mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-13 15:55:46 +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>
113 lines
9.3 KiB
TypeScript
113 lines
9.3 KiB
TypeScript
import { createHash } from "node:crypto";
|
|
import { lstat, readFile, readdir, realpath } from "node:fs/promises";
|
|
import path from "node:path";
|
|
import pins from "./sdk-canary-build-pins.json" with { type: "json" };
|
|
import inputs from "./source-sdk-public-inputs.json" with { type: "json" };
|
|
import { verifySdkPublicSource } from "./source-sdk-public-type.js";
|
|
import { verifyRetainedSourceSdk } from "./source-sdk-evidence.js";
|
|
import { chatCatalogCaseFingerprint, normalizedChatCatalogExecution } from "./chat-catalog-fingerprint.js";
|
|
import { findChatCatalogCase } from "./chat-catalog-registry.js";
|
|
import { CHAT_OPENCODE_MODEL, CHAT_OPENCODE_PROFILE, reviewedChatProfile } from "./chat-runtime-profile.js";
|
|
import type { MatrixExecution } from "./types.js";
|
|
const hash = (value: string | Buffer) => createHash("sha256").update(value).digest("hex");
|
|
const hex = (value: unknown) => typeof value === "string" && /^[a-f0-9]{64}$/.test(value);
|
|
function fail(): never { throw new Error("SDK canary pre-fixture prerequisite is unqualified"); }
|
|
export const SDK_CANARY = "chat-features.runner-opencode.local.plugins.use-plugin-tool";
|
|
type Fingerprint = ReturnType<typeof chatCatalogCaseFingerprint>;
|
|
interface FileIdentity { file: string; size: number; sha256: string }
|
|
export interface SdkCanaryBuildObservation {
|
|
sourceRevision: string; inputFiles: number; gitInputFiles: number; frozenInputManifestSha256: string; copiedInputManifestSha256: string;
|
|
publicInputsFileSha256: string; helperSha256: string; lockSha256: string; compilerPackageSha256: string;
|
|
compiledFiles: number; compiledManifestSha256: string; retainedManifestSha256: string;
|
|
certificates: typeof pins.certificates;
|
|
}
|
|
export function assertSdkCanaryBuild(observed: SdkCanaryBuildObservation, expected = pins) {
|
|
if (!/^[a-f0-9]{40}$/.test(observed.sourceRevision)) fail();
|
|
for (const key of ["inputFiles", "gitInputFiles", "frozenInputManifestSha256", "copiedInputManifestSha256", "publicInputsFileSha256", "helperSha256", "lockSha256", "compilerPackageSha256", "compiledFiles", "compiledManifestSha256"] as const)
|
|
if (observed[key] !== expected[key]) fail();
|
|
if (observed.retainedManifestSha256 !== expected.compiledManifestSha256 || JSON.stringify(observed.certificates) !== JSON.stringify(expected.certificates)) fail();
|
|
}
|
|
export async function sdkCanaryOutputManifest(root: string) {
|
|
const files: FileIdentity[] = []; let total = 0;
|
|
const visit = async (relative: string) => {
|
|
const file = path.join(root, relative), info = await lstat(file);
|
|
if (info.isSymbolicLink() || await realpath(file) !== file) fail();
|
|
if (info.isDirectory()) { for (const child of await readdir(file)) await visit(`${relative}/${child}`); }
|
|
else {
|
|
if (!info.isFile() || info.size > 8 * 1024 * 1024 || files.length >= 4096 || (total += info.size) > 64 * 1024 * 1024) fail();
|
|
const bytes = await readFile(file); files.push({ file: relative, size: bytes.length, sha256: hash(bytes) });
|
|
}
|
|
};
|
|
for (const certificate of pins.certificates) await visit(certificate.target);
|
|
return files.sort((a, b) => a.file.localeCompare(b.file));
|
|
}
|
|
export async function captureSdkCanaryBuild(input: { repositoryRoot: string; temporaryRoot: string; sourceRevision: string | null; secrets: readonly string[] }) {
|
|
if (!input.sourceRevision || await realpath(input.repositoryRoot) !== input.repositoryRoot) fail();
|
|
const source = await verifySdkPublicSource(input.repositoryRoot);
|
|
const publicInputsFileSha256 = hash(await readFile(path.join(input.repositoryRoot, "tests/runner-e2e/source-sdk-public-inputs.json")));
|
|
const copied = [];
|
|
for (const [file, gitBlobSha1] of Object.entries(inputs.inputs)) {
|
|
const bytes = await readFile(path.join(input.repositoryRoot, file));
|
|
copied.push({ path: file, bytes: bytes.length, sha256: hash(bytes), gitBlobSha1, restoredLockOverlay: file === "pnpm-lock.yaml" });
|
|
}
|
|
copied.sort((a, b) => a.path.localeCompare(b.path));
|
|
const retained = await verifyRetainedSourceSdk(input.temporaryRoot, SDK_CANARY, input.sourceRevision, input.secrets, input.repositoryRoot);
|
|
if (retained.receipt.schema !== "paperclip.source-plugin-sdk/v2" || retained.receipt.scanAudit?.bounded !== true || retained.receipt.scanAudit.suppressedFindings !== 0
|
|
|| !Array.isArray(retained.receipt.certifiedTargets) || retained.receipt.certifiedTargets.length !== pins.certificates.length) fail();
|
|
const files = await sdkCanaryOutputManifest(input.repositoryRoot);
|
|
const retainedFiles = retained.files.filter(row => row.file !== "receipt.json").map(({ file, size, sha256 }) => ({ file, size, sha256 })).sort((a, b) => a.file.localeCompare(b.file));
|
|
const certificates = [];
|
|
for (const expected of pins.certificates) {
|
|
const marker = await readFile(path.join(input.repositoryRoot, expected.target, ".paperclip-build-complete"));
|
|
const certificate = JSON.parse(marker.toString("utf8"));
|
|
const saved = retained.receipt.certifiedTargets.find(row => row.directory === expected.target);
|
|
if (saved?.sources !== certificate.sources || saved?.outputs !== certificate.outputs) fail();
|
|
certificates.push({ target: expected.target, sources: certificate.sources, outputs: certificate.outputs, markerSha256: hash(marker) });
|
|
}
|
|
const observed: SdkCanaryBuildObservation = { sourceRevision: input.sourceRevision, inputFiles: copied.length, gitInputFiles: copied.length - 1,
|
|
frozenInputManifestSha256: hash(JSON.stringify({ frozenRevision: "cb31ebccf0411f3e260159b1d39e6c047a39f4e9", inputs: inputs.inputs, schema: "paperclip.sdk-diagnostic-frozen-inputs/v1" })),
|
|
copiedInputManifestSha256: hash(JSON.stringify(copied)), publicInputsFileSha256,
|
|
helperSha256: hash(await readFile(path.join(input.repositoryRoot, "scripts/ensure-plugin-build-deps.mjs"))),
|
|
lockSha256: source.identity.lockSha256, compilerPackageSha256: source.identity.compilerPackageSha256,
|
|
compiledFiles: files.length, compiledManifestSha256: hash(JSON.stringify(files)), retainedManifestSha256: hash(JSON.stringify(retainedFiles)), certificates };
|
|
if (retained.receipt.buildHelperSha256 !== observed.helperSha256 || retained.receipt.lockSha256 !== observed.lockSha256) fail();
|
|
assertSdkCanaryBuild(observed); return observed;
|
|
}
|
|
function completeFingerprint(fingerprint: Fingerprint) {
|
|
if (!fingerprint || fingerprint.fingerprintKind !== "complete-execution-and-oracle-closure" || !hex(fingerprint.digest)
|
|
|| fingerprint.definitionDigest !== fingerprint.digest || fingerprint.graderDigest !== fingerprint.digest || !Array.isArray(fingerprint.sources) || !fingerprint.sources.length
|
|
|| fingerprint.sources.some(row => !row.file || !hex(row.sha256)) || new Set(fingerprint.sources.map(row => row.file)).size !== fingerprint.sources.length
|
|
|| !fingerprint.sources.some(row => row.file === "packages/plugins/sdk/dist/index.js")) fail();
|
|
}
|
|
const entered = new WeakSet<MatrixExecution>();
|
|
const fingerprints = new WeakMap<MatrixExecution, Fingerprint>();
|
|
export async function withSdkCanaryPreFixture<T>(input: { execution: MatrixExecution; sourceRevision: string | null;
|
|
capture(): Promise<{ build: SdkCanaryBuildObservation; fingerprint: Fingerprint }>;
|
|
retain(proof: unknown): Promise<void>; fixtures(): Promise<T>;
|
|
}) {
|
|
if (input.execution.id !== SDK_CANARY) return input.fixtures();
|
|
if (entered.has(input.execution)) fail(); entered.add(input.execution);
|
|
if (!reviewedChatProfile(input.execution.profile) || input.execution.profile.id !== CHAT_OPENCODE_PROFILE || input.execution.profile.provider !== "opencode"
|
|
|| input.execution.profile.model !== CHAT_OPENCODE_MODEL || input.execution.profile.generation !== "native" || input.execution.suite.id !== "chat-features"
|
|
|| input.execution.environment.id !== "local" || input.execution.task.flow !== "agent_chat" || input.execution.task.turnTimeoutMs !== 0
|
|
|| input.execution.task.id !== "plugins.use-plugin-tool" || input.execution.task.expectedRunCount !== 2 || input.execution.task.attemptTimeoutMs.local !== 480_000) fail();
|
|
const captured = await input.capture(); assertSdkCanaryBuild(captured.build); completeFingerprint(captured.fingerprint);
|
|
if (captured.build.sourceRevision !== input.sourceRevision) fail();
|
|
await input.retain({ schema: "paperclip.sdk-canary-prefixture/v1", executionId: SDK_CANARY, sourceRevision: input.sourceRevision,
|
|
authenticatedMetadataSha256: pins.authenticatedMetadataSha256, build: captured.build, fingerprint: captured.fingerprint });
|
|
fingerprints.set(input.execution, structuredClone(captured.fingerprint));
|
|
return input.fixtures();
|
|
}
|
|
export async function captureSdkCanaryPreFixture(input: { execution: MatrixExecution; repositoryRoot: string; temporaryRoot: string; sourceRevision: string | null; secrets: readonly string[] }) {
|
|
const build = await captureSdkCanaryBuild(input);
|
|
const descriptor = findChatCatalogCase(input.execution.task.id);
|
|
if (!descriptor || descriptor.expectedHumanMessageCount !== 2 || descriptor.expectedRunCount !== 2 || descriptor.attemptTimeoutMs !== 480_000) fail();
|
|
return { build, fingerprint: chatCatalogCaseFingerprint(descriptor, normalizedChatCatalogExecution(input.execution)) };
|
|
}
|
|
export function assertSdkCanaryFingerprintMatches(execution: MatrixExecution, fingerprint: Fingerprint) {
|
|
if (execution.id !== SDK_CANARY) return;
|
|
completeFingerprint(fingerprint);
|
|
const before = fingerprints.get(execution);
|
|
if (!before || JSON.stringify(before) !== JSON.stringify(fingerprint)) fail();
|
|
}
|