mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-14 00:26:28 +01:00
## 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>
185 lines
13 KiB
TypeScript
185 lines
13 KiB
TypeScript
import { readChatReadinessReceipts } from "./chat-readiness-observer.js";
|
|
import { legacyChatFeatureFingerprint } from "./chat-catalog-fingerprint.js";
|
|
import { runChatCatalogFlow } from "./chat-catalog-flow.js";
|
|
import { chatCatalogScenarios } from "./chat-catalog-inventory.js";
|
|
import { expect } from "@playwright/test";
|
|
import { pollUntil } from "./api.js";
|
|
import { collectChatRunEvidence, sendChatMessage, type ChatFlowInput, type ChatIssue, type ChatRun } from "./chat-flow.js";
|
|
import { FixtureRegistry } from "./fixture-registry.js";
|
|
import { codexSubscriptionAgentConfig } from "./codex-subscription-fixture.js";
|
|
import { CHAT_FEATURE_BUDGET_CENTS, CHAT_FEATURE_GRADER, chatFeatureScenarios, type ChatFeatureCase } from "./chat-feature-cases.js";
|
|
import { chatFeatureExecutionRuns, gradeChatFeature, linksFeatureTask, type ChatFeatureObservation, type FeatureDocument, type FeatureTask } from "./chat-feature-scoring.js";
|
|
|
|
export async function runChatFeatureFlow(context: {
|
|
input: ChatFlowInput; issue(): ChatIssue | undefined; allRuns(): Promise<ChatRun[]>; refreshIssue(): Promise<void>; selectSlackSource?(issueId: string): Promise<void>;
|
|
}) {
|
|
const { input } = context;
|
|
const { api, page, fixtures: f, execution, nonce } = input;
|
|
if (!["create-backlog", "delegate-announcement"].includes(execution.task.id)) {
|
|
const scenario = chatCatalogScenarios.find(row => row.id === execution.task.id);
|
|
if (!scenario) throw new Error(`Unknown catalog scenario: ${execution.task.id}`);
|
|
return runChatCatalogFlow({ ...context, scenario });
|
|
}
|
|
const caseId = execution.task.id as ChatFeatureCase;
|
|
await input.evidence("chat-catalog-definition.json", { scenarioId: `tasks.${caseId}`,
|
|
...legacyChatFeatureFingerprint(execution), suiteDefinitionHash: execution.suiteDefinitionHash, journeyKind: "agent-chat" });
|
|
const scenario = chatFeatureScenarios.find(row => row.id === `tasks.${caseId}`);
|
|
if (!scenario) throw new Error(`Unregistered chat feature scenario: ${caseId}`);
|
|
const marker = execution.task.buildVisibleMarker(nonce);
|
|
const brief = `Acorn Notes launches on November 12. The invitation code is ${marker}. Write a welcoming announcement that includes the product name, date and invitation code.`;
|
|
const company = `/api/companies/${f.company.id}`;
|
|
const setupObservations: Array<Record<string, unknown>> = [];
|
|
const registry = new FixtureRegistry();
|
|
registry.register<{ id: string }>({ id: "project", setup: () => api.post(`${company}/projects`, { name: "Launch", description: brief }) });
|
|
registry.register<{ id: string }>({ id: "worker", dependencies: ["project"], async setup() {
|
|
const base = execution.profile.buildAgent({ environmentId: f.environment.id, environmentFixtureId: execution.environment.id,
|
|
workspacePath: input.workspacePath, secretRefs: f.secretRefs, executionId: nonce });
|
|
const worker = { ...base, name: "Morgan", role: "engineer", title: "Engineer", reportsTo: f.agent.id,
|
|
capabilities: "Writes clear product announcements using the project brief.",
|
|
runtimeConfig: { ...(base.runtimeConfig as Record<string, unknown>),
|
|
heartbeat: { enabled: false, maxConcurrentRuns: 1, maxDailyRuns: 1, maxDailyCostCents: CHAT_FEATURE_BUDGET_CENTS } } };
|
|
return api.post(`${company}/agents`, execution.profile.credential === "CODEX_AUTH_JSON"
|
|
? codexSubscriptionAgentConfig(worker, process.env, f.company.id) : worker);
|
|
} });
|
|
const fixtures = await registry.setupAll();
|
|
const projectId = (fixtures.values.get("project") as { id: string }).id;
|
|
const workerId = (fixtures.values.get("worker") as { id: string }).id;
|
|
let observation: ChatFeatureObservation = { caseId, companyId: f.company.id, leadId: f.agent.id, workerId, projectId,
|
|
source: null, tasks: [], runs: [], comments: [], documents: [], activity: [], query: scenario.query, marker, renderedLinks: [] };
|
|
let failure: unknown;
|
|
async function refresh() {
|
|
await context.refreshIssue();
|
|
const source = context.issue();
|
|
const [tasks, runs, comments] = await Promise.all([
|
|
api.get<FeatureTask[]>(`${company}/issues?includeAgentCreatedTasks=true&limit=100`), context.allRuns(),
|
|
source ? api.get<ChatFeatureObservation["comments"]>(`/api/issues/${source.id}/comments?order=asc`) : [],
|
|
]);
|
|
const task = tasks.length === 1 ? tasks[0] : undefined;
|
|
const documents: FeatureDocument[] = [];
|
|
if (task) {
|
|
const summaries = await api.get<Array<{ key: string }>>(`/api/issues/${task.id}/documents`);
|
|
for (const summary of summaries) {
|
|
const endpoint = `/api/issues/${task.id}/documents/${encodeURIComponent(summary.key)}`;
|
|
const [detail, revisions] = await Promise.all([api.get<Omit<FeatureDocument, "revisions">>(endpoint), api.get<FeatureDocument["revisions"]>(`${endpoint}/revisions`)]);
|
|
documents.push({ ...detail, issueId: task.id, revisions });
|
|
}
|
|
}
|
|
const renderedLinks: ChatFeatureObservation["renderedLinks"] = [];
|
|
for (const comment of comments.filter(row => row.authorAgentId === f.agent.id)) {
|
|
const reply = page.locator(`[id=${JSON.stringify(`comment-${comment.id}`)}]`);
|
|
const links = await reply.locator("a[href]").evaluateAll(elements => elements.map(element => element.getAttribute("href")!));
|
|
for (const href of links) {
|
|
const url = new URL(href, api.baseURL);
|
|
renderedLinks.push({ commentId: comment.id, href: url.origin === api.baseURL ? `${url.pathname}${url.search}${url.hash}` : href });
|
|
}
|
|
}
|
|
observation = { ...observation, source: source ?? null, tasks, runs, comments, documents, renderedLinks,
|
|
activity: task ? await api.get<ChatFeatureObservation["activity"]>(`/api/issues/${task.id}/activity`) : [] };
|
|
return gradeChatFeature(observation);
|
|
}
|
|
async function saveEvidence() {
|
|
await input.evidence("chat-feature.json", { schema: CHAT_FEATURE_GRADER, scenarioId: scenario!.id,
|
|
capabilities: scenario!.capabilities, observation, grade: gradeChatFeature(observation),
|
|
setupObservations, setupProbeReceipts: await readChatReadinessReceipts(process.env.PAPERCLIP_RUNNER_E2E_TEMP_ROOT ?? "", f.company.id),
|
|
setupUsageCoverage: "Production agent hello probes are outside heartbeat usage; their detailed tokens and cost are unavailable." });
|
|
}
|
|
try {
|
|
expect(await api.get(`${company}/issues?includeAgentCreatedTasks=true`)).toEqual([]);
|
|
for (const agentId of [f.agent.id, workerId]) {
|
|
await pollUntil({ label: `chat feature agent readiness ${agentId}`, deadlineAt: Date.now() + 45_000, intervalMs: 500,
|
|
load: async () => {
|
|
const agent = await api.get<{ lifecycleState: string; status: string; lifecycleError?: string; lifecycleHolds?: string[] }>(`/api/agents/${agentId}`);
|
|
setupObservations.push({ agentId, lifecycleState: agent.lifecycleState, status: agent.status,
|
|
lifecycleErrorPresent: !!agent.lifecycleError, lifecycleHeld: !!agent.lifecycleHolds?.length });
|
|
return agent;
|
|
},
|
|
accept: agent => agent.lifecycleState === "ready" && ["idle", "active"].includes(agent.status),
|
|
reject: agent => agent.lifecycleError ? "Agent setup failed; inspect retained setup state"
|
|
: agent.lifecycleHolds?.length ? "Agent setup held" : undefined,
|
|
});
|
|
}
|
|
const budgets = await Promise.all([api.get<Record<string, unknown>>(company), ...[f.agent.id, workerId].map(id => api.get<Record<string, unknown>>(`/api/agents/${id}`))]);
|
|
expect(budgets.every(record => record.budgetMonthlyCents === CHAT_FEATURE_BUDGET_CENTS), "Saved company and both agent budget limits").toBe(true);
|
|
const scopes = [{ scopeType: "company", scopeId: f.company.id }, ...[f.agent.id, workerId].map(scopeId => ({ scopeType: "agent", scopeId }))];
|
|
for (const scope of scopes) await api.post(`${company}/budgets/policies`, { ...scope,
|
|
amount: CHAT_FEATURE_BUDGET_CENTS, hardStopEnabled: true, isActive: true, unpricedUsagePolicy: "allow" });
|
|
const { policies } = await api.get<{ policies: Array<{ scopeType: string; scopeId: string; amount: number; hardStopEnabled: boolean; isActive: boolean; unpricedUsagePolicy: string }> }>(`${company}/budgets/overview`);
|
|
const funding = policies.filter(policy => scopes.some(scope => scope.scopeType === policy.scopeType && scope.scopeId === policy.scopeId))
|
|
.map(({ scopeType, scopeId, amount, hardStopEnabled, isActive, unpricedUsagePolicy }) => ({ scopeType, scopeId, amount, hardStopEnabled, isActive, unpricedUsagePolicy }));
|
|
expect(funding).toHaveLength(3);
|
|
expect(funding.every(policy => policy.amount === CHAT_FEATURE_BUDGET_CENTS && policy.hardStopEnabled && policy.isActive && policy.unpricedUsagePolicy === "allow")).toBe(true);
|
|
await input.evidence("chat-feature-setup.json", { scenarioId: scenario.id, companyId: f.company.id,
|
|
leadId: f.agent.id, workerId, projectId, brief, budgetMonthlyCents: CHAT_FEATURE_BUDGET_CENTS,
|
|
maximumProviderRuns: execution.task.expectedRunCount, providerRunTimeoutMs: execution.task.turnTimeoutMs,
|
|
setupObservations, funding, instructions: "Production role defaults; no fixture procedure or expected outcome is injected." });
|
|
await sendChatMessage(page, scenario.query);
|
|
let settledAt: number | undefined;
|
|
let finishedWorkAt: number | undefined;
|
|
await pollUntil({ label: `chat scenario ${scenario.id}`, deadlineAt: Date.now() + 300_000, intervalMs: 1_000,
|
|
load: refresh,
|
|
accept: grade => {
|
|
if (!grade.passed) { settledAt = undefined; return false; }
|
|
settledAt ??= Date.now();
|
|
return Date.now() - settledAt >= 5_000;
|
|
},
|
|
reject: grade => {
|
|
const runs = chatFeatureExecutionRuns(observation);
|
|
if (runs.length > execution.task.expectedRunCount) return "Chat exceeded its declared run bound";
|
|
if (runs.some(run => ["failed", "cancelled", "timed_out"].includes(run.status))) return "An observed provider run failed";
|
|
const finished = runs.length === execution.task.expectedRunCount && runs.every(run => run.status === "succeeded")
|
|
&& observation.source?.conversationState === "waiting";
|
|
if (!finished) finishedWorkAt = undefined;
|
|
else finishedWorkAt ??= Date.now();
|
|
if (!grade.passed && finishedWorkAt && Date.now() - finishedWorkAt >= 15_000)
|
|
return `Settled chat failed: ${grade.checks.filter(check => !check.passed).map(check => check.id).join(", ")}`;
|
|
return undefined;
|
|
},
|
|
timeoutDetail: grade => grade?.checks.filter(check => !check.passed).map(check => check.id).join(", "),
|
|
});
|
|
await saveEvidence();
|
|
const savedTask = observation.tasks[0]!;
|
|
const savedDocuments = observation.documents.map(document => ({ id: document.id, body: document.body, revision: document.latestRevisionId }));
|
|
await page.reload({ waitUntil: "domcontentloaded" });
|
|
await expect(page.getByTestId("task-chat-composer-input")).toBeVisible();
|
|
await pollUntil({ label: "chat result survives browser reload", deadlineAt: Date.now() + 30_000, load: refresh, accept: grade => grade.passed });
|
|
expect(observation.tasks[0]!.id).toBe(savedTask.id);
|
|
expect(observation.documents.map(document => ({ id: document.id, body: document.body, revision: document.latestRevisionId }))).toEqual(savedDocuments);
|
|
const targetLink = observation.renderedLinks.find(link => linksFeatureTask(link.href, savedTask))!;
|
|
const sourceUrl = page.url();
|
|
const reply = page.locator(`[id=${JSON.stringify(`comment-${targetLink.commentId}`)}]`);
|
|
await expect(reply).toBeVisible();
|
|
const linkIndex = await reply.locator("a[href]").evaluateAll((elements, target) => elements.findIndex(element =>
|
|
new URL(element.getAttribute("href")!, target.base).href === new URL(target.href, target.base).href),
|
|
{ base: api.baseURL, href: targetLink.href });
|
|
expect(linkIndex).toBeGreaterThanOrEqual(0);
|
|
await reply.locator("a[href]").nth(linkIndex).click();
|
|
await expect(page.getByTestId("issue-detail-header").getByRole("heading", { name: savedTask.title, exact: true })).toBeVisible();
|
|
expect((await api.get<FeatureTask>(`/api/issues/${encodeURIComponent(savedTask.identifier ?? savedTask.id)}`)).id).toBe(savedTask.id);
|
|
await page.goto(sourceUrl, { waitUntil: "domcontentloaded" });
|
|
await expect(page.getByTestId("task-chat-composer-input")).toBeVisible();
|
|
await input.capture("final-state", "Created task and retained source conversation", "final-state.png");
|
|
input.check?.("browser-result-access", true, "The exact linked task loaded in the browser and the source conversation survived reload.");
|
|
return { issue: observation.source!, runs: chatFeatureExecutionRuns(observation) };
|
|
} catch (error) {
|
|
failure = error;
|
|
throw error;
|
|
} finally {
|
|
try {
|
|
await refresh();
|
|
await saveEvidence();
|
|
const runEvidence = await Promise.all(observation.runs.map(run => collectChatRunEvidence(api, run)));
|
|
await input.evidence("chat-feature-runs.json", runEvidence);
|
|
await input.evidence("api-state.json", { issue: observation.source, runs: observation.runs,
|
|
comments: observation.comments, activity: observation.activity, runEvidence });
|
|
} catch (error) {
|
|
if (!failure) throw error;
|
|
await input.evidence("chat-feature-evidence-error.json", { incomplete: true,
|
|
message: error instanceof Error ? error.message : "Evidence collection failed" });
|
|
} finally {
|
|
try {
|
|
for (const check of gradeChatFeature(observation).checks) input.check?.(check.id, check.passed, check.detail);
|
|
} finally { await fixtures.teardown(); }
|
|
}
|
|
}
|
|
}
|