mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-15 10:21:41 +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>
137 lines
13 KiB
TypeScript
137 lines
13 KiB
TypeScript
import { describe, expect, it, vi } from "vitest";
|
|
import { SkillSourceUpdateReadError } from "../api.js";
|
|
import { readFileSync } from "node:fs";
|
|
import { accessibilityUpdatePins as pins, gradeIntegratorSkillUpdate, requireSkillUpdatePrerequisite, integratorSkillUpdateCases, acceptedSkillReviewRun, collectVisibleSkillReviewFallback } from "./integrator-skill-update.js";
|
|
// Unmodified MIT-licensed source bytes, with LICENSE.md in the same fixture
|
|
// directory. Product setup still fetches the actual GitHub commit snapshots.
|
|
import { createHash } from "node:crypto";
|
|
import { saveSkillConversationContext } from "./integrator-skill-context.js";
|
|
vi.mock("./integrator-skill-context.js", () => ({ saveSkillConversationContext: vi.fn(async () => ({})) }));
|
|
vi.mock("./integrator-skill-source.js", () => ({ editableLocalNotes: async () => ({ id: "local", currentVersionId: "local-version" }) }));
|
|
vi.mock("./integrator-skill-versions.js", () => ({ readSkillVersionHistory: async () => [] }));
|
|
|
|
const bytes = { old: readFileSync(new URL("../fixtures/accessibility-source/0.4.SKILL.md", import.meta.url), "utf8"), next: readFileSync(new URL("../fixtures/accessibility-source/0.5.SKILL.md", import.meta.url), "utf8") };
|
|
function valid(old: string, next: string) {
|
|
const upstream = { id: "up", companyId: "company", name: "accessibility", editable: false, sourceRef: pins.oldCommit, currentVersionId: "old", markdown: old };
|
|
const local = { id: "fork", editable: true, forkedFromSkillId: "up", currentVersionId: "local", markdown: `${old}\nORCHID customer notes` };
|
|
const versions = [{ id: "old", fileInventory: [{ path: "SKILL.md", content: old }] }]; const localVersions = [{ id: "local" }];
|
|
const base = { scenarioId: "skills.review-update", origin: "https://fixture.test", companyId: "company", leadId: "ceo", state: { upstream: structuredClone(upstream), localFork: structuredClone(local), localVersions: structuredClone(localVersions) }, upstream, localFork: local, upstreamVersions: versions, forkVersions: localVersions,
|
|
source: { id: "source", lastScanCommit: pins.oldCommit, trackingRef: pins.newCommit, automaticUpdates: false }, updateStatus: { supported: true, hasUpdate: true, currentRef: pins.oldCommit, latestRef: pins.newCommit },
|
|
tasks: [], projects: [], goals: [], documents: [], revisions: [], comments: [], humanQueries: [], links: [], runs: [], activity: [],
|
|
reply: "The currently installed version is 0.4. The new revision adds continuous accessibility regression tests and splits long forms into smaller steps; it also discourages crowded interface patterns. I have not applied it." };
|
|
return { ...base, upstream: { ...upstream, sourceRef: pins.newCommit, currentVersionId: "new", markdown: next }, source: { ...base.source, lastScanCommit: pins.newCommit },
|
|
upstreamVersions: [...versions, { id: "new", fileInventory: [{ path: "SKILL.md", content: next }] }], activity: [{ agentId: "ceo", action: "company.skill_source_refreshed", entityId: "source" }], state: { ...base.state, reviewCheckpoint: structuredClone(base) },
|
|
reply: `${base.reply}\nI updated the accessibility skill to the reviewed source version. The local copy is unchanged.` };
|
|
}
|
|
describe("genuine immutable public skill update qualification", () => {
|
|
it("retains two independently qualified distinct public hashes without credential prerequisites", () => {
|
|
expect(pins.oldCommit).not.toBe(pins.newCommit); expect(pins.oldSha256).not.toBe(pins.newSha256);
|
|
expect(pins.repositoryURL).toBe("https://github.com/KreerC/ACCESSIBILITY.md");
|
|
});
|
|
it("accepts substantive review, normal saved version update and preserved fork using actual source bytes", () => {
|
|
expect(createHash("sha256").update(bytes.old).digest("hex")).toBe(pins.oldSha256); expect(createHash("sha256").update(bytes.next).digest("hex")).toBe(pins.newSha256);
|
|
expect(gradeIntegratorSkillUpdate(valid(bytes.old, bytes.next)).passed).toBe(true); expect(gradeIntegratorSkillUpdate(undefined).passed).toBe(false);
|
|
const natural = valid(bytes.old, bytes.next); natural.reply = `${natural.state.reviewCheckpoint.reply}\nThe accessibility skill is now on revision 0.5; our local fork is preserved.`; expect(gradeIntegratorSkillUpdate(natural).passed).toBe(true);
|
|
});
|
|
it("accepts a truthful not-applied review but rejects a later positive claim or changed saved version", () => {
|
|
const truthful = valid(bytes.old, bytes.next);
|
|
truthful.state.reviewCheckpoint.reply = truthful.state.reviewCheckpoint.reply.replace("I have not applied it.", "No update applied yet.");
|
|
truthful.reply = `${truthful.state.reviewCheckpoint.reply}\nI updated the accessibility skill to the reviewed source version. The local copy is unchanged.`;
|
|
expect(gradeIntegratorSkillUpdate(truthful).passed).toBe(true);
|
|
const contradictory = structuredClone(truthful);
|
|
contradictory.state.reviewCheckpoint.reply += " The update is applied.";
|
|
expect(gradeIntegratorSkillUpdate(contradictory).checks.find(check => check.id === "review-diff-before-update")?.passed).toBe(false);
|
|
const changed = structuredClone(truthful);
|
|
changed.state.reviewCheckpoint.upstream.currentVersionId = "premature";
|
|
expect(gradeIntegratorSkillUpdate(changed).checks.find(check => check.id === "review-diff-before-update")?.passed).toBe(false);
|
|
});
|
|
|
|
it("keeps a complete persisted preview on the original path without any DOM read", async () => {
|
|
const first = valid(bytes.old, bytes.next).state.reviewCheckpoint;
|
|
const evaluate = vi.fn(async () => { throw new Error("No additional native bubble exists"); });
|
|
const ctx = { input: { page: { evaluate } } } as unknown as Parameters<typeof collectVisibleSkillReviewFallback>[0];
|
|
expect(await collectVisibleSkillReviewFallback(ctx, first, "source")).toBeUndefined();
|
|
expect(evaluate).not.toHaveBeenCalled();
|
|
expect(gradeIntegratorSkillUpdate(valid(bytes.old, bytes.next)).passed).toBe(true);
|
|
});
|
|
|
|
it("admits only a visible committed current-run review alongside the shorter final comment", () => {
|
|
const value = valid(bytes.old, bytes.next);
|
|
const first = value.state.reviewCheckpoint as typeof value.state.reviewCheckpoint & { visibleReview?: unknown };
|
|
Object.assign(first.state, { sourceIssueId: "source" });
|
|
const summary = first.reply;
|
|
const run = { id: "review-run", runtimeMode: "native", companyId: "company", agentId: "ceo", status: "succeeded", nativeIssueId: "source",
|
|
contextSnapshot: { issueId: "source", commentId: "human" }, resultJson: { summary, finalizationPhase: "committed",
|
|
finalizationReasonCode: "conversation_turn_finished", prpRunTerminalState: "succeeded", prpTurnTerminalState: "completed",
|
|
nativeResult: { schema: "paperclip.run_result.v1", reportedWorkDisposition: "done", summary } } };
|
|
Object.assign(first, { runs: [run], reply: "The complete diff is shown above. No update applied yet." });
|
|
value.reply = `${first.reply}\nI updated the accessibility skill to the reviewed source version. The local copy is unchanged.`;
|
|
const check = () => gradeIntegratorSkillUpdate(value).checks.find(row => row.id === "review-diff-before-update")?.passed;
|
|
expect(check()).toBe(false); // An API-only summary never substitutes for visible review.
|
|
first.visibleReview = { runId: run.id, issueId: "source", humanCommentId: "human", summary, visibleText: summary };
|
|
expect(check()).toBe(true);
|
|
expect(gradeIntegratorSkillUpdate(value).passed).toBe(true); // Raw final-comment prefix still proves later apply.
|
|
for (const patch of [{ runId: "old" }, { issueId: "foreign" }, { humanCommentId: "old-human" }, { summary: "foreign" }, { visibleText: "" }, { visibleText: "Installed 0.4; latest 0.5 exists." }]) {
|
|
const original = first.visibleReview; first.visibleReview = { ...original as object, ...patch };
|
|
expect(check()).toBe(false); first.visibleReview = original;
|
|
}
|
|
for (const patch of [{ companyId: "foreign" }, { agentId: "foreign" }, { status: "failed" }, { nativeIssueId: "foreign" }, { contextSnapshot: { issueId: "source", commentId: "other" } }, { resultJson: { ...run.resultJson, finalizationPhase: "pending" } }]) {
|
|
const original = first.runs; Object.assign(first, { runs: [{ ...run, ...patch }] });
|
|
expect(check()).toBe(false); first.runs = original;
|
|
}
|
|
const originalVersion = first.upstream.currentVersionId; first.upstream.currentVersionId = "premature";
|
|
expect(check()).toBe(false); first.upstream.currentVersionId = originalVersion;
|
|
expect(acceptedSkillReviewRun(first, "foreign")).toBeUndefined();
|
|
});
|
|
|
|
it("rejects premature apply, listing-only review, stale installation and overwritten local work", () => {
|
|
const early = valid(bytes.old, bytes.next); early.state.reviewCheckpoint.upstream.currentVersionId = "new"; expect(gradeIntegratorSkillUpdate(early).passed).toBe(false);
|
|
const listing = valid(bytes.old, bytes.next); listing.state.reviewCheckpoint.reply = "Current installed 0.4. Latest 0.5 is available."; expect(gradeIntegratorSkillUpdate(listing).passed).toBe(false);
|
|
const stale = valid(bytes.old, bytes.next); stale.upstream.markdown = bytes.old; expect(gradeIntegratorSkillUpdate(stale).passed).toBe(false);
|
|
const overwritten = valid(bytes.old, bytes.next); overwritten.localFork.markdown = bytes.next; expect(gradeIntegratorSkillUpdate(overwritten).passed).toBe(false);
|
|
const lying = valid(bytes.old, bytes.next); lying.reply = `${lying.state.reviewCheckpoint.reply}\nI haven't updated the accessibility skill, only fetched its source.`; expect(gradeIntegratorSkillUpdate(lying).passed).toBe(false);
|
|
});
|
|
});
|
|
|
|
it("requires the actual pinned update prerequisite before admitting the human request", async () => {
|
|
const real = { supported: true, hasUpdate: true, currentRef: pins.oldCommit, latestRef: pins.newCommit, trackingRef: pins.newCommit };
|
|
const evidence = vi.fn(async (_name: string, _value: unknown) => {});
|
|
const read = vi.fn(async (): Promise<{ supported: boolean; hasUpdate: boolean; currentRef: string; latestRef: string; trackingRef: string }> => real);
|
|
const ctx = { observationDeadlineAt: Date.now() + 30_000, input: { api: { getSkillSourceUpdateStatus: read }, fixtures: { company: { id: "company" } }, evidence } } as unknown as Parameters<typeof requireSkillUpdatePrerequisite>[0];
|
|
await requireSkillUpdatePrerequisite(ctx, "selected"); expect(read).toHaveBeenCalledWith("company", "selected", { timeout: expect.any(Number) });
|
|
expect(evidence).toHaveBeenCalledWith("chat-skill-update-prerequisite.json", expect.objectContaining({ currentPinnedRevisionMatched: true, latestPinnedRevisionMatched: true }));
|
|
for (const bad of [{ ...real, supported: false }, { ...real, hasUpdate: false }, { ...real, currentRef: pins.newCommit }, { ...real, latestRef: pins.oldCommit }, { ...real, trackingRef: "wrong" }]) {
|
|
read.mockResolvedValueOnce(bad); await expect(requireSkillUpdatePrerequisite(ctx, "selected")).rejects.toThrow("unavailable");
|
|
}
|
|
read.mockRejectedValueOnce(new Error("closed HTTP error")); await expect(requireSkillUpdatePrerequisite(ctx, "selected")).rejects.toThrow("closed HTTP error");
|
|
expect(evidence).toHaveBeenCalledOnce();
|
|
});
|
|
|
|
function setupFixture(unavailable: boolean) {
|
|
const source = { id: "source", revision: 1, entries: [{ path: pins.path, skillId: "up" }], lastScanCommit: pins.oldCommit, trackingRef: pins.newCommit, automaticUpdates: false };
|
|
const prerequisite = vi.fn(async () => { if (unavailable) throw new SkillSourceUpdateReadError(422, "skill_source_ref_unavailable", 422);
|
|
return { supported: true, hasUpdate: true, currentRef: pins.oldCommit, latestRef: pins.newCommit, trackingRef: pins.newCommit }; });
|
|
const ctx = { observationDeadlineAt: Date.now() + 30_000, input: { fixtures: { company: { id: "company" } }, evidence: vi.fn(async (_name: string, _value: unknown) => {}), api: {
|
|
post: async () => ({ source, imported: [{ id: "up" }] }), patch: async () => source,
|
|
get: async () => ({ id: "up", sourceType: "github", sourceRef: pins.oldCommit, markdown: bytes.old, editable: false }), getSkillSourceUpdateStatus: prerequisite,
|
|
} } } as unknown as Parameters<typeof requireSkillUpdatePrerequisite>[0];
|
|
return { ctx, prerequisite };
|
|
}
|
|
it("actual review-update setup blocks an unavailable dependency before preparing the canonical human request", async () => {
|
|
vi.mocked(saveSkillConversationContext).mockClear();
|
|
vi.stubGlobal("fetch", vi.fn(async (url: string) => new Response(url.includes(pins.oldCommit) ? bytes.old : bytes.next)));
|
|
try {
|
|
const f = setupFixture(true); await expect(integratorSkillUpdateCases[0]!.setup(f.ctx)).rejects.toMatchObject({ name: "SkillSourceUpdateReadError", reason: "skill_source_ref_unavailable" });
|
|
expect(f.prerequisite).toHaveBeenCalledOnce(); expect(saveSkillConversationContext).not.toHaveBeenCalled();
|
|
} finally { vi.unstubAllGlobals(); }
|
|
});
|
|
it("actual review-update setup keeps the real prerequisite and then prepares the unchanged source-aware request", async () => {
|
|
vi.mocked(saveSkillConversationContext).mockClear();
|
|
vi.stubGlobal("fetch", vi.fn(async (url: string) => new Response(url.includes(pins.oldCommit) ? bytes.old : bytes.next)));
|
|
try {
|
|
const f = setupFixture(false); const result = await integratorSkillUpdateCases[0]!.setup(f.ctx);
|
|
expect(integratorSkillUpdateCases[0]!.oracleVersion).toBe("paperclip.chat-catalog.integrator.v3");
|
|
expect(f.prerequisite).toHaveBeenCalledOnce(); expect(saveSkillConversationContext).toHaveBeenCalledOnce(); expect(result.source).toMatchObject({ trackingRef: pins.newCommit, lastScanCommit: pins.oldCommit });
|
|
} finally { vi.unstubAllGlobals(); }
|
|
});
|