Files
PaperClipAI/tests/runner-e2e/chat-setup-health.ts
T
f4337bb5fb test(evals): land the chat catalog eval harness on master (#19614)
## Thinking Path

> - Paperclip is the open source app people use to manage AI agents for
work
> - The chat-feature evals measure whether agents complete real chat
requests through the product
> - The chat catalog eval harness and its repairs live on the
`chat-eval/master-candidate` branch, not on master
> - Today's work made 139 of 242 chat evals pass in verified single-case
live runs. Each product fix landed on master as its own pull request
> - The eval-side changes (setups, graders, fault seams, evidence
capture) still exist only on the candidate
> - This pull request lands the harness on master. Master stays the
source of truth for product behavior
> - The benefit is one tree: evals run against master commits, and
product fixes reach the evals as soon as they merge

## Linked Issues or Issue Description

Refs #19441, Refs #19599, Refs #20270, Refs #20334, Refs #20342, Refs
#20344, Refs #20354, Refs #20360, Refs #20373, Refs #20382, Refs #20409

**What existing behavior does this improve?**
The chat catalog eval harness and the product it evaluates live on a
long-running branch outside master. Eval fixes do not reach master, so
master-based eval runs use old setups and graders.

**Subsystem affected**
`tests/runner-e2e` chat catalog harness, plus a small set of product
changes that exist only on the candidate.

**Proposed change**
Land the harness on master. Every product fix from this eval work
already merged to master as its own pull request. Conflicts keep
master's product behavior.

## What Changed

- Most of the diff (808 of 851 files) is under `tests/runner-e2e/`:
chat-catalog domains, scenarios, descriptors, the TestOps manifest,
private evidence tooling, and their tests.
- Graders now check saved state, setup literals and rendered or clicked
links. Many keyword and regex checks on free-form replies are removed.
Each changed grader has a bumped `oracleVersion` and calibration tests
(a valid reply, a reworded valid reply, and broken variants that must
fail).
- The shared link helper `hasLink` no longer treats a list link (for
example `/COM/issues?view=mine`) as a link to every task without a URL
key.
- Setup fixes build real state through supported product routes: routine
history, agent grants through the board grant route, budget holds with a
real cap, and a storage-layer archive fault for
`inbox-search.archive-partial-failure`.
- Browser boundary journeys (sign-in, chat-disabled, onboarding) now
write the standard pass evidence files.
- Retargeted cases, approved by the maintainer:
`agent-chat.chat-disabled` (Conference Room gate),
`agents.restore-instructions` (no saved instruction history),
`access.manage-membership` (member permission edits stay board-only).
- The private evidence trusted files (`manual-private-handoff.ts`,
`manual-private-retention.ts`, `manual-private-report-assets.ts`) are
byte-identical to master.
- Product files outside `tests/` (43 files, +1,328 −49) carry
candidate-only changes that master does not yet have. Most are tests.
The main code changes are in `request-human-input.ts`,
`heartbeat-run-summary.ts`, `native-completion-feedback.ts`,
`native-run-finalizer.ts`, `openapi.ts`, `secrets.ts` and
`IssueChatLongThreadPerf.tsx`.

## Verification

- `pnpm vitest run chat-catalog-domains chat-catalog-contract.test.ts
chat-catalog-consistency.test.ts unit-launcher.test.mjs
--testTimeout=180000` in `tests/runner-e2e`: 269 files pass, 3,550 tests
pass.
- `node --test chat-full-catalog-testops-manifest.test.mjs`: 10 pass.
- `node tests/runner-e2e/source-sdk-public-inputs.mjs <HEAD> --check`:
no change.
- `tsc` for `server` and `tests/runner-e2e`: clean.
- Live single-case verification: 139 of 242 cases passed on paid runs,
and a person read the saved evidence for each. These are single-case
runs, not one full-cohort run.

## Risks

- Large diff. Most of it is test-only, but the 43 non-test files change
product code paths. Review those files first.
- Some graders are now less strict on reply wording. A person must read
replies for meaning on those cases.
- 42 cases need disposable outside accounts (Notion, Slack, GitHub,
Daytona and others) and stay blocked. `access.invite-teammate` stays
failing until local agent isolation lands, because the runner's durable
state keeps a live invite token.
- The paid workflow compares the trusted private-evidence files with
master. They are unchanged, so this risk is low.

## Model Used

Anthropic Claude Opus 5.5 (`claude-opus-5-5`), with tool use and code
execution, in Claude Code. Subagents used the same model.

## 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>
Co-authored-by: Paperclip <noreply@paperclip.ing>
2026-11-10 08:02:11 -06:00

135 lines
10 KiB
TypeScript

import type { Page } from "@playwright/test";
import { createHash } from "node:crypto";
import { constants, readFileSync } from "node:fs";
import { open, realpath, stat, writeFile } from "node:fs/promises";
import path from "node:path";
import { pollUntil, type RunnerApi } from "./api.js";
import { CHAT_OPENCODE_MODEL } from "./chat-runtime-profile.js";
import { readChatReadinessReceipts } from "./chat-readiness-observer.js";
import type { LiveFixtureValues } from "./live-fixtures.js";
import type { RunnerTaskFixture } from "./types.js";
export const CHAT_SETUP_HEALTH_SUITE = "chat-setup-health";
export const CHAT_SETUP_HEALTH_ID = "chat-setup-health.runner-opencode.local.setup-only";
export class SetupHealthAbort extends Error {
constructor() { super("Setup health cleanup failed; owned supervisor stop requested, no further API observations"); }
}
const stopPath = (root: string) => path.join(root, "artifacts-private", "chat-setup-health-stop.json");
export async function requestSetupHealthStop(root: string) {
await writeFile(stopPath(root), JSON.stringify({ executionId: CHAT_SETUP_HEALTH_ID, temporaryRoot: root }), { flag: "wx", mode: 0o600 });
}
/** Exact one-cell request, consumed only by the existing owned supervisor. */
export async function stopSetupHealthIfRequested(input: { executionIds: readonly string[]; temporaryRoot: string; stop(): Promise<void> }) {
if (input.executionIds.length !== 1 || input.executionIds[0] !== CHAT_SETUP_HEALTH_ID) return false;
let file;
try {
const root = await realpath(input.temporaryRoot), rootInfo = await stat(root);
if (root !== input.temporaryRoot || rootInfo.uid !== process.getuid?.() || (rootInfo.mode & 0o077)) return false;
file = await open(stopPath(root), constants.O_RDONLY | constants.O_NOFOLLOW);
const info = await file.stat();
if (!info.isFile() || info.uid !== rootInfo.uid || (info.mode & 0o077) || info.size > 1024) return false;
const row = JSON.parse(await file.readFile("utf8"));
if (row?.executionId !== CHAT_SETUP_HEALTH_ID || row?.temporaryRoot !== root) return false;
} catch { return false; }
finally { await file?.close(); }
await input.stop();
return true;
}
export const chatSetupHealthTask: RunnerTaskFixture = {
id: "setup-only", label: "One normal OpenCode setup verification; no catalog request", groups: [], flow: "setup_health", workMode: "standard",
expectedRunCount: 0, automaticRetryPolicy: "single_attempt", turnTimeoutMs: 60_000,
attemptTimeoutMs: { local: 180_000, daytona: 180_000 }, expectedTerminalState: { issue: "in_review", run: "succeeded" },
buildTitle: () => "Setup health only", buildPrompt: () => "", buildVisibleMarker: () => "",
buildMatchers: () => [],
};
export const chatSetupHealthDigest = createHash("sha256").update([
"chat-setup-health.ts", "chat-setup-credential.ts", "chat-readiness-observer.ts", "server-entry.ts", "runner.spec.ts", "live-fixtures.ts", "chat-remote-host.ts", "server.ts",
].map(file => readFileSync(new URL(`./${file}`, import.meta.url), "utf8")).join("\n")).digest("hex");
interface HealthAgent { id: string; companyId: string; lifecycleState?: string; lifecycleError?: string | null; status: string;
defaultEnvironmentId?: string | null; adapterType: string; adapterConfig?: Record<string, unknown> }
export function gradeSetupHealth(state: {
ready: boolean; agentCount: number; configMatches: boolean; issues: number; conversations: number; comments: number; humanComments: number; runs: number;
probes: Awaited<ReturnType<typeof readChatReadinessReceipts>>;
}) {
return [
{ id: "ordinary-agent-ready", passed: state.ready && state.agentCount === 1 && state.configMatches },
{ id: "one-ordinary-hello", passed: state.probes.length === 1 && !state.probes[0]!.thrown
&& ["pass", "warn"].includes(state.probes[0]!.status) && state.probes[0]!.checks.some(check => check.code === "opencode_hello_probe_passed" && check.level === "info") },
{ id: "zero-chat-or-task-work", passed: state.issues === 0 && state.conversations === 0 && state.comments === 0 && state.humanComments === 0 && state.runs === 0 },
];
}
export async function runChatSetupHealth(input: { api: RunnerApi; fixtures: LiveFixtureValues; temporaryRoot: string;
evidence(name: string, value: unknown): Promise<void> }) {
const { api, fixtures: f } = input;
let ready = false;
try {
await pollUntil({ label: "ordinary setup-health agent readiness", deadlineAt: Date.now() + 75_000, intervalMs: 250,
load: () => api.get<HealthAgent>(`/api/agents/${f.agent.id}`, { timeout: 5_000 }),
accept: agent => agent.lifecycleState === "ready" && ["idle", "active"].includes(agent.status),
reject: agent => agent.lifecycleError ? "Actual setup-health lifecycle verification failed" : undefined });
ready = true;
} finally {
// GET only: no chat resolution/creation, task dispatch or explicit second probe.
const company = `/api/companies/${f.company.id}`;
let diagnosticFailure = false;
try {
const probes = await readChatReadinessReceipts(input.temporaryRoot, f.company.id);
await input.evidence("chat-setup-health.json", { schema: "paperclip.chat-setup-health/v1", scope: "diagnostic-outside-244",
companyId: f.company.id, agentId: f.agent.id, probes, publicState: "not-yet-verified",
setupProbeUsage: "unavailable outside heartbeat accounting; never zero" });
} catch { diagnosticFailure = true; }
// Always fence future automatic verification, even when diagnostic I/O fails.
try { await api.post(`/api/agents/${f.agent.id}/pause`, {}, { timeout: 5_000 }); }
catch {
let stopRequested = false;
try { await requestSetupHealthStop(input.temporaryRoot); stopRequested = true; } catch { /* Preserve uncertain cleanup. */ }
try { await input.evidence("chat-setup-health-cleanup.json", { scope: "diagnostic-outside-244", pauseFailed: true, stopRequested, cleanup: "failed" }); } catch { /* Do not delay owned stop for diagnostics. */ }
throw new SetupHealthAbort();
}
if (diagnosticFailure) throw new Error("Setup health diagnostics unavailable after normal public pause");
// Read after the pause fence too: a late second completion must fail the count.
const probes = await readChatReadinessReceipts(input.temporaryRoot, f.company.id);
const settings = await api.get<{ enableAgentChat: boolean }>("/api/instance/settings/experimental", { timeout: 5_000 });
if (!settings.enableAgentChat) await api.patch("/api/instance/settings/experimental", { enableAgentChat: true });
const [agents, issues, conversations, runs] = await Promise.all([
api.get<HealthAgent[]>(`${company}/agents`, { timeout: 5_000 }), api.get<Array<{ id: string }>>(`${company}/issues?includeAgentCreatedTasks=true&limit=100`, { timeout: 5_000 }),
api.get<Array<{ id: string }>>(`${company}/chats`, { timeout: 5_000 }), api.get<unknown[]>(`${company}/heartbeat-runs?limit=100`, { timeout: 5_000 }),
]);
const sourceIds = [...new Set([...issues, ...conversations].map(row => row.id))];
const comments = (await Promise.all(sourceIds.map(id => api.get<Array<{ authorUserId?: string | null }>>(`/api/issues/${id}/comments?order=asc`, { timeout: 5_000 })))).flat();
const agent = agents.find(row => row.id === f.agent.id);
const state = { ready, agentCount: agents.length, configMatches: agent?.companyId === f.company.id && agent.defaultEnvironmentId === f.environment.id
&& agent.adapterType === "paperclip_runner" && agent.adapterConfig?.provider === "opencode" && agent.adapterConfig?.model === CHAT_OPENCODE_MODEL,
issues: issues.length, conversations: conversations.length, comments: comments.length, humanComments: comments.filter(row => row.authorUserId).length,
runs: runs.length, probes };
const checks = gradeSetupHealth(state);
await input.evidence("chat-setup-health.json", { schema: "paperclip.chat-setup-health/v1", scope: "diagnostic-outside-244", companyId: f.company.id,
agentId: f.agent.id, state, checks, canonicalQueries: state.humanComments,
setupProbeUsage: "One ordinary automatic lifecycle hello is expected. Provider tokens and cost are unavailable outside heartbeat accounting; never reported as zero." });
await input.evidence("api-state.json", { scope: "diagnostic-outside-244", capturePhase: "after-public-pause",
companyId: f.company.id, agentId: f.agent.id, observed: state, checks });
if (checks.some(check => !check.passed)) throw new Error("Setup health failed; inspect retained closed readiness diagnostics");
}
}
/** Genuine readonly diagnostic UI evidence, never published as a task screenshot. */
export async function captureSetupHealthProfile(input: {
executionId: string; api: RunnerApi; page: Page; companyId: string; agentId: string;
capturePrivate(id: string, filename: string): Promise<void>;
}) {
if (input.executionId !== CHAT_SETUP_HEALTH_ID) throw new Error("Setup health evidence requires its exact diagnostic cell");
const [company, agent] = await Promise.all([
input.api.get<{ id: string; issuePrefix: string }>(`/api/companies/${input.companyId}`),
input.api.get<{ id: string; companyId: string; name: string; urlKey: string }>(`/api/agents/${input.agentId}`),
]);
if (company.id !== input.companyId || agent.id !== input.agentId || agent.companyId !== company.id
|| !company.issuePrefix || !agent.name || !agent.urlKey) throw new Error("Setup health evidence resource mismatch");
const origin = new URL(input.api.baseURL);
if (origin.protocol !== "http:" || origin.hostname !== "127.0.0.1" || origin.username || origin.password)
throw new Error("Setup health evidence requires the owned loopback app");
const url = new URL(`/${encodeURIComponent(company.issuePrefix)}/agents/${encodeURIComponent(agent.urlKey)}/overview`, origin).href;
await input.page.goto(url, { waitUntil: "domcontentloaded" });
await input.page.getByRole("main").getByRole("heading", { name: agent.name, exact: true }).waitFor({ state: "visible" });
if (input.page.url() !== url) throw new Error("Setup health evidence route mismatch");
await input.capturePrivate("setup-health-final-state", "final-state.png");
}