mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-13 07:12:00 +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>
162 lines
15 KiB
TypeScript
162 lines
15 KiB
TypeScript
import defaultInputs from "./source-sdk-public-inputs.json" with { type: "json" };
|
|
import { createHash } from "node:crypto";
|
|
import { createRequire } from "node:module";
|
|
import { lstat, readFile, readdir, realpath } from "node:fs/promises";
|
|
import path from "node:path";
|
|
import { findSecretLeakMatch, secretShapeObservations } from "./redaction.js";
|
|
|
|
const hash = (v: Buffer | string) => createHash("sha256").update(v).digest("hex");
|
|
export interface SdkPublicInputs {
|
|
inputs: Record<string, string>; lockSha256: string; compilerPackageSha256: string;
|
|
parserPackageSha256: string; parserEntrySha256: string; zodPackageSha256: string; zodEntrySha256: string; zodTypeFiles: Record<string, string>;
|
|
}
|
|
interface PublicSource {
|
|
root: string; parser: { parse(text: string, options: unknown): any }; constructors: Record<string, unknown>; factories: Record<string, unknown>;
|
|
inputs: SdkPublicInputs; identity: { inputs: number; manifestSha256: string; lockSha256: string; compilerPackageSha256: string; parserPackageSha256: string; parserEntrySha256: string; zodPackageSha256: string; zodEntrySha256: string; zodTypeGraphSha256: string };
|
|
}
|
|
const verified = new WeakSet<PublicSource>();
|
|
export async function verifySdkPublicSource(root: string, expectations?: SdkPublicInputs): Promise<PublicSource> {
|
|
const supplied = expectations ?? defaultInputs;
|
|
const inputs: SdkPublicInputs = Object.freeze({ ...supplied, inputs: Object.freeze({ ...supplied.inputs }), zodTypeFiles: Object.freeze({ ...supplied.zodTypeFiles }) });
|
|
if (await realpath(root) !== root) throw new Error("SDK public source proof unavailable");
|
|
for (const [file, expected] of Object.entries(inputs.inputs)) {
|
|
if (!/^[A-Za-z0-9._/-]+$/.test(file) || file.split("/").includes("..")) throw new Error("SDK public source proof unavailable");
|
|
const filename = path.join(root, file), stat = await lstat(filename);
|
|
const bytes = await readFile(filename);
|
|
// The authenticated restored lock is the sole overlay on frozen Git inputs.
|
|
if (!stat.isFile() || stat.isSymbolicLink() || await realpath(filename) !== filename
|
|
|| (file === "pnpm-lock.yaml" ? hash(bytes) !== inputs.lockSha256 : createHash("sha1").update(`blob ${bytes.length}\0`).update(bytes).digest("hex") !== expected)) throw new Error("SDK public source proof unavailable");
|
|
}
|
|
const actualSources: string[] = [];
|
|
const collect = async (directory: string) => {
|
|
for (const entry of await readdir(path.join(root, directory), { withFileTypes: true })) {
|
|
const file = `${directory}/${entry.name}`;
|
|
if (entry.isDirectory()) await collect(file);
|
|
else if (entry.isFile()) actualSources.push(file);
|
|
else throw new Error("SDK public source proof unavailable");
|
|
}
|
|
};
|
|
for (const directory of ["packages/shared/src", "packages/plugins/sdk/src"]) await collect(directory);
|
|
const expectedSources = Object.keys(inputs.inputs).filter(file => file.startsWith("packages/shared/src/") || file.startsWith("packages/plugins/sdk/src/"));
|
|
if (JSON.stringify(actualSources.sort()) !== JSON.stringify(expectedSources.sort())) throw new Error("SDK public source proof unavailable");
|
|
const matches = async (file: string, expected: string) => { if (hash(await readFile(file)) !== expected) throw new Error("SDK public source proof unavailable"); };
|
|
await matches(path.join(root, "pnpm-lock.yaml"), inputs.lockSha256);
|
|
await matches(path.join(root, "node_modules/typescript/package.json"), inputs.compilerPackageSha256);
|
|
const parserRoot = path.join(root, "node_modules/.pnpm/@babel+parser@7.29.8/node_modules/@babel/parser");
|
|
await matches(path.join(parserRoot, "package.json"), inputs.parserPackageSha256); await matches(path.join(parserRoot, "lib/index.js"), inputs.parserEntrySha256);
|
|
const require = createRequire(path.join(root, "packages/shared/package.json")), zodPackage = require.resolve("zod/package.json"), zodEntry = require.resolve("zod");
|
|
await matches(zodPackage, inputs.zodPackageSha256); await matches(zodEntry, inputs.zodEntrySha256);
|
|
const parser = createRequire(path.join(parserRoot, "package.json"))("./lib/index.js");
|
|
const publicTypes = new Map<string, any>();
|
|
for (const [relative, expected] of Object.entries(inputs.zodTypeFiles)) {
|
|
if (!/^[A-Za-z0-9._/-]+$/.test(relative) || relative.split("/").includes("..")) throw new Error("SDK public source proof unavailable");
|
|
const file = path.join(path.dirname(zodPackage), relative); await matches(file, expected);
|
|
const ast = parser.parse(await readFile(file, "utf8"), { sourceType: "module", plugins: [["typescript", { dts: true }]], errorRecovery: true });
|
|
if (ast.errors.length) throw new Error("SDK public source proof unavailable"); publicTypes.set(relative, ast);
|
|
}
|
|
const resolves = (parent: string, specifier: unknown, child: string) => typeof specifier === "string" && path.posix.normalize(path.posix.join(path.posix.dirname(parent), specifier)).replace(/\.(?:cjs|js)$/, ".d.cts") === child;
|
|
const entry = publicTypes.get("index.d.cts"), external = publicTypes.get("v4/classic/external.d.cts"), schemas = publicTypes.get("v4/classic/schemas.d.cts");
|
|
if (!entry || !external || !schemas || !entry.program.body.some((n: any) => n.type === "ExportAllDeclaration" && resolves("index.d.cts", n.source?.value, "v4/classic/external.d.cts"))
|
|
|| !external.program.body.some((n: any) => n.type === "ExportAllDeclaration" && resolves("v4/classic/external.d.cts", n.source?.value, "v4/classic/schemas.d.cts"))) throw new Error("SDK public source proof unavailable");
|
|
const namespaces = entry.program.body.filter((n: any) => n.type === "ImportDeclaration" && resolves("index.d.cts", n.source?.value, "v4/classic/external.d.cts")).flatMap((n: any) => n.specifiers.filter((s: any) => s.type === "ImportNamespaceSpecifier").map((s: any) => s.local.name));
|
|
if (!entry.program.body.some((n: any) => n.type === "ExportNamedDeclaration" && n.specifiers?.some((s: any) => s.exported?.name === "z" && namespaces.includes(s.local?.name)))) throw new Error("SDK public source proof unavailable");
|
|
const declaredTypes = new Set(schemas.program.body.filter((n: any) => n.type === "ExportNamedDeclaration" && ["TSInterfaceDeclaration", "TSTypeAliasDeclaration", "ClassDeclaration"].includes(n.declaration?.type)).map((n: any) => n.declaration.id.name));
|
|
const declaredFactories = new Set(schemas.program.body.filter((n: any) => n.type === "ExportNamedDeclaration" && ["TSDeclareFunction", "FunctionDeclaration"].includes(n.declaration?.type)).map((n: any) => n.declaration.id.name));
|
|
const zod = require("zod"), namespace = zod.z ?? zod;
|
|
const constructors = Object.freeze(Object.fromEntries(Object.entries(namespace).filter(([name, value]) => declaredTypes.has(name) && typeof value === "function")));
|
|
const factories = Object.freeze(Object.fromEntries(Object.entries(namespace).filter(([name, value]) => declaredFactories.has(name) && typeof value === "function")));
|
|
const source = { root, parser, constructors, factories, inputs, identity: { inputs: Object.keys(inputs.inputs).length, manifestSha256: hash(JSON.stringify(inputs)),
|
|
lockSha256: inputs.lockSha256, compilerPackageSha256: inputs.compilerPackageSha256, parserPackageSha256: inputs.parserPackageSha256,
|
|
parserEntrySha256: inputs.parserEntrySha256, zodPackageSha256: inputs.zodPackageSha256, zodEntrySha256: inputs.zodEntrySha256, zodTypeGraphSha256: hash(JSON.stringify(inputs.zodTypeFiles)) } };
|
|
Object.freeze(source.identity); Object.freeze(source); verified.add(source); return source;
|
|
}
|
|
function nodes(node: any, fn: (node: any) => void) {
|
|
if (!node || typeof node !== "object") return;
|
|
if (Array.isArray(node)) { for (const child of node) nodes(child, fn); return; }
|
|
if (typeof node.type !== "string") return;
|
|
fn(node);
|
|
for (const [key, value] of Object.entries(node)) if (!["loc", "extra", "errors", "tokens", "comments", "leadingComments", "trailingComments", "innerComments"].includes(key)) nodes(value, fn);
|
|
}
|
|
function aliases(ast: any): string[] {
|
|
const found: string[] = [];
|
|
for (const row of ast.program.body) if (row.type === "ImportDeclaration" && row.source.value === "zod") {
|
|
for (const s of row.specifiers) if (s.type === "ImportNamespaceSpecifier" || s.type === "ImportSpecifier" && s.imported.name === "z") found.push(s.local.name);
|
|
}
|
|
if (new Set(found).size !== found.length) return [];
|
|
let shadowed = false;
|
|
nodes(ast, n => {
|
|
if (n.type === "TSModuleDeclaration") shadowed = true;
|
|
if (["VariableDeclarator", "FunctionDeclaration", "ClassDeclaration", "TSModuleDeclaration", "TSTypeAliasDeclaration", "TSInterfaceDeclaration", "TSEnumDeclaration"].includes(n.type) && found.includes(n.id?.name)) shadowed = true;
|
|
if (n.type === "TSTypeParameter" && found.includes(typeof n.name === "string" ? n.name : n.name?.name)) shadowed = true;
|
|
if (Array.isArray(n.params) && n.params.some((p: any) => p.type === "Identifier" && found.includes(p.name))) shadowed = true;
|
|
});
|
|
return shadowed ? [] : found;
|
|
}
|
|
function parsed(source: PublicSource, text: string, declaration: boolean) {
|
|
const ast = source.parser.parse(text, { sourceType: "module", plugins: [["typescript", { dts: declaration }]], errorRecovery: true });
|
|
if (ast.errors.length) throw new Error("SDK public source parse unavailable"); return ast;
|
|
}
|
|
const primitive = (node: any) => node?.type === "NumericLiteral" && Number.isFinite(node.value) || ["BooleanLiteral", "NullLiteral"].includes(node?.type) || node?.type === "StringLiteral" && node.value === "";
|
|
function publicSourceProperty(source: PublicSource, ast: any, key: string) {
|
|
const publicAliases = aliases(ast), properties: any[] = [];
|
|
nodes(ast, n => { if (n.type === "ObjectProperty" && !n.computed && (n.key?.name ?? n.key?.value) === key) properties.push(n); });
|
|
if (properties.length !== 1) return false;
|
|
let value = properties[0].value, factory: string | undefined;
|
|
while (value?.type === "CallExpression" && value.callee?.type === "MemberExpression" && !value.callee.computed && value.callee.property?.type === "Identifier") {
|
|
if (!value.arguments.every(primitive)) return false;
|
|
factory = value.callee.property.name; value = value.callee.object;
|
|
}
|
|
return value?.type === "Identifier" && publicAliases.includes(value.name) && !!factory && Object.hasOwn(source.factories, factory) && typeof source.factories[factory] === "function";
|
|
}
|
|
export async function sdkPublicTypeSyntax(source: PublicSource, file: string, text: string, span: { start: number; end: number }) {
|
|
if (!verified.has(source) || Buffer.byteLength(text) > 8 * 1024 * 1024) return false;
|
|
const original = file.replace("/dist/", "/src/").replace(/\.d\.ts$/, ".ts");
|
|
if (original === file || !source.inputs.inputs[original]) return false;
|
|
const filename = path.join(source.root, original), sourceText = await readFile(filename, "utf8");
|
|
if (createHash("sha1").update(`blob ${Buffer.byteLength(sourceText)}\0`).update(sourceText).digest("hex") !== source.inputs.inputs[original]) return false;
|
|
try {
|
|
const ast = parsed(source, text, true), publicAliases = aliases(ast), candidates: any[] = [];
|
|
nodes(ast, n => { if (n.type === "TSPropertySignature" && n.start <= span.start && n.end >= span.end) candidates.push(n); });
|
|
candidates.sort((a, b) => (a.end - a.start) - (b.end - b.start)); const property = candidates[0], type = property?.typeAnnotation?.typeAnnotation;
|
|
if (!property || property.computed || property.key?.type !== "Identifier" || property.initializer || property.value || type?.type !== "TSTypeReference"
|
|
|| span.start < property.key.start || span.start >= property.key.end || type.typeName?.type !== "TSQualifiedName" || span.end !== type.typeName.end
|
|
|| !/^\s*:\s*$/.test(text.slice(property.key.end, type.typeName.start))
|
|
|| (ast.comments ?? []).some((c: any) => c.start < property.end && c.end > property.start)) return false;
|
|
let bound = true, references = 0;
|
|
nodes(property.typeAnnotation, n => {
|
|
if (!["TSTypeAnnotation", "TSTypeReference", "TSQualifiedName", "TSTypeParameterInstantiation", "Identifier"].includes(n.type)) bound = false;
|
|
if (n.type === "TSTypeReference") {
|
|
references++; const name = n.typeName;
|
|
if (name?.type !== "TSQualifiedName" || name.left?.type !== "Identifier" || name.right?.type !== "Identifier" || !publicAliases.includes(name.left.name)
|
|
|| !Object.hasOwn(source.constructors, name.right.name) || typeof source.constructors[name.right.name] !== "function") bound = false;
|
|
}
|
|
});
|
|
return bound && references > 0 && publicSourceProperty(source, parsed(source, sourceText, false), property.key.name);
|
|
} catch { return false; }
|
|
}
|
|
export interface SdkScanObservation { pathSha256: string; fileSha256: string; detector: number; start: number; end: number; classification: "private_exact" | "direct_token" | "unresolved_context" | "public_schema_type" }
|
|
export function createSdkArtifactScanner(repositoryRoot: string, secrets: readonly string[], expectations?: SdkPublicInputs) {
|
|
const observations: SdkScanObservation[] = []; let bounded = true; let proof: Promise<PublicSource | null> | undefined;
|
|
const source = () => proof ??= verifySdkPublicSource(repositoryRoot, expectations).catch(() => null);
|
|
return {
|
|
async scan(file: string, bytes: Buffer) {
|
|
const base = { pathSha256: hash(file), fileSha256: hash(bytes) };
|
|
if (findSecretLeakMatch(bytes, secrets, { includeShapes: false })) { observations.push({ ...base, detector: -1, start: 0, end: bytes.length, classification: "private_exact" }); return false; }
|
|
const raw = secretShapeObservations(bytes.toString("utf8")); if (!raw.bounded) { bounded = false; for (const span of raw.matches) observations.push({ ...base, ...span, classification: span.detector === 3 ? "unresolved_context" : "direct_token" }); return false; }
|
|
if (raw.matches.some(span => span.detector !== 3)) {
|
|
for (const span of raw.matches) observations.push({ ...base, ...span, classification: span.detector === 3 ? "unresolved_context" : "direct_token" });
|
|
return false;
|
|
}
|
|
let safe = true;
|
|
for (const span of raw.matches) {
|
|
const publicSource = span.detector === 3 ? await source() : null;
|
|
const syntax = !!publicSource && await sdkPublicTypeSyntax(publicSource, file, bytes.toString("utf8"), span);
|
|
const classification = span.detector !== 3 ? "direct_token" : syntax ? "public_schema_type" : "unresolved_context";
|
|
observations.push({ ...base, ...span, classification }); if (!syntax) safe = false;
|
|
}
|
|
return safe;
|
|
},
|
|
async audit() { const p = proof ? await proof : null; return { schema: "paperclip.source-sdk-scan/v1", suppressedFindings: 0, bounded, publicSource: p?.identity ?? null, observations: structuredClone(observations).sort((a, b) => a.pathSha256.localeCompare(b.pathSha256) || a.detector - b.detector || a.start - b.start) }; },
|
|
};
|
|
}
|