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>
62 lines
7.7 KiB
TypeScript
62 lines
7.7 KiB
TypeScript
import { createHash } from "node:crypto";
|
|
import { describe, expect, it } from "vitest";
|
|
import { savedDesignInstruction, validRenderedCommentReference, waitForRenderedReply } from "./internal-comment-reference.js";
|
|
import { grade_tasks } from "./internal-tasks.js";
|
|
import type { InternalObservation } from "./internal-support.js";
|
|
function valid(body = "Follow the approved design task RUN-1 as the source of truth for this landing page build."): InternalObservation {
|
|
const task = { id: "landing", companyId: "company", title: "Landing page", status: "backlog" };
|
|
const design = { id: "design", companyId: "company", identifier: "RUN-1", title: "Approved design", status: "done" };
|
|
const comment = { id: "note", issueId: task.id, body, authorAgentId: "ceo", createdByRunId: "run" };
|
|
return { scenarioId: "tasks.comment-with-reference", origin: "http://127.0.0.1:3100", companyId: "company", leadId: "ceo", state: { task, design,
|
|
savedCommentReference: { commentId: comment.id, taskId: task.id, designId: design.id, bodySha256: createHash("sha256").update(body).digest("hex"), href: "/RUN/issues/RUN-1", visible: true, destinationMatched: true, publicReadMatched: true } },
|
|
tasks: [task, design], comments: [comment], projects: [], goals: [], activity: [], documents: [], revisions: [], reply: "Saved the note on the landing page task.", links: ["/RUN/issues/landing"], humanQueries: ["Add a note."], runs: [{ id: "run", agentId: "ceo", companyId: "company", status: "succeeded" }] };
|
|
}
|
|
it("accepts the retained natural bare-identifier note and explicit Markdown links with actual comment-scoped proof", () => {
|
|
for (const body of [valid().comments[0]!.body!, "Use [the approved design](/RUN/issues/RUN-1) when implementing the landing page.", "Build according to `RUN-1`.", "Follow [the design](http://127.0.0.1:3100/RUN/issues/RUN-1)."]) expect(grade_tasks(valid(body)).passed).toBe(true);
|
|
});
|
|
it("does not accept wrong/foreign/hidden or negated design instructions", () => {
|
|
for (const body of ["Follow RUN-2.", "Follow [RUN-1](https://foreign.invalid/issues/RUN-1).", "Follow [RUN-1](/FOREIGN/issues/RUN-1).", "Do not follow RUN-1.", "Ignore the approved design RUN-1.", "Follow RUN-1. Ignore that design.", "```\nFollow RUN-1\n```", "<!-- Follow RUN-1 -->", "RUN-1 is the design, but this note gives no instruction."]) {
|
|
const o = valid(body); expect(savedDesignInstruction(o, body, o.state.design as typeof o.tasks[0]), body).toBe(false); expect(grade_tasks(o).passed).toBe(false);
|
|
}
|
|
});
|
|
it("requires exact saved-comment, body, task, destination and readable-resource proof instead of aggregate links", () => {
|
|
for (const change of [{ commentId: "other-note" }, { taskId: "other-task" }, { designId: "other-design" }, { bodySha256: "stale" }, { href: "https://foreign.invalid/issues/RUN-1" }, { visible: false }, { destinationMatched: false }, { publicReadMatched: false }]) {
|
|
const o = valid(); Object.assign(o.state.savedCommentReference!, change); o.links.push("/RUN/issues/RUN-1"); expect(grade_tasks(o).passed).toBe(false);
|
|
}
|
|
const o = valid(); delete o.state.savedCommentReference; expect(validRenderedCommentReference(o, o.comments[0]!, o.tasks[1]!)).toBe(false);
|
|
});
|
|
it("retains exactly one real agent/run-attributed comment", () => {
|
|
for (const change of [{ authorAgentId: "human" }, { createdByRunId: "unrelated" }, { issueId: "foreign-task" }]) { const o = valid(); Object.assign(o.comments[0]!, change); expect(grade_tasks(o).passed).toBe(false); }
|
|
const o = valid(); o.comments.push({ ...o.comments[0]!, id: "duplicate" }); expect(grade_tasks(o).passed).toBe(false);
|
|
});
|
|
|
|
// Retain the original affirmative/negative instruction calibration matrix.
|
|
function observation(body: string): InternalObservation {
|
|
const task = { id: "task", companyId: "company", title: "Landing page", status: "backlog" }, design = { id: "design", identifier: "ACO-2", companyId: "company", title: "Approved design", status: "done" };
|
|
return { scenarioId: "tasks.comment-with-reference", origin: "https://fixture.test", companyId: "company", leadId: "ceo", state: { task: structuredClone(task), design: structuredClone(design), savedCommentReference: { commentId: "note", taskId: "task", designId: "design", bodySha256: createHash("sha256").update(body).digest("hex"), href: "/ACO/issues/design", visible: true, destinationMatched: true, publicReadMatched: true } }, tasks: [task, design], projects: [], goals: [], documents: [], revisions: [], activity: [], reply: "Note added.", humanQueries: [], links: ["/ACO/issues/task"], comments: [{ id: "note", issueId: "task", authorAgentId: "ceo", createdByRunId: "run", body }], runs: [{ id: "run", companyId: "company", agentId: "ceo", status: "succeeded" }] };
|
|
}
|
|
for (const body of ["Follow [the approved design](/ACO/issues/design).", "Please use [this design](https://fixture.test/ACO/issues/ACO-2) for implementation.", "Adhere to [the approved design](/ACO/issues/design?tab=documents).", "Implement according to [the design](/ACO/issues/design).", "Build in line with [the approved design](/ACO/issues/design).", "Follow [the approved design](/ACO/issues/design) and avoid unrelated layout changes.", "Do not forget to follow [the approved design](/ACO/issues/design)."])
|
|
it(`accepts affirmative actual design instruction: ${body}`, () => expect(grade_tasks(observation(body)).passed).toBe(true));
|
|
for (const body of ["Do not follow [the design](/ACO/issues/design).", "Never use [the design](/ACO/issues/design).", "Ignore [the design](/ACO/issues/design), but follow [the design](/ACO/issues/design).", "Follow [the design](/ACO/issues/design) but ignore it.", "Follow [the wrong task](/ACO/issues/other); the intended design ID is design.", "Follow [the design](https://foreign.test/ACO/issues/design).", "Follow [the design](/ACO/issues/design), but do not follow it.", "Here is [the design](/ACO/issues/design).", "Follow [another task](/ACO/issues/other). Reference [the design](/ACO/issues/design)."])
|
|
it(`rejects negated, unrelated or merely mentioned reference: ${body}`, () => expect(grade_tasks(observation(body)).passed).toBe(false));
|
|
it("preserves exact note authorship and duplicate rejection", () => { const o = observation("Follow [the design](/ACO/issues/design)."); o.comments[0]!.authorAgentId = "other"; expect(grade_tasks(o).passed).toBe(false); o.comments[0]!.authorAgentId = "ceo"; o.comments.push(structuredClone(o.comments[0]!)); expect(grade_tasks(o).passed).toBe(false); });
|
|
|
|
describe("waitForRenderedReply", () => {
|
|
const fixture = (renderAfterChecks: number, comments = [{ id: "reply", authorAgentId: "ceo", createdByRunId: "run" }]) => {
|
|
let checks = 0;
|
|
const locator = { filter: () => locator, count: async () => (++checks > renderAfterChecks ? 1 : 0) };
|
|
const ctx = { issue: () => ({ id: "chat" }), input: { api: { get: async () => comments }, page: { locator: () => locator } } };
|
|
return { ctx: ctx as unknown as Parameters<typeof waitForRenderedReply>[0], checks: () => checks };
|
|
};
|
|
const o = { leadId: "ceo", runs: [{ id: "run", agentId: "ceo", companyId: "company", status: "succeeded" }] } as Parameters<typeof waitForRenderedReply>[1];
|
|
it("waits until the saved reply is on the page instead of reading links at run settle", async () => {
|
|
const f = fixture(3); await waitForRenderedReply(f.ctx, o, 2_000, 1); expect(f.checks()).toBe(4);
|
|
});
|
|
it("names the unrendered reply on timeout instead of returning silently", async () => {
|
|
await expect(waitForRenderedReply(fixture(Number.POSITIVE_INFINITY).ctx, o, 20, 1)).rejects.toThrow(/Rendered reply wait: saved reply reply did not render/);
|
|
});
|
|
it("fails by name when the settled run saved no reply", async () => {
|
|
await expect(waitForRenderedReply(fixture(0, [{ id: "other", authorAgentId: "ceo", createdByRunId: "unrelated" }]).ctx, o, 20, 1)).rejects.toThrow(/saved no reply/);
|
|
});
|
|
});
|