Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/external-teams.ts
T
f4337bb5fb test(evals): land the chat catalog eval harness on master (#19614)
## 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>
2026-11-10 08:02:11 -06:00

95 lines
8.1 KiB
TypeScript

import type { CatalogTeam, CatalogTeamImportPreviewResult } from "../../../packages/shared/src/types/teams-catalog.js";
import { definition, externalObservation, observeExternal, root, sameRows, checks, verifyRenderedReply } from "./external-support.js";
import { collectRunEvents } from "../run-observations.js";
const LIST_TEAMS = new Set(["GET /api/teams/catalog", "GET /api/teams/catalog/{catalogId}"]);
const PREVIEW_TEAM = "POST /api/companies/{companyId}/teams/catalog/{catalogId}/preview";
const INSTALL_TEAM = "POST /api/companies/{companyId}/teams/catalog/{catalogId}/install";
/** Successful call_api operations the agent's own runs executed, from their recorded tool results. */
export function teamApiOperations(events: readonly { payload?: Record<string, unknown> | null }[]) {
const operations: string[] = [];
for (const event of events) {
const prp = event.payload?.prpEvent as { payload?: { item?: { type?: string; result?: { ok?: unknown; apiOperationId?: unknown } } } } | undefined;
const item = prp?.payload?.item;
if (item?.type === "tool_result" && item.result?.ok === true && typeof item.result.apiOperationId === "string") operations.push(item.result.apiOperationId);
}
return operations;
}
/** Identity check only: the catalog's own name, slug, key or id, ignoring Markdown emphasis and case. */
export function namesCatalogTeam(reply: string, team: Pick<CatalogTeam, "name" | "slug" | "key" | "id">) {
const text = reply.replace(/[*_`]/g, "").toLowerCase();
return [team.name, team.slug, team.slug.replaceAll("-", " "), team.key, team.id].some(name => !!name && text.includes(name.toLowerCase()));
}
/** Identity check only: each required skill is named by a literal the catalog returned for it (ref, id or key). */
export function namesRequiredSkills(reply: string, team: Pick<CatalogTeam, "requiredSkills">) {
const text = reply.replace(/[*_`]/g, "").toLowerCase();
const names = (skill: CatalogTeam["requiredSkills"][number]) => [skill.ref, skill.ref.replaceAll("-", " "),
skill.catalogSkillId, skill.catalogSkillKey];
return team.requiredSkills.every(skill => names(skill).some(name => !!name && text.includes(name.toLowerCase())));
}
/** The web app routes no teams catalog page, so a /teams-catalog link opens Page not found. */
export function linksMissingTeamCatalogPage(o: { links: string[]; origin: string }) {
return o.links.some(href => { try { return new URL(href, o.origin).pathname.split("/").includes("teams-catalog"); } catch { return false; } });
}
// The reply's summary of roles, requirements and counts is read by a person; it is
// not matched with keywords. Grading uses the run's own catalog reads, the saved
// state and the preview the product computes. The web app has no catalog page to link.
export function gradeTeams(value: unknown) {
const c = checks(); const o = externalObservation(value);
c.add("complete-team-catalog-evidence", !!o, "Public catalog metadata, before/after company resources and reply links are required."); if (!o) return c.result();
const team = o.state.team as CatalogTeam | undefined;
const operations = o.state.runEventsRead === true && Array.isArray(o.state.teamApiOperations) ? o.state.teamApiOperations as string[] : undefined;
c.add("existing-package-grounding", !!team && team.compatibility === "compatible" && team.agentSlugs.length >= 2 && team.counts.projects > 0
&& !!operations?.some(operation => LIST_TEAMS.has(operation)), "The agent read the real teams catalog in its own run before answering.");
c.add("names-catalog-team", !!team && namesCatalogTeam(o.reply, team), "The reply identifies the selected catalog team by a name, slug or id the catalog returned.");
// The expected outcome calls out requirements in the package metadata. The required skill refs are
// catalog literals, so naming each one is an identity check, not wording.
if (o.scenarioId === "teams.find-team") c.add("names-required-skills", !!team && team.requiredSkills.length > 0 && namesRequiredSkills(o.reply, team),
"The reply names every required skill the catalog lists for the recommended team.");
c.add("no-broken-catalog-link", !linksMissingTeamCatalogPage(o), "The reply does not link a teams catalog page, which the web app does not route.");
c.add("no-premature-installation", sameRows(o.state.installedBefore, o.state.installedAfter) && sameRows(o.state.agentIdsBefore, o.state.agentIdsAfter)
&& sameRows(o.state.projectIdsBefore, o.state.projectIdsAfter) && !operations?.includes(INSTALL_TEAM),
"Catalog discovery and preview cannot create agents, projects or an installation.");
if (o.scenarioId === "teams.preview-install") {
const preview = o.state.preview as CatalogTeamImportPreviewResult | undefined;
c.add("preview-requirements-grounded", !!preview && preview.team.contentHash === team?.contentHash && Array.isArray(preview.errors),
"The product's own preview of the selected package is current and readable.");
c.add("preview-resources-described", !!operations?.includes(PREVIEW_TEAM), "The agent ran the supported install preview in its own run instead of describing an install from memory.");
}
return c.result();
}
export const teamExternalCases = ["find-team", "preview-install"].map(id => definition(`teams.${id}`, async ctx => {
const catalog = await ctx.input.api.get<CatalogTeam[]>("/api/teams/catalog");
const team = catalog.find(row => row.slug === "product-engineering" && row.compatibility === "compatible");
if (!team || team.agentSlugs.length < 2 || team.counts.projects < 1) throw new Error("Setup blocker: supported Product Engineering catalog package with agents/project is absent");
const [agents, projects, installed] = await Promise.all([
ctx.input.api.get<Array<{ id: string }>>(`${root(ctx)}/agents`), ctx.input.api.get<Array<{ id: string }>>(`${root(ctx)}/projects`),
ctx.input.api.get(`${root(ctx)}/teams/catalog/installed`),
]);
// Context selection is a prerequisite, not a preview or installation result.
await ctx.input.api.post(`${root(ctx)}/chats/${ctx.input.fixtures.agent.id}`); await ctx.refreshIssue();
if (!ctx.issue()) throw new Error("Standing chat context was not created");
await ctx.input.api.patch(`/api/issues/${ctx.issue()!.id}`, { description: `We are choosing a team to build a website. ${id === "preview-install" ? `The selected package is ${team.name}, catalog key ${team.key}. No installation has been accepted.` : "No package has been installed."}` });
return { team, agentIdsBefore: agents.map(row => row.id).sort(), projectIdsBefore: projects.map(row => row.id).sort(), installedBefore: installed,
};
}, gradeTeams, {
async observe(ctx, state) {
const [agents, projects, installed] = await Promise.all([
ctx.input.api.get<Array<{ id: string }>>(`${root(ctx)}/agents`), ctx.input.api.get<Array<{ id: string }>>(`${root(ctx)}/projects`),
ctx.input.api.get(`${root(ctx)}/teams/catalog/installed`),
]);
state.agentIdsAfter = agents.map(row => row.id).sort(); state.projectIdsAfter = projects.map(row => row.id).sort(); state.installedAfter = installed;
if (id === "preview-install") state.preview = await ctx.input.api.post(`${root(ctx)}/teams/catalog/${encodeURIComponent((state.team as CatalogTeam).id)}/preview`, { targetManagerAgentId: ctx.input.fixtures.agent.id });
const events: { seq?: number; payload?: Record<string, unknown> | null }[] = [];
for (const run of await ctx.allRuns()) events.push(...await collectRunEvents<{ seq?: number; payload?: Record<string, unknown> | null }>((afterSeq, limit) =>
ctx.input.api.getRunObservation(`/api/heartbeat-runs/${run.id}/events?afterSeq=${afterSeq}&limit=${limit}`)));
state.runEventsRead = true; state.teamApiOperations = teamApiOperations(events);
return observeExternal(ctx, state);
},
// The reply is the result: the catalog has no web page to open from it.
async verifyResult(ctx, _state, value) { await verifyRenderedReply(ctx, value); },
oracleVersion: "paperclip.chat-catalog.team-discovery.v3",
sourceFiles: ["chat-catalog-domains/external-support.ts", "chat-catalog-domains/internal-support.ts", "chat-catalog-domains/external-teams.ts",
"chat-scenarios/teams.yaml", "run-observations.ts"],
}));