mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-14 08:53:19 +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>
53 lines
9.4 KiB
TypeScript
53 lines
9.4 KiB
TypeScript
import { expect, it } from "vitest";
|
|
import { gradeDelegation, completeLaunchBrief, launchPageReflectsBrief, linksOpenTask, RETURN_DELIVERY_LATENCY_MS, type LaunchPageFacts } from "./runtime-delegation.js";
|
|
function valid() {
|
|
const task = { id: "website", parentId: "launch", companyId: "company", assigneeAgentId: "morgan", status: "done", description: "Implement the approved design brief and report the completed result back." };
|
|
return { scenarioId: "task-coordination.delegate-with-context", companyId: "company", leadId: "ceo", agents: [], links: [], humanQueries: [], reply: "",
|
|
tasks: [{ id: "launch" }, task], activity: [{ id: "created", action: "issue.created", entityId: "website", agentId: "ceo" }],
|
|
runs: [{ id: "context", agentId: "ceo", status: "succeeded", contextSnapshot: { issueId: "chat" } }, { id: "delegate", agentId: "ceo", status: "succeeded", contextSnapshot: { issueId: "chat" } },
|
|
{ id: "implement", agentId: "morgan", status: "succeeded", contextSnapshot: { issueId: "website" } }, { id: "report", agentId: "ceo", status: "succeeded", contextSnapshot: { issueId: "chat" } }],
|
|
state: { parent: { id: "launch" }, task, worker: { id: "morgan" }, humanId: "human", briefCommentId: "human-brief", briefHash: "original-hash", briefDownloadedHash: "original-hash",
|
|
briefAttachment: { id: "brief", issueId: "launch", companyId: "company", createdByAgentId: null, createdByUserId: "human", issueCommentId: "human-brief", sha256: "original-hash" },
|
|
parentOwnedByCeo: true, parentContextRetained: true, createdFromSource: true, childHasBrief: true, linkedToLaunch: false, workerBriefReadStatus: 200, workerBriefHash: "original-hash", briefContextKind: "direct-attachment-reference", workerTaskReadStatus: 200, workerParentReadStatus: 200, parentLinkInspectable: true, durableBriefComplete: false,
|
|
workerWasReallyWaiting: true, actualWebsiteArtifact: true, returnDeliveryLatencyMs: 40_000 as number,
|
|
delivery: { checks: [{ id: "correlated", passed: true }], response: { body: "Morgan completed Website; the actual saved page is available on the child task." } } } };
|
|
}
|
|
it("accepts one genuinely contextual child and correlated actual result", () => expect(gradeDelegation(valid()).passed).toBe(true));
|
|
it("accepts natural correct-worker result reporting", () => { const o = valid(); o.state.delivery.response.body = "The page Morgan built is ready. Open Website for the delivered result from that implementation."; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("accepts an independently verified inspectable link as the supported alternative to a child edge", () => { const o = valid(); o.state.task.parentId = ""; o.state.linkedToLaunch = true; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("rejects missing actual delegation evidence", () => expect(gradeDelegation(undefined).passed).toBe(false));
|
|
it("rejects a delegated task unrelated to the original Launch work", () => { const o = valid(); o.state.task.parentId = "other"; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects a brief still available only in source chat", () => { const o = valid(); o.state.childHasBrief = false; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects a worker unable to fetch the actual brief", () => { const o = valid(); o.state.workerBriefReadStatus = 403; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects copying a similarly named brief with different bytes", () => { const o = valid(); o.state.workerBriefHash = "wrong"; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects completion with no real delivered page", () => { const o = valid(); o.state.actualWebsiteArtifact = false; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects a result reported on an uncorrelated source", () => { const o = valid(); o.state.delivery.checks[0]!.passed = false; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("does not grade the reply's wording: a terse correct correlated delivery passes", () => { const o = valid(); o.state.delivery.response.body = "Here it is: COM-3."; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("does not grade the child description's wording: the created-from-source return path is the completion request", () => { const o = valid(); o.state.task.description = "Build the landing page from the attached brief."; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("accepts delivery within one ordinary turn of Morgan finishing", () => { const o = valid(); o.state.returnDeliveryLatencyMs = RETURN_DELIVERY_LATENCY_MS; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it.each([RETURN_DELIVERY_LATENCY_MS + 1, 300_000, Number.NaN, -1])("rejects a late, missing or pre-completion return delivery (%s ms)", latency => { const o = valid(); o.state.returnDeliveryLatencyMs = latency; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("accepts a bare Launch identifier the task page renders as a link", () => expect(linksOpenTask(["/COM/issues/COM-1"], "http://127.0.0.1:3100", { id: "launch-uuid", identifier: "COM-1" })).toBe(true));
|
|
it("accepts a Launch UUID link", () => expect(linksOpenTask(["[Launch](/COM/issues/launch-uuid)".slice(9, -1)], "http://127.0.0.1:3100", { id: "launch-uuid", identifier: "COM-1" })).toBe(true));
|
|
it.each([["/COM/issues/COM-2"], ["https://elsewhere.invalid/COM/issues/COM-1"], ["/COM/agents/COM-1"], []])("rejects links that do not open Launch here: %s", (...targets) => expect(linksOpenTask(targets.flat(), "http://127.0.0.1:3100", { id: "launch-uuid", identifier: "COM-1" })).toBe(false));
|
|
it("rejects a hidden actual parent execution", () => { const o = valid(); o.runs[2]!.contextSnapshot.issueId = "launch"; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects duplicate work tasks", () => { const o = valid(); o.tasks.push({ id: "duplicate" }); expect(gradeDelegation(o).passed).toBe(false); });
|
|
|
|
it("accepts an actual worker-readable parent reference and its exact human attachment", () => { const o = valid(); o.state.briefContextKind = "readable-parent-reference"; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("rejects an inaccessible parent even when the board can read its attachment", () => { const o = valid(); o.state.briefContextKind = "readable-parent-reference"; o.state.workerParentReadStatus = 403; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects a parent reference with the wrong brief bytes", () => { const o = valid(); o.state.briefContextKind = "readable-parent-reference"; o.state.workerBriefHash = "other-design"; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("accepts a complete durable brief with actual grounded delivery without copying its attachment", () => { const o = valid(); o.state.briefContextKind = "faithful-durable-copy"; o.state.durableBriefComplete = true; o.state.workerBriefReadStatus = 404; o.state.workerBriefHash = ""; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("rejects an incomplete durable brief", () => { const o = valid(); o.state.briefContextKind = "faithful-durable-copy"; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it("rejects a durable copy without the worker's actual access to its task", () => { const o = valid(); o.state.briefContextKind = "faithful-durable-copy"; o.state.durableBriefComplete = true; o.state.workerTaskReadStatus = 403; expect(gradeDelegation(o).passed).toBe(false); });
|
|
it.each(["Notify the CEO in this conversation when the page is ready", "Implement the brief and let me know when done"])("accepts a clear natural completion request: %s", description => { const o = valid(); o.state.task.description = description; expect(gradeDelegation(o).passed).toBe(true); });
|
|
it("recognizes complete substantive durable brief content", () => expect(completeLaunchBrief("Reference: Launch brief nonce. Build semantic markup with accessible input labels, one primary join button, and the headline Research without the busywork.", "Launch brief nonce")).toBe(true));
|
|
it.each(["Unrelated headline; semantic HTML, accessible labels, one primary signup action. Launch brief nonce", "Research without the busywork. Launch brief nonce", "Research without the busywork; semantic HTML, accessible labels, one primary signup action. Other reference"])("rejects wrong or incomplete durable content", text => expect(completeLaunchBrief(text, "Launch brief nonce")).toBe(false));
|
|
|
|
const page: LaunchPageFacts = { headline: "Research without the busywork", landmarks: 4, unlabelledControls: 0, namedSignupActions: 1, signupForms: 1 };
|
|
it("accepts the rendered page facts of a brief-faithful page", () => expect(launchPageReflectsBrief(page)).toBe(true));
|
|
it("accepts a signup form whose button is not named with signup words", () => expect(launchPageReflectsBrief({ ...page, namedSignupActions: 0 })).toBe(true));
|
|
it("accepts a named signup link without a form", () => expect(launchPageReflectsBrief({ ...page, signupForms: 0 })).toBe(true));
|
|
it("accepts the headline split across inline markup and whitespace", () => expect(launchPageReflectsBrief({ ...page, headline: "\n Research without the busywork " })).toBe(true));
|
|
it.each<[string, Partial<LaunchPageFacts>]>([["wrong headline", { headline: "Unrelated result" }], ["no h1", { headline: null }], ["no semantic landmarks", { landmarks: 0 }],
|
|
["an unlabelled control", { unlabelledControls: 1 }], ["no signup action", { namedSignupActions: 0, signupForms: 0 }]])("rejects a page with %s", (_, change) => expect(launchPageReflectsBrief({ ...page, ...change })).toBe(false));
|
|
it("rejects a page that was never rendered", () => expect(launchPageReflectsBrief(undefined)).toBe(false));
|