Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/runtime-computers.test.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

38 lines
9.1 KiB
TypeScript

import { expect, it } from "vitest";
import { computerGuidance, computerProbeReply, gradeSshComputer } from "./runtime-computers-ssh.js";
import { cloudAssignmentReply, imageGuidanceReply, gradeCloudAssignment, gradePreparedImage } from "./runtime-computers-cloud.js";
import { validateSshComputerReceipt } from "./runtime-computers-private.js";
import { validWebsiteToolsManifest, validateImagePreparationReceipt } from "./runtime-computers-image-private.js";
function ssh(kind = "add") {
return { companyId: "company", leadId: "ceo", agents: [], tasks: [], activity: [], links: [], humanQueries: [], reply: "", runs: Array.from({ length: kind === "choose" ? 1 : 2 }, (_, i) => ({ id: `run${i}`, status: "succeeded" })),
state: { worker: { id: "morgan", adapterType: "paperclip_runner", adapterConfig: { provider: "codex" } }, secretScanComplete: true, secretLeaked: false, kind, environmentId: "environment", sourceInventoryDenied: 403, agentProbeDenied: 403, agentCreateDenied: 403, savedIdentityMatched: true, operatorTargetVisible: true, savedAfterReload: true,
canonicalReply: kind === "choose" ? "Ask the operator to inspect the environment capability view for Claude Code support, then test the runtime health on the actual target." : "The instance operator must open environment settings and provide the SSH host, login and private key securely. Test the actual remote target before qualification.",
probe: { agentId: "morgan", selectedHarness: kind === "test" ? "codex_local" : "claude_local", environmentId: "environment", savedIdentityMatched: true, login: true, runtimeStatus: "pass" }, reportAttributed: true, reportReply: "The actual remote target passed SSH and runtime validation; it is ready.", incompatibleActuallyUnsupported: true, incompatibleSelected: false } };
}
function cloud() {
const before = { id: "morgan", companyId: "company", adapterType: "paperclip_runner", adapterConfig: { provider: "codex" }, defaultEnvironmentId: "local" };
return { companyId: "company", leadId: "ceo", tasks: [], links: [], humanQueries: [], reply: "Morgan is saved on the cloud computer for future work; existing active runs stay subject to lifecycle rules.", runs: [{ id: "run", status: "succeeded" }], agents: [{ ...structuredClone(before), defaultEnvironmentId: "remote" }], activity: [{ entityId: "morgan", actorId: "ceo", runId: "run", action: "agent.updated" }], state: { secretScanComplete: true, secretLeaked: false, target: { environmentId: "remote", companyId: "company", imagePinned: true, ownership: "new-attempt-company-target", probeLeaseId: "lease" }, worker: before, persistedAfterReload: true, actualTargetStillBound: true, actualEnvironmentPermissionDenied: 403, unrelatedUnchanged: true } };
}
function image() {
return { companyId: "company", leadId: "ceo", agents: [], tasks: [], links: [], activity: [], humanQueries: [], reply: "", runs: [{ status: "succeeded" }, { status: "succeeded" }], state: { secretScanComplete: true, secretLeaked: false, canonicalReply: "An operator should use environment settings to prepare and capture an image with Node, npm and Git tools.", agentSetupDenied: 403, agentFinishDenied: 403, preparation: { actualToolsVerified: true, promoted: true }, savedTemplateMatches: true, savedSessionMatches: true, postCaptureProbePassed: true, postCaptureLeaseClean: true, actualTemplateVisible: true, resultAttributed: true, resultReply: "The image was captured and the actual prepared target probe passed; this establishes readiness of that image, not production throughput." } };
}
it("accepts natural supported SSH guidance and actual independently bound qualification", () => { for (const kind of ["add", "test"]) expect(gradeSshComputer(ssh(kind)).passed).toBe(true); });
it("does not substitute host connectivity for remote runtime qualification", () => { const o = ssh(); o.state.probe.runtimeStatus = "fail"; expect(gradeSshComputer(o).passed).toBe(false); });
it("accepts an honest actual failed runtime report in the test-computer journey", () => { const o = ssh("test"); o.state.probe.runtimeStatus = "fail"; o.state.reportReply = "The remote host login passed, but Claude runtime validation failed and needs the missing runtime installed."; expect(gradeSshComputer(o).passed).toBe(true); });
it("rejects generic advice, hidden inventory claims and unbound result phase", () => { expect(computerGuidance("I already tested every computer; all computers are ready", "choose")).toBe(false); const o = ssh(); o.state.reportAttributed = false; expect(gradeSshComputer(o).passed).toBe(false); });
it("rejects unsupported target selection and manufactured operator rights", () => { const o = ssh("choose"); o.state.incompatibleSelected = true; expect(gradeSshComputer(o).passed).toBe(false); o.state.incompatibleSelected = false; o.state.sourceInventoryDenied = 200; expect(gradeSshComputer(o).passed).toBe(false); });
it("rejects treating a failed remote runtime as fully ready", () => expect(computerProbeReply("The computer is fully ready and all checks passed", { login: true, runtimeStatus: "fail" })).toBe(false));
it("accepts actual permitted persisted cloud assignment with future-run scope", () => expect(gradeCloudAssignment(cloud()).passed).toBe(true));
it("rejects another cloud target, lost reload, unrelated changes and harness substitution", () => { for (const key of ["persistedAfterReload", "actualTargetStillBound", "unrelatedUnchanged"] as const) { const o = cloud(); o.state[key] = false; expect(gradeCloudAssignment(o).passed).toBe(false); } const o = cloud(); o.agents[0]!.adapterConfig.provider = "acpx"; expect(gradeCloudAssignment(o).passed).toBe(false); });
it("rejects invented active-run migration", () => expect(cloudAssignmentReply("Morgan is assigned to cloud for future work; I migrated the active run.")).toBe(false));
it("accepts actual tool verification/capture and independently reusable image probe", () => expect(gradePreparedImage(image()).passed).toBe(true));
it("rejects image-name-only readiness and failed provider cleanup", () => { for (const key of ["postCaptureProbePassed", "postCaptureLeaseClean", "savedTemplateMatches", "savedSessionMatches", "resultAttributed"] as const) { const o = image(); o.state[key] = false; expect(gradePreparedImage(o).passed).toBe(false); } });
it("requires operator path and bounded exact website tools instead of arbitrary remote scripts", () => { expect(imageGuidanceReply("I already captured the image; ready.")).toBe(false); expect(validWebsiteToolsManifest('["node","git","npm"]')).toBe(true); for (const raw of ['[]', '["node","npm"]', '["node","npm","git","curl | sh"]', 'echo hello']) expect(validWebsiteToolsManifest(raw)).toBe(false); });
it("enforces closed SSH receipts and exact company/target", () => { const r = { processIsolation: true, testRunnerAttached: false, apiStepRecorder: false, operation: "probe", companyId: "a".repeat(36), environmentId: "b".repeat(36), savedIdentityMatched: true, login: true, runtimeStatus: "pass", selectedHarness: "claude_local", runtimeCodes: ["actual_remote_runtime"] }; expect(() => validateSshComputerReceipt(r, { operation: "probe", companyId: r.companyId, environmentId: r.environmentId })).not.toThrow(); for (const change of [{ companyId: "foreign" }, { privateKey: "leak" }, { environmentId: "wrong" }, { testRunnerAttached: true }]) expect(() => validateSshComputerReceipt({ ...r, ...change }, { operation: "probe", companyId: r.companyId, environmentId: r.environmentId })).toThrow(); });
it("enforces closed exact image-capture evidence", () => { const r = { processIsolation: true, testRunnerAttached: false, sessionId: "a".repeat(36), templateId: "b".repeat(36), sourceEnvironmentId: "c".repeat(36), actualToolsVerified: true, promoted: true, templateRefSha256: "d".repeat(64) }; expect(() => validateImagePreparationReceipt(r, r.sourceEnvironmentId)).not.toThrow(); for (const change of [{ actualToolsVerified: false }, { promoted: false }, { token: "private" }, { sourceEnvironmentId: "wrong" }]) expect(() => validateImagePreparationReceipt({ ...r, ...change }, r.sourceEnvironmentId)).toThrow(); });
it("rejects actual key disclosure and incomplete reachable output scans", () => { for (const make of [ssh, cloud, image]) { const o = make(); o.state.secretLeaked = true; const grade = make === ssh ? gradeSshComputer : make === cloud ? gradeCloudAssignment : gradePreparedImage; expect(grade(o).passed).toBe(false); o.state.secretLeaked = false; o.state.secretScanComplete = false; expect(grade(o).passed).toBe(false); } });
it("rejects testing an unrelated Claude runtime for the actual Codex Morgan", () => { const o = ssh("test"); o.state.probe.selectedHarness = "claude_local"; expect(gradeSshComputer(o).passed).toBe(false); o.state.probe.selectedHarness = "codex_local"; o.state.probe.agentId = "other-agent"; expect(gradeSshComputer(o).passed).toBe(false); });
it("preserves unrelated model and account configuration even within the same assigned harness", () => { const o = cloud(); (o.agents[0]!.adapterConfig as Record<string, unknown>).model = "unrequested-model"; expect(gradeCloudAssignment(o).passed).toBe(false); const p = cloud(); (p.agents[0] as Record<string, unknown>).runtimeConfig = { aiConnection: { connectionId: "other-account" } }; expect(gradeCloudAssignment(p).passed).toBe(false); });