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>
125 lines
12 KiB
TypeScript
125 lines
12 KiB
TypeScript
import { mkdir } from "node:fs/promises";
|
|
import path from "node:path";
|
|
import { expect } from "@playwright/test";
|
|
import { RunnerApi } from "../api.js";
|
|
import { setupLiveFixtures, type LiveFixtureValues } from "../live-fixtures.js";
|
|
import { createAuthenticatedPerson } from "./identity-auth.js";
|
|
import { checks } from "./internal-support.js";
|
|
import type { IdentityBoundaryCase, IdentityBoundaryContext } from "./identity-boundary-contract.js";
|
|
/**
|
|
* Agent Chat is always on since #19360, so this boundary now covers the gate that still exists:
|
|
* the Conference Room experiment (enableConferenceRoomChat, default off, opt-in hidden from
|
|
* settings by #8237). A real signed-in person follows an old Conference Room link. The product
|
|
* redirects to the company dashboard rather than a broken route, does not list the Conference
|
|
* Room, still offers ordinary Chat with Morgan, refuses the board discussion API, and an
|
|
* ordinary agent cannot flip the instance setting. The opt-in is deliberately not offered in
|
|
* the UI today, so no visible operator control is required here.
|
|
*/
|
|
type Person = Awaited<ReturnType<typeof createAuthenticatedPerson>>;
|
|
const owned = new WeakMap<IdentityBoundaryContext, { person: Person; api: RunnerApi; companyId?: string; fixtures?: LiveFixtureValues; keyId?: string }>();
|
|
export interface ConferenceRoomOffObservation { companyId: string; companyPrefix: string; agentId: string; personId: string; canonicalIntent: string;
|
|
conferenceRoomEnabled: boolean; oldLinkLandedPath: string; conferenceComposerVisible: boolean; conferenceNavOffered: boolean;
|
|
chatAlternativePath: string; chatAlternativeAgentRefs: string[]; chatComposerVisible: boolean;
|
|
boardChatStatus: number; boardChatErrorCode: string; boardChatProbePersonId: string;
|
|
agentSettingsStatus: number; gateAfterProbe: boolean;
|
|
beforeTaskIds: string[]; afterTasks: Array<{ id: string; conversationAgentId: string | null }>; runIds: string[] }
|
|
/** Probe the board discussion API with the genuine person's same-origin session wrapper.
|
|
* Return only the closed refusal code, never the raw response or credential. */
|
|
export async function probeBoardChatAsPerson(person: Pick<Person, "id" | "request">, companyId: string, message: string) {
|
|
const response = await person.request("/api/board/chat/stream", "POST", { companyId, message });
|
|
const status = response.status;
|
|
const body = await response.json().catch(() => undefined) as { code?: unknown; error?: unknown } | undefined;
|
|
const code = body?.code === "FEATURE_DISABLED" ? "FEATURE_DISABLED" : body?.code === "DEPLOYMENT_MODE_UNSUPPORTED" ? "DEPLOYMENT_MODE_UNSUPPORTED"
|
|
: body?.error === "Board mutation requires trusted browser origin" ? "trusted-origin-required" : "unrecognized-response";
|
|
return { boardChatStatus: status, boardChatErrorCode: code, boardChatProbePersonId: person.id };
|
|
}
|
|
export function gradeConferenceRoomOff(value: unknown) {
|
|
const c = checks(); const o = value as ConferenceRoomOffObservation | undefined;
|
|
c.add("actual-conference-room-off-context", !!o && !!o.companyId && !!o.companyPrefix && !!o.agentId && !!o.personId && o.personId !== "local-board"
|
|
&& o.canonicalIntent === "Open the conference room so I can talk with Morgan." && o.conferenceRoomEnabled === false,
|
|
"The real signed-in human follows the Conference Room link with the instance experiment actually off; canonical intent remains unsent.");
|
|
c.add("old-link-lands-on-working-page", !!o && o.oldLinkLandedPath === `/${o.companyPrefix}/dashboard` && !o.conferenceComposerVisible && !o.conferenceNavOffered,
|
|
"The old Conference Room link opens this company's working dashboard, not a broken route or an unusable board composer, and navigation does not offer the Conference Room.");
|
|
c.add("supported-chat-alternative", !!o && o.chatComposerVisible && o.chatAlternativeAgentRefs.includes(o.chatAlternativePath.split("/").at(-1) ?? "")
|
|
&& o.chatAlternativePath.startsWith(`/${o.companyPrefix}/chats/`),
|
|
"Chat in the same navigation opens an ordinary, working chat with Morgan.");
|
|
c.add("board-discussion-refused", !!o && o.boardChatStatus === 403 && o.boardChatErrorCode === "FEATURE_DISABLED" && o.boardChatProbePersonId === o.personId,
|
|
"The same person's board discussion request is refused by the Conference Room gate itself, not by origin or another check.");
|
|
c.add("ordinary-agent-cannot-change-instance-gate", !!o && o.agentSettingsStatus === 403 && o.gateAfterProbe === false,
|
|
"The real ordinary agent cannot enable the Conference Room experiment.");
|
|
const before = new Set(o?.beforeTaskIds ?? []);
|
|
c.add("no-disguised-task-chat-or-run", !!o && Array.isArray(o.afterTasks) && o.beforeTaskIds.every(id => o.afterTasks.some(row => row.id === id))
|
|
&& o.afterTasks.filter(row => !before.has(row.id)).every(row => row.conversationAgentId === o.agentId)
|
|
&& Array.isArray(o.runIds) && !o.runIds.length,
|
|
"Nothing creates a task posing as the conference room or starts a run; the only new record allowed is Morgan's own ordinary chat.");
|
|
return c.result();
|
|
}
|
|
export const runtimeDisabledChatCases: readonly IdentityBoundaryCase[] = [{
|
|
scenarioId: "agent-chat.chat-disabled", journeyKind: "browser-boundary", boundaryTarget: "authenticated-harness", authentication: "human-session",
|
|
oracleVersion: "paperclip.chat-catalog.runtime-conference-off.v3", expectedRunCount: 0, expectedHumanMessageCount: 0, turnTimeoutMs: 120_000, attemptTimeoutMs: 300_000,
|
|
sourceFiles: ["chat-catalog-domains/runtime-chat-disabled.ts", "chat-catalog-domains/identity-auth.ts", "chat-catalog-domains/identity-private-browser.ts", "chat-catalog-domains/identity-private-browser-worker.ts", "chat-catalog-domains/identity-wizard.ts", "chat-catalog-domains/identity-boundary-contract.ts", "live-fixtures.ts", "chat-scenarios/agent-chat.yaml"],
|
|
grade: gradeConferenceRoomOff,
|
|
async run(ctx) {
|
|
const browser = ctx.page.context().browser(); if (!browser) throw new Error("Conference Room boundary requires the real human browser");
|
|
const person = await createAuthenticatedPerson({ api: ctx.api, browser, secrets: ctx.privateSecrets, name: "Alex" });
|
|
const api = new RunnerApi(person.context.request); person.bindApiSession(api); const resource: { person: Person; api: RunnerApi; companyId?: string; fixtures?: LiveFixtureValues; keyId?: string } = { person, api }; owned.set(ctx, resource);
|
|
await api.post("/api/bootstrap/claim");
|
|
const workspacePath = path.join(ctx.privateDir, "conference-off-owned-workspace"); await mkdir(workspacePath, { recursive: true, mode: 0o700 });
|
|
const fixtures = await setupLiveFixtures({ api, execution: ctx.execution, executionNonce: ctx.nonce, workspacePath, credentials: ctx.credentials,
|
|
async onCompanyCreated(id) { resource.companyId = id; ctx.observeResources({ companyId: id, runIds: [] }); } });
|
|
resource.fixtures = fixtures;
|
|
const prefix = fixtures.company.issuePrefix; if (!prefix) throw new Error("Conference Room boundary company has no route prefix");
|
|
await api.patch(`/api/agents/${fixtures.agent.id}`, { name: "Morgan" });
|
|
await api.patch("/api/instance/settings/experimental", { enableConferenceRoomChat: false });
|
|
const settings = await api.get<{ enableConferenceRoomChat: boolean }>("/api/instance/settings/experimental");
|
|
const morgan = await api.get<{ id: string; urlKey?: string | null }>(`/api/agents/${fixtures.agent.id}`);
|
|
const root = `/api/companies/${fixtures.company.id}`;
|
|
const beforeTaskIds = (await api.get<Array<{ id: string }>>(`${root}/issues?includeAgentCreatedTasks=true&limit=1000`)).map(row => row.id).sort();
|
|
// The boundary capture callback is bound to ctx.page; authenticate that page with the same person.
|
|
await person.installSessionOn(ctx.page.context());
|
|
person.bindApiSession(ctx.api);
|
|
const page = ctx.page;
|
|
await page.goto(`/${prefix}/board-chat`, { waitUntil: "domcontentloaded" });
|
|
await expect(page).toHaveURL(url => url.pathname === `/${prefix}/dashboard`);
|
|
const oldLinkLandedPath = new URL(page.url()).pathname;
|
|
const conferenceComposerVisible = await page.getByPlaceholder("Ask anything about your organization...").isVisible();
|
|
const chatLink = page.locator(`a[href="/${prefix}/chats"]`).first();
|
|
await expect(chatLink).toBeVisible();
|
|
const conferenceNavOffered = await page.locator(`a[href="/${prefix}/board-chat"]`).count() > 0;
|
|
ctx.observeResources({ companyId: fixtures.company.id, agentId: fixtures.agent.id, runIds: [] });
|
|
await ctx.capture("conference-room-off", "The old Conference Room link lands on the working company dashboard; canonical intent remains unsent", "runtime-conference-room-off.png");
|
|
// Supported alternative: the person's own Chat entry opens an ordinary chat with Morgan.
|
|
await chatLink.click();
|
|
const morganChoice = page.getByRole("button", { name: /Morgan/ });
|
|
await expect.poll(async () => /\/chats\/[^/]+$/.test(new URL(page.url()).pathname) || await morganChoice.isVisible(), { timeout: 30_000 }).toBe(true);
|
|
if (!/\/chats\/[^/]+$/.test(new URL(page.url()).pathname)) await morganChoice.click();
|
|
await expect(page).toHaveURL(url => /\/chats\/[^/]+$/.test(url.pathname));
|
|
await expect(page.getByTestId("task-chat-composer-input")).toBeVisible();
|
|
const chatAlternativePath = new URL(page.url()).pathname;
|
|
await ctx.captureFinalState(page);
|
|
const key = await api.postSensitive<{ id: string; token: string }>(`/api/agents/${fixtures.agent.id}/keys`, { name: "conference-off-permission-probe" });
|
|
ctx.privateSecrets.push(key.token); resource.keyId = key.id;
|
|
const denied = await fetch(new URL("/api/instance/settings/experimental", api.baseURL), { method: "PATCH", redirect: "error", signal: AbortSignal.timeout(15_000),
|
|
headers: { Authorization: `Bearer ${key.token}`, "content-type": "application/json" }, body: JSON.stringify({ enableConferenceRoomChat: true }) });
|
|
const agentSettingsStatus = denied.status; await denied.body?.cancel();
|
|
const boardProbe = await probeBoardChatAsPerson(person, fixtures.company.id, ctx.scenario.query);
|
|
const gateAfterProbe = (await api.get<{ enableConferenceRoomChat: boolean }>("/api/instance/settings/experimental")).enableConferenceRoomChat;
|
|
const tasks = await api.get<Array<{ id: string; conversationAgentId?: string | null }>>(`${root}/issues?includeAgentCreatedTasks=true&limit=1000`);
|
|
const runIds = (await api.get<Array<{ id: string }>>(`${root}/heartbeat-runs?limit=100`)).map(row => row.id);
|
|
ctx.observeResources({ companyId: fixtures.company.id, agentId: fixtures.agent.id, runIds });
|
|
const observation: ConferenceRoomOffObservation = { companyId: fixtures.company.id, companyPrefix: prefix, agentId: fixtures.agent.id, personId: person.id,
|
|
canonicalIntent: ctx.scenario.query, conferenceRoomEnabled: settings.enableConferenceRoomChat, oldLinkLandedPath, conferenceComposerVisible, conferenceNavOffered,
|
|
chatAlternativePath, chatAlternativeAgentRefs: [morgan.id, ...(morgan.urlKey ? [morgan.urlKey] : [])], chatComposerVisible: true,
|
|
...boardProbe, agentSettingsStatus, gateAfterProbe, beforeTaskIds,
|
|
afterTasks: tasks.map(row => ({ id: row.id, conversationAgentId: row.conversationAgentId ?? null })).sort((a, b) => a.id.localeCompare(b.id)), runIds };
|
|
await ctx.evidence("runtime-conference-room-off-boundary.json", observation); return observation;
|
|
},
|
|
async cleanup(ctx) {
|
|
const resource = owned.get(ctx); if (!resource) return;
|
|
try { if (resource.keyId && resource.fixtures) await resource.api.delete(`/api/agents/${resource.fixtures.agent.id}/keys/${resource.keyId}`, { allowNotFound: true });
|
|
await resource.fixtures?.teardown();
|
|
if (resource.companyId) await resource.api.delete(`/api/companies/${resource.companyId}`, { allowNotFound: true });
|
|
} finally { await resource.person.dispose(); owned.delete(ctx); }
|
|
},
|
|
}];
|