receipts.test-fixtures.js";
function record(overrides: Record<string, unknown> = {}) {
return { id: "recovery", companyId: "company", sourceIssueId: "website", kind: "active_run_watchdog", status: "resolved", outcome: "restored", cause: "native_session_retry_exhausted",
ownerType: "system", ownerAgentId: null, ownerUserId: null, returnOwnerAgentId: "morgan", recoveryIssueId: null, wakePolicy: null,
nextAction: "The task continues in a fresh run from its last saved work.", resolutionNote: "continued_fresh_from_last_saved",
evidence: { runId: "run", continuationDelivery: "delivered", continuationRunId: "next",
freshContinuation: { schema: "paperclip.fresh-continuation.v1", sourceRunId: "run", trigger: "recovery_refused", failureCode: "native_runner_process_exited", delivery: "recorded", isolation: "fresh_isolated_workspace" },
freshContinuationSettlement: { schema: "paperclip.fresh-continuation.v1", actorType: "system", actorId: "fresh-continuation", decision: "continue_in_fresh_run", replay: "none", sourceRunId: "run" } }, ...overrides };
}
function valid(id = "recovery.explain-stall") {
return { scenarioId: id, origin: "http://paperclip.test", companyId: "company", leadId: "ceo", agents: [], links: ["/COM/issues/COM-2"], humanQueries: [], reply: "", tasks: [{ id: "website", status: "done", assigneeAgentId: "morgan" }], activity: [],
runs: [{ id: "run", agentId: "morgan", status: "failed" }, { id: "next", agentId: "morgan", status: "succeeded" }, { id: "source", agentId: "ceo", status: "succeeded" }],
state: { task: { id: "website", identifier: "COM-2" }, queryRecoveryRead: true, worker: { id: "morgan" },
failure: { id: "run", status: "failed", resultJson: { freshContinuationSupersededAt: "2026-11-12T21:50:53.372Z" } },
continuation: record(), recoveryAfter: { active: null, actions: [record()] }, continuationRunId: "next",
continuationRun: { id: "next", companyId: "company", agentId: "morgan", retryOfRunId: "run", contextSnapshot: { issueId: "website" } },
activeHoldAtReply: null as unknown, taskStatusAtReply: "in_progress", continuationResult: true,
setupFaults: [1, 2, 3].map(pid => ({ runId: "run", pid, errorCode: "native_runner_process_exited" })),
queryReply: "Morgan's Website run lost its Runner three times. Paperclip continued the task in a fresh run from its saved work; nothing is needed from you." } };
}
const grade = (o: ReturnType<typeof valid>) => gradeInfrastructureRecovery(o).passed;
it("accepts one system-owned fresh continuation that finishes the original task", () => expect(grade(valid())).toBe(true));
it("accepts the same contract for the source-action question", () => expect(grade(valid("recovery.source-action"))).toBe(true));
it("does not grade reply wording", () => { const o = valid(); o.state.queryReply = "It got cut off and picked back up on its own; see the task."; expect(grade(o)).toBe(true); });
it("accepts the recovery record that keeps an earlier hold as the continuation's primary record", () => {
const o = valid(), hold = record({ kind: "active_run_watchdog", cause: "native_session_retry_exhausted", evidence: { ...record().evidence, automaticRecoveryHistory: { policy: "preserve_without_replay_v1" } } });
o.state.continuation = hold; o.state.recoveryAfter = { active: null, actions: [hold] }; expect(grade(o)).toBe(true);
});
it("accepts a system-owned active hold at the reply", () => { const o = valid(); o.state.activeHoldAtReply = { ownerType: "system" }; expect(grade(o)).toBe(true); });
it("rejects a task left blocked for a person", () => { const o = valid(); o.state.taskStatusAtReply = "blocked"; expect(grade(o)).toBe(false); });
it("rejects a board-owned recovery action the person must act on", () => { const o = valid(); o.state.activeHoldAtReply = { ownerType: "board" }; expect(grade(o)).toBe(false); });
it("rejects a continuation record owned by a person", () => { const o = valid(), owned = record({ ownerType: "board" }); o.state.recoveryAfter = { active: null, actions: [owned] }; expect(grade(o)).toBe(false); });
it("rejects a continuation that replays the interrupted turn", () => {
const o = valid(), replay = record({ evidence: { ...record().evidence, freshContinuationSettlement: { ...record().evidence.freshContinuationSettlement, replay: "exact" } } });
o.state.recoveryAfter = { active: null, actions: [replay] }; expect(grade(o)).toBe(false);
});
it("rejects two continuation records for one interrupted run", () => { const o = valid(); o.state.recoveryAfter = { active: null, actions: [record(), record({ id: "second" })] }; expect(grade(o)).toBe(false); });
it("rejects a replaced continuation record", () => { const o = valid(); o.state.recoveryAfter = { active: null, actions: [record({ id: "replacement" })] }; expect(grade(o)).toBe(false); });
it("rejects missing continuation evidence", () => { const o = valid(); o.state.recoveryAfter = { active: null, actions: [] }; expect(grade(o)).toBe(false); });
it("rejects an interrupted run that keeps its write authority", () => { const o = valid(); o.state.failure.resultJson = {} as never; expect(grade(o)).toBe(false); });
it("rejects a second continuation run", () => { const o = valid(); o.runs.push({ id: "again", agentId: "morgan", status: "succeeded" }); expect(grade(o)).toBe(false); });
it("rejects a run that does not continue the interrupted one", () => { const o = valid(); o.state.continuationRun.retryOfRunId = "other"; expect(grade(o)).toBe(false); });
it("rejects a continuation that never finishes", () => { const o = valid(); o.runs[1]!.status = "failed"; expect(grade(o)).toBe(false); });
it("rejects a continuation that never saves its result", () => { const o = valid(); o.state.continuationResult = false; expect(grade(o)).toBe(false); });
it("rejects an unfinished original task", () => { const o = valid(); o.tasks[0]!.status = "in_progress"; expect(grade(o)).toBe(false); });
it("rejects relabelled unavailable-model failure as infrastructure loss", () => { const o = valid(); o.state.setupFaults[0]!.errorCode = "native_provider_model_rejected"; expect(grade(o)).toBe(false); });
it("rejects a continuation recorded for a person-owned failure", () => {
const o = valid(), person = record({ evidence: { ...record().evidence, freshContinuation: { ...record().evidence.freshContinuation, failureCode: "native_provider_model_rejected" } } });
o.state.recoveryAfter = { active: null, actions: [person] }; expect(grade(o)).toBe(false);
});
it("rejects three receipts for one reused PID", () => { const o = valid(); o.state.setupFaults[1]!.pid = 1; expect(grade(o)).toBe(false); });
it("rejects an answer whose run never read the saved recovery record", () => { const o = valid(); o.state.queryRecoveryRead = false; expect(grade(o)).toBe(false); });
it("rejects an answer that does not link the original task", () => { const o = valid(); o.links = ["/COM/issues/COM-9"]; expect(grade(o)).toBe(false); });
it("selects only the continuation of this run in this task", () => {
const run = { id: "run", status: "failed", resultJson: { freshContinuationSupersededAt: "2026-11-12T21:50:53.372Z" } } as never;
expect(freshContinuationRecord({ actions: [record() as never] }, run, "company", "website", "morgan")?.id).toBe("recovery");
expect(freshContinuationRecord({ actions: [record() as never] }, run, "company", "other", "morgan")).toBeUndefined();
expect(freshContinuationRecord({ actions: [record() as never] }, run, "company", "website", "riley")).toBeUndefined();
expect(freshContinuationRecord({ actions: [record({ status: "active" }) as never] }, run, "company", "website", "morgan")).toBeUndefined();
});
it("accepts a genuine call_api recovery read after the port redacts its body, bound to the source task or failed run", () => {
const read = (operationId: string, pathParams: Record<string, string>, status = 200) => savedCall({ operationId, pathParams }, status, "3", "run-1", { sourceIssueId: "website", runId: "run" });
expect(readSourceRecoveryRecord(read("GET /api/issues/{id}/recovery-actions", { id: "website" }), "website", "run", "COM-1")).toBe(true);
expect(readSourceRecoveryRecord(read("GET /api/issues/{id}", { id: "COM-1" }), "website", "run", "COM-1")).toBe(true);
expect(readSourceRecoveryRecord(read("GET /api/heartbeat-runs/{runId}", { runId: "run" }), "website", "run", "COM-1")).toBe(true);
expect(readSourceRecoveryRecord(read("GET /api/issues/{id}", { id: "COM-2" }), "website", "run", "COM-1")).toBe(false);
expect(readSourceRecoveryRecord(read("GET /api/heartbeat-runs/{runId}", { runId: "other" }), "website", "run", "COM-1")).toBe(false);
expect(readSourceRecoveryRecord(read("GET /api/heartbeat-runs/{runId}", { runId: "next" }), "website", "run", "COM-1", "next")).toBe(true);
expect(readSourceRecoveryRecord(read("GET /api/heartbeat-runs/{runId}", { runId: "next" }), "website", "run", "COM-1")).toBe(false);
expect(readSourceRecoveryRecord(read("GET /api/issues/{id}/recovery-actions", { id: "website" }, 403), "website", "run", "COM-1")).toBe(false);
});
it("reads the recovery record only from exact product tool output", () => {
const event = (output: string, status = "completed") => ({ eventType: "tool.execution.completed", payload: { prpEvent: { sourceKind: "runner", payload: { status, output } } } });
const actions = '{"apiOperationId":"GET /api/issues/{id}/recovery-actions","body":{"active":{"sourceIssueId":"website"}}}';
const run = '{\\"apiOperationId\\": \\"GET /api/heartbeat-runs/{runId}\\", \\"id\\": \\"run\\"}';
expect(readSourceRecoveryRecord([event(actions)], "website", "run")).toBe(true);
expect(readSourceRecoveryRecord([event(run)], "website", "run")).toBe(true);
expect(readSourceRecoveryRecord([event(actions, "failed")], "website", "run")).toBe(false);
expect(readSourceRecoveryRecord([event(actions)], "other", "run")).toBe(false);
expect(readSourceRecoveryRecord([event('{"apiOperationId":"GET /api/issues/{id}/recovery-actions/x"} website')], "website", "run")).toBe(false);
expect(readSourceRecoveryRecord([event("I read the website recovery actions")], "website", "run")).toBe(false);
// The task read counts only when its saved blocker names this failed run.
const task = (body: string) => `{"apiOperationId":"GET /api/issues/{id}","data":${body}}`;
expect(readSourceRecoveryRecord([event(task('{"id":"website","executionBlocker":{"runId":"run","cause":"native_session_retry_exhausted"}}'))], "website", "run")).toBe(true);
expect(readSourceRecoveryRecord([event(task('{"id":"website","executionBlocker":null}'))], "website", "run")).toBe(false);
expect(readSourceRecoveryRecord([event(task('{"id":"website","executionRunId":"next"}'))], "website", "run", null, "next")).toBe(true);
expect(readSourceRecoveryRecord([event(task('{"id":"other","executionBlocker":{"runId":"run"}}'))], "website", "run")).toBe(false);
expect(readSourceRecoveryRecord([event(task('{"id":"website","executionBlocker":{"runId":"run"}}'), "failed")], "website", "run")).toBe(false);
expect(readSourceRecoveryRecord([event('{"apiOperationId":"GET /api/issues/{id}/documents/{key}","data":"website run"}')], "website", "run")).toBe(false);
});
it("rejects duplicate repair chains", () => { const o = valid(); o.tasks.push({ id: "repair", status: "todo", assigneeAgentId: "morgan" }); expect(grade(o)).toBe(false); });
it("rejects missing actual source recovery evidence", () => expect(gradeInfrastructureRecovery(undefined).passed).toBe(false));
function lossContext(successorAfter: number, successor: Record<string, unknown> = {}) {
let faults = 0, pid = 100;
const next = { id: "next", companyId: "company", agentId: "morgan", retryOfRunId: "run", contextSnapshot: { issueId: "website" }, ...successor };
return { input: { fixtures: { company: { id: "company" } }, evidence: async () => {}, api: { get: async (path: string) => {
if (path === "/api/heartbeat-runs/next") return next;
return { id: "run", status: "running", processPid: ++pid, nativePhase: "running" };
} } },
allRuns: async () => { faults++; return [{ id: "run" }, ...(faults >= successorAfter ? [{ id: "next" }] : [])]; } };
}
it("accepts the fresh continuation of the interrupted run appearing right after the last Runner loss", async () => {
await expect(nativeInfrastructureLoss(lossContext(3) as never)).resolves.toMatchObject({ setupFaults: [{}, {}, {}] });
});
it("rejects a replacement run before the last Runner loss", async () => {
await expect(nativeInfrastructureLoss(lossContext(2) as never)).rejects.toThrow(/undeclared replacement/);
});
it("rejects a run after the last Runner loss that does not continue the interrupted run", async () => {
await expect(nativeInfrastructureLoss(lossContext(3, { retryOfRunId: "other" }) as never)).rejects.toThrow(/undeclared replacement/);
await expect(nativeInfrastructureLoss(lossContext(3, { agentId: "riley" }) as never)).rejects.toThrow(/undeclared replacement/);
});