mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-14 08:53:19 +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>
96 lines
11 KiB
TypeScript
96 lines
11 KiB
TypeScript
import { expect } from "@playwright/test";
|
|
import { uploadFactualQueryAttachment, verifyChatQueryAttachments, type ChatQueryAttachmentReceipt, type ChatQueryHumanComment } from "../chat-query-attachments.js";
|
|
import type { ChatCatalogContext } from "../chat-catalog-contract.js";
|
|
import type { ToolConnection } from "../../../packages/shared/src/types/tool-access.js";
|
|
import { collectRunEvents } from "../run-observations.js";
|
|
import { privateScanJson, bytesContainSecret, scanReachableSecretOutputs } from "./external-secret-scan.js";
|
|
import { startNotesProvider, type NotesProvider } from "./external-provider.js";
|
|
import { definition, externalObservation, observeExternal, root, checks, verifyRenderedReply } from "./external-support.js";
|
|
import { waitForReply } from "./identity-support.js";
|
|
import { privateBrowserEntry } from "./external-private-browser.js";
|
|
const fixtures = new WeakMap<ChatCatalogContext, NotesProvider>();
|
|
/** Nonsecret account selection supplied by the real fixture human. No route,
|
|
* credential, tool invocation or expected setup result is provided to the model. */
|
|
export function secureProviderFacts(providerURL: string) {
|
|
const provider = new URL(providerURL);
|
|
if (provider.protocol !== "http:" || provider.hostname !== "127.0.0.1" || !provider.port || provider.pathname !== "/" || provider.search || provider.hash || provider.username || provider.password)
|
|
throw new Error("Account context must identify only the actual owned provider origin");
|
|
return `# Selected account\n\nAccount: Synthetic notes service (controlled test provider).\nAuthorized server origin: ${provider.origin}\n`;
|
|
}
|
|
export function gradeSecureKey(value: unknown) {
|
|
const c = checks(), o = externalObservation(value);
|
|
c.add("complete-secure-key-evidence", !!o, "Actual secure form submission, saved secret references, provider health and private output/draft scans are required."); if (!o) return c.result();
|
|
c.add("actual-human-account-context", o.state.factualContextVerified === true,
|
|
"The selected nonsecret account is delivered as an actual human-uploaded canonical attachment with independently verified source and bytes.");
|
|
c.add("actual-secure-credential-control", o.state.challengeVisible === true && o.state.passwordInput === true && o.state.autocompleteOff === true && o.state.humanSubmittedSecurely === true
|
|
&& o.state.secureBrowserUntraced === true && o.state.visibleReady === true, "A separate untraced human browser enters the credential only into the product's password control and sees actual completed setup.");
|
|
const connection = o.state.connectionAfter as { companyId: string; createdByUserId: string | null; providerURL: unknown; id: string; status: string; enabled: boolean; healthStatus: string; secretIds: string[]; serializedValueLeaked: boolean } | undefined;
|
|
c.add("stored-reference-and-verified-connectivity", !!connection && connection.companyId === o.companyId && connection.createdByUserId === o.state.personId && connection.providerURL === o.state.serverUrl && connection.id === o.state.connectionId && connection.status === "active" && connection.enabled === true
|
|
&& connection.healthStatus === "ok" && connection.secretIds.length >= 1 && connection.serializedValueLeaked === false && Number(o.state.providerAuthorizedCatalogReads) > 0
|
|
&& o.connections.length === 1 && o.state.secureSecretsExist === true,
|
|
"Exactly one active connection stores real credential references and a successful authenticated provider catalog check; no value is copied into public connection configuration.");
|
|
c.add("no-key-in-history-or-drafts", o.state.storageClean === true && o.state.secretScanComplete === true && o.state.secretLeaked === false,
|
|
"Private scans check actual browser storage, persisted comments, documents/revisions, attachments/artifacts, run events and run results for the actual credential bytes.");
|
|
c.add("supported-secure-entry-handoff", o.links.some(href => { try { const url = new URL(href, o.origin); return url.origin === o.origin && url.pathname === o.state.setupPath; } catch { return false; } }),
|
|
"The model actually links the supported secure setup entry rather than merely mentioning that credentials exist.");
|
|
// Reply wording is not graded: the required secure-entry link is the handoff, and the private scans
|
|
// prove the credential never reached chat, documents, attachments or run output.
|
|
return c.result();
|
|
}
|
|
export const secureKeyExternalCases = [definition("connection-setup.secure-api-key", async ctx => {
|
|
if (!ctx.input.privateSecrets) throw new Error("Secure setup requires private credential redaction ownership");
|
|
const provider = await startNotesProvider("Synthetic key-authenticated fixture"); fixtures.set(ctx, provider); ctx.input.privateSecrets.push(provider.token);
|
|
try {
|
|
await ctx.input.api.post(`${root(ctx)}/chats/${ctx.input.fixtures.agent.id}`); await ctx.refreshIssue();
|
|
const factualContextAttachment = await uploadFactualQueryAttachment(ctx, {
|
|
filename: "provider-account.md", body: secureProviderFacts(provider.url),
|
|
});
|
|
return { serverUrl: provider.url, personId: factualContextAttachment.createdByUserId, factualContextAttachment, setupPath: `/${ctx.input.fixtures.company.issuePrefix}/apps/byo` };
|
|
} catch (error) { fixtures.delete(ctx); await provider.close(); throw error; }
|
|
}, gradeSecureKey, {
|
|
oracleVersion: "paperclip.chat-catalog.external.v9", authentication: "human-session",
|
|
queryAttachments: [{ stateKey: "factualContextAttachment", filename: "provider-account.md" }],
|
|
attemptTimeoutMs: 420_000,
|
|
sourceFiles: ["chat-catalog-domains/external-secure-key.ts", "chat-query-attachments.ts", "../../ui/src/App.tsx", "../../ui/src/pages/apps/app-connect-policy.ts", "../../ui/src/features/connections/ConnectionSetupFlow.tsx", "chat-catalog-domains/external-private-browser.ts", "chat-catalog-domains/external-private-browser-worker.ts", "chat-catalog-domains/external-provider.ts", "chat-catalog-domains/external-secret-scan.ts", "run-observations.ts", "chat-catalog-domains/external-support.ts", "chat-catalog-domains/internal-support.ts", "chat-catalog-domains/identity-support.ts", "chat-scenarios/connection-setup.yaml"],
|
|
async act(ctx, state) {
|
|
await waitForReply(ctx); const provider = fixtures.get(ctx); if (!provider) throw new Error("Secure provider fixture missing");
|
|
// The saved reply can arrive before it renders; wait for the rendered link instead of reading anchors once.
|
|
const renderedTarget = async () => (await ctx.input.page.locator(`[id^="comment-"] a`).evaluateAll(anchors => anchors.map(anchor => (anchor as HTMLAnchorElement).href)))
|
|
.find(href => { try { const url = new URL(href); return url.origin === new URL(ctx.input.api.baseURL).origin && url.pathname === state.setupPath; } catch { return false; } });
|
|
let target: string | undefined;
|
|
await expect.poll(async () => (target = await renderedTarget()) !== undefined, { timeout: 30_000 }).toBe(true).catch(() => undefined);
|
|
if (!target) throw new Error("The model did not link the intended secure setup entry");
|
|
const receipt = await privateBrowserEntry({ mode: "secure-key", baseURL: ctx.input.api.baseURL, setupURL: target, providerURL: provider.url,
|
|
credential: provider.token, cookies: await ctx.input.page.context().cookies(ctx.input.api.baseURL) });
|
|
Object.assign(state, receipt, { secureBrowserUntraced: receipt.processIsolation && !receipt.testRunnerAttached && !receipt.tracingEnabled && !receipt.videoEnabled && !receipt.apiStepRecorder });
|
|
await ctx.input.evidence("chat-catalog-secure-entry-boundary.json", { processIsolation: receipt.processIsolation, testRunnerAttached: receipt.testRunnerAttached,
|
|
tracingEnabled: receipt.tracingEnabled, videoEnabled: receipt.videoEnabled, apiStepRecorder: receipt.apiStepRecorder, connectionId: receipt.connectionId });
|
|
},
|
|
async observe(ctx, state) {
|
|
const provider = fixtures.get(ctx); if (!provider) throw new Error("Secure fixture missing");
|
|
const connection = await ctx.input.api.get<ToolConnection>(`/api/tool-connections/${state.connectionId}`);
|
|
const sourceIssue = ctx.issue();
|
|
if (!sourceIssue) throw new Error("Secure account context has no canonical source");
|
|
const comments = await ctx.input.api.get<ChatQueryHumanComment[]>(`/api/issues/${sourceIssue.id}/comments?order=asc`);
|
|
const human = comments.find(row => row.authorUserId && !row.authorAgentId);
|
|
state.factualContextVerified = !!human && await verifyChatQueryAttachments(ctx, [state.factualContextAttachment as ChatQueryAttachmentReceipt], human);
|
|
const secretIds = [...new Set(connection.credentialSecretRefs.map(ref => ref.secretId))];
|
|
state.connectionAfter = { companyId: connection.companyId, createdByUserId: connection.createdByUserId, providerURL: connection.config?.url, id: connection.id, status: connection.status, enabled: connection.enabled, healthStatus: connection.healthStatus, secretIds, serializedValueLeaked: JSON.stringify(connection).includes(provider.token) };
|
|
state.secretIds = secretIds; const o = await observeExternal(ctx, state); state.secureSecretsExist = secretIds.length > 0 && secretIds.every(id => o.secrets.some(secret => secret.id === id));
|
|
state.providerAuthorizedCatalogReads = provider.calls.filter(call => call.authorized && call.method === "tools/list").length;
|
|
const events: unknown[] = []; for (const run of o.runs) events.push(...await collectRunEvents(afterSeq => privateScanJson(ctx, `/api/heartbeat-runs/${run.id}/events?afterSeq=${afterSeq}&limit=1000`)));
|
|
const source = ctx.issue(); if (!source) throw new Error("Secure output scan needs the actual source task"); const reachable = new Set([source.id]);
|
|
for (;;) { const count = reachable.size; for (const task of o.tasks) if (task.parentId && reachable.has(task.parentId)) reachable.add(task.id); if (count === reachable.size) break; }
|
|
const scan = await scanReachableSecretOutputs(ctx, [...reachable], provider.token);
|
|
state.secretScanComplete = scan.complete; state.secretLeaked = scan.leaked || bytesContainSecret(Buffer.from(JSON.stringify([events, o.runs])), provider.token);
|
|
return { ...o, state };
|
|
},
|
|
async verifyResult(ctx, _state, value) { await verifyRenderedReply(ctx, value); },
|
|
async cleanup(ctx, state) {
|
|
const provider = fixtures.get(ctx); fixtures.delete(ctx);
|
|
try { if (typeof state.connectionId === "string") await ctx.input.api.delete(`/api/tool-connections/${state.connectionId}`, { allowNotFound: true });
|
|
for (const id of Array.isArray(state.secretIds) ? state.secretIds : []) if (typeof id === "string") await ctx.input.api.delete(`/api/secrets/${id}`, { allowNotFound: true });
|
|
} finally { await provider?.close(); }
|
|
},
|
|
})];
|