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>
160 lines
13 KiB
TypeScript
160 lines
13 KiB
TypeScript
import { createHash } from "node:crypto";
|
|
import { copyFile, mkdir, mkdtemp, readFile, realpath, rm, symlink, writeFile } from "node:fs/promises";
|
|
import path from "node:path";
|
|
import os from "node:os";
|
|
import { beforeAll, afterAll, expect, it } from "vitest";
|
|
import pins from "./source-sdk-public-inputs.json" with { type: "json" };
|
|
import { createSdkArtifactScanner, sdkPublicTypeSyntax, verifySdkPublicSource } from "./source-sdk-public-type.js";
|
|
import { findSecretLeakMatch, normalizedSecrets } from "./redaction.js";
|
|
import { diagnosticSecrets } from "./private-diagnostics.js";
|
|
const hash = (v: string | Buffer) => createHash("sha256").update(v).digest("hex");
|
|
const blob = (v: string | Buffer) => createHash("sha1").update(`blob ${Buffer.byteLength(v)}\0`).update(v).digest("hex");
|
|
const originalRoot = path.resolve(import.meta.dirname, "../.."), file = "packages/shared/dist/validators/synthetic-unit.d.ts", inputFile = file.replace("/dist/", "/src/").replace(/\.d\.ts$/, ".ts");
|
|
const source = 'import {z} from "zod";export const schema=z.object({daytona_api_key:z.string().min(1)});';
|
|
const declaration = 'import {z} from "zod";export declare const schema:{daytona_api_key:z.ZodOptional<z.ZodString>};';
|
|
let root: string; let fixtureLockSha256: string;
|
|
const fixtureInputs: Record<string, string> = {};
|
|
beforeAll(async () => {
|
|
root = await realpath(await mkdtemp(path.join(os.tmpdir(), "sdk-public-type-synthetic-")));
|
|
for (const [relative] of Object.entries(pins.inputs)) { const dest = path.join(root, relative); await mkdir(path.dirname(dest), { recursive: true }); await copyFile(path.join(originalRoot, relative), dest); fixtureInputs[relative] = blob(await readFile(dest)); }
|
|
await copyFile(path.join(originalRoot, "pnpm-lock.yaml"), path.join(root, "pnpm-lock.yaml")); fixtureLockSha256 = hash(await readFile(path.join(root, "pnpm-lock.yaml")));
|
|
await symlink(await realpath(path.join(originalRoot, "node_modules")), path.join(root, "node_modules"));
|
|
await symlink(await realpath(path.join(originalRoot, "packages/shared/node_modules")), path.join(root, "packages/shared/node_modules"));
|
|
});
|
|
afterAll(async () => { if (root) await rm(root, { recursive: true, force: true }); });
|
|
async function fixture(sourceText = source) {
|
|
await mkdir(path.dirname(path.join(root, inputFile)), { recursive: true }); await writeFile(path.join(root, inputFile), sourceText);
|
|
return { ...pins, lockSha256: fixtureLockSha256, inputs: { ...fixtureInputs, [inputFile]: blob(sourceText) } };
|
|
}
|
|
it("default public proof accepts the reviewed build lock and still rejects a changed lock", async () => {
|
|
const directory = await realpath(await mkdtemp(path.join(os.tmpdir(), "sdk-reviewed-runtime-")));
|
|
try {
|
|
for (const relative of Object.keys(pins.inputs)) {
|
|
const destination = path.join(directory, relative);
|
|
await mkdir(path.dirname(destination), { recursive: true });
|
|
await copyFile(path.join(originalRoot, relative), destination);
|
|
}
|
|
await symlink(await realpath(path.join(originalRoot, "node_modules")), path.join(directory, "node_modules"));
|
|
await symlink(await realpath(path.join(originalRoot, "packages/shared/node_modules")), path.join(directory, "packages/shared/node_modules"));
|
|
const reviewed = JSON.parse(await readFile(path.join(originalRoot, "tests/runner-e2e/report-lock.json"), "utf8"));
|
|
const lock = await readFile(path.join(directory, "pnpm-lock.yaml"));
|
|
expect(hash(lock)).toBe(reviewed.outputSha256);
|
|
// Use the real default registry, without injecting alternate expected pins.
|
|
const proof = await verifySdkPublicSource(directory);
|
|
expect(proof.identity.lockSha256).toBe(hash(lock));
|
|
await writeFile(path.join(directory, "pnpm-lock.yaml"), Buffer.concat([lock, Buffer.from("\n# changed owning runtime input\n")]));
|
|
await expect(verifySdkPublicSource(directory)).rejects.toThrow("SDK public source proof unavailable");
|
|
} finally { await rm(directory, { recursive: true, force: true }); }
|
|
});
|
|
it("classifies varied genuinely public schema types while retaining the raw observation", async () => {
|
|
for (const text of [declaration, declaration.replaceAll("z.", "schemaLibrary.").replace("{z}", "{z as schemaLibrary}"), declaration.replace("ZodString", "ZodNumber")]) {
|
|
const scanner = createSdkArtifactScanner(root, [], await fixture()); expect(await scanner.scan(file, Buffer.from(text))).toBe(true);
|
|
const audit = await scanner.audit(); expect(audit.suppressedFindings).toBe(0); expect(audit.observations).toHaveLength(1); expect(audit.observations[0]!.classification).toBe("public_schema_type");
|
|
}
|
|
});
|
|
it("classifies the shipped agent env schema declaration and keeps a literal at that position strict", async () => {
|
|
// Plugin cells retain packages/shared/dist/validators/agent.d.ts, whose
|
|
// `OPENROUTER_API_KEY: z.ZodOptional<z.ZodString>` matches the OpenRouter key
|
|
// shape. Stale default pins left it unresolved and blocked web-server startup.
|
|
const repository = await realpath(originalRoot), agentFile = "packages/shared/dist/validators/agent.d.ts";
|
|
const shipped = 'import { z } from "zod";\nexport declare const env: z.ZodObject<{\n OPENROUTER_API_KEY: z.ZodOptional<z.ZodString>;\n}>;\n';
|
|
const scanner = createSdkArtifactScanner(repository, []); expect(await scanner.scan(agentFile, Buffer.from(shipped))).toBe(true);
|
|
expect((await scanner.audit()).observations.map(row => row.classification)).toEqual(["public_schema_type"]);
|
|
for (const value of ['"syntheticCredentialValue123456789"', "syntheticCredentialValue123456789", 'z.ZodOptional<z.ZodString> | "syntheticCredential1"']) {
|
|
const strict = createSdkArtifactScanner(repository, []);
|
|
expect(await strict.scan(agentFile, Buffer.from(shipped.replace("z.ZodOptional<z.ZodString>", value)))).toBe(false);
|
|
expect((await strict.audit()).observations.map(row => row.classification)).toContain("unresolved_context");
|
|
}
|
|
});
|
|
it("protects an exact private value even when equal to a public schema symbol", async () => {
|
|
const scanner = createSdkArtifactScanner(root, ["ZodString"], await fixture()); expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false);
|
|
expect((await scanner.audit()).observations[0]!.classification).toBe("private_exact");
|
|
});
|
|
it("keeps direct token shapes unconditional inside declarations", async () => {
|
|
const scanner = createSdkArtifactScanner(root, [], await fixture()); expect(await scanner.scan(file, Buffer.from(declaration+'declare const privateValue:"sk-ABCDEFGHIJKLMNOPQRSTUVWX123456789";'))).toBe(false);
|
|
expect((await scanner.audit()).observations.some(row => row.classification === "direct_token")).toBe(true);
|
|
});
|
|
it.each([
|
|
declaration.replace("z.ZodOptional<z.ZodString>", '"syntheticCredentialValue123456789"'),
|
|
declaration.replace("z.ZodString", '"syntheticCredentialValue123456789"'),
|
|
declaration.replace('from "zod"', 'from "unverified-library"'),
|
|
declaration.replace("z.ZodString", "UnboundPrivateType"),
|
|
declaration.replace("z.ZodString", "typeof privateValue"),
|
|
declaration+'declare namespace z {}',
|
|
declaration.replace("<z.ZodString>", "</* synthetic context */z.ZodString>"),
|
|
'const daytona_api_key="syntheticCredentialValue123456789";',
|
|
'// daytona_api_key: syntheticCredentialValue123456789',
|
|
declaration.slice(0, -2),
|
|
])("keeps unknown/literal/initializer/comment/shadow/parse evidence strict", async text => {
|
|
const scanner = createSdkArtifactScanner(root, [], await fixture()); expect(await scanner.scan(file, Buffer.from(text))).toBe(false);
|
|
});
|
|
it("does not classify unproven files or a forged proof object", async () => {
|
|
const expectations = await fixture(), scanner = createSdkArtifactScanner(root, [], expectations);
|
|
expect(await scanner.scan("arbitrary/credential.d.ts", Buffer.from(declaration))).toBe(false);
|
|
expect(await sdkPublicTypeSyntax({} as never, file, declaration, { start: 0, end: 1 })).toBe(false);
|
|
});
|
|
it("preserves unknown credential roots/private enum-equal leaves and their encoded values", async () => {
|
|
const raw = JSON.stringify({ unknown: { kind: "identity" }, token: "synthetic-private-value-71983521" });
|
|
const scanner = createSdkArtifactScanner(root, normalizedSecrets([raw]), await fixture());
|
|
for (const text of [declaration+'type PublicKind="identity";', declaration+'type Value="synthetic-private-value-71983521";']) expect(await scanner.scan(file, Buffer.from(text))).toBe(false);
|
|
});
|
|
it.each(["lockSha256", "compilerPackageSha256", "parserPackageSha256", "parserEntrySha256", "zodPackageSha256", "zodEntrySha256"] as const)("rejects mismatched %s without classifying syntax", async field => {
|
|
const expectations = await fixture(), scanner = createSdkArtifactScanner(root, [], { ...expectations, [field]: "0".repeat(64) });
|
|
expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false); expect((await scanner.audit()).publicSource).toBeNull();
|
|
});
|
|
it("rejects changed source and unknown source initializers", async () => {
|
|
const expectations = await fixture(); await writeFile(path.join(root, inputFile), source+'\n');
|
|
const scanner = createSdkArtifactScanner(root, [], expectations); expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false);
|
|
const other = createSdkArtifactScanner(root, [], await fixture(source.replace("z.string().min(1)", "privateFactory()"))); expect(await other.scan(file, Buffer.from(declaration))).toBe(false);
|
|
});
|
|
it("keeps stale, missing and extra source graph inputs ineligible after pin synchronization", async () => {
|
|
const expectations = await fixture();
|
|
const missing = { ...expectations.inputs }; delete missing[inputFile];
|
|
for (const inputs of [{ ...expectations.inputs, [inputFile]: "0".repeat(40) }, missing]) {
|
|
const scanner = createSdkArtifactScanner(root, [], { ...expectations, inputs });
|
|
expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false);
|
|
expect((await scanner.audit()).publicSource).toBeNull();
|
|
}
|
|
const extra = path.join(root, "packages/shared/src/unpinned-synthetic.ts");
|
|
try {
|
|
await writeFile(extra, "export const synthetic = true;");
|
|
const scanner = createSdkArtifactScanner(root, [], expectations);
|
|
expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false);
|
|
expect((await scanner.audit()).publicSource).toBeNull();
|
|
} finally { await rm(extra, { force: true }); }
|
|
const scanner = createSdkArtifactScanner(root, [], expectations);
|
|
expect(await scanner.scan(file, Buffer.from(declaration))).toBe(true);
|
|
});
|
|
it("keeps shape overflow blocked with bounded retained observations", async () => {
|
|
const scanner = createSdkArtifactScanner(root, [], await fixture()); expect(await scanner.scan(file, Buffer.from((declaration+'\n').repeat(40)))).toBe(false);
|
|
const audit = await scanner.audit(); expect(audit.bounded).toBe(false); expect(audit.observations).toHaveLength(32);
|
|
});
|
|
|
|
it("requires the authenticated public type export graph", async () => {
|
|
const expected = await fixture();
|
|
const scanner = createSdkArtifactScanner(root, [], { ...expected, zodTypeFiles: { ...expected.zodTypeFiles, "v4/classic/schemas.d.cts": "0".repeat(64) } });
|
|
expect(await scanner.scan(file, Buffer.from(declaration))).toBe(false);
|
|
expect((await scanner.audit()).publicSource).toBeNull();
|
|
});
|
|
it("rejects non-exported source factories and declaration constructors independently", async () => {
|
|
const unknownFactory = createSdkArtifactScanner(root, [], await fixture(source.replace("z.string()", "z.syntheticUnknownFactory()")));
|
|
expect(await unknownFactory.scan(file, Buffer.from(declaration))).toBe(false);
|
|
const unknownType = createSdkArtifactScanner(root, [], await fixture());
|
|
expect(await unknownType.scan(file, Buffer.from(declaration.replace("ZodString", "syntheticUnknownType")))).toBe(false);
|
|
});
|
|
it("does not apply typed syntax classification to API, stream, JavaScript or ZIP-shaped content", async () => {
|
|
for (const untyped of ["api-state.json", "persisted-run.ndjson", "entry.js", "blob.zip"]) {
|
|
const scanner = createSdkArtifactScanner(root, [], await fixture());
|
|
expect(await scanner.scan(untyped, Buffer.from(declaration))).toBe(false);
|
|
}
|
|
expect(findSecretLeakMatch(declaration, [])).toBeTruthy();
|
|
});
|
|
it("keeps raw and encoded credential leaves protected during a synthetic scan roundtrip", async () => {
|
|
const roots = diagnosticSecrets([JSON.stringify({ private: { kind: "identity", value: "synthetic-private-value-71983521" } })]).values;
|
|
for (const value of ["identity", "synthetic-private-value-71983521", Buffer.from("synthetic-private-value-71983521").toString("base64"), encodeURIComponent("synthetic-private-value-71983521")]) {
|
|
const scanner = createSdkArtifactScanner(root, roots, await fixture());
|
|
const bytes = Buffer.from(declaration+'type SyntheticPrivateValue="'+value+'";');
|
|
expect(await scanner.scan(file, bytes)).toBe(false);
|
|
}
|
|
});
|