Files
PaperClipAI/tests/runner-e2e/chat-catalog-registry.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

114 lines
12 KiB
TypeScript

import { validNativeStopDeclaration, assertNativeStopCallback } from "./chat-native-stop-boundary.js";
import { validSlackIngressDeclaration } from "./chat-slack-ingress.js";
import { validCompanionDeclaration } from "./chat-companion-surface.js";
import { assertExpectedWorkerCancellation } from "./chat-catalog-run-policy.js";
import { installedChatScenarioIds } from "./chat-installed-lane.js";
import { existsSync, readFileSync } from "node:fs";
import { validChatCatalogHumanCounts } from "./chat-catalog-contract.js";
import type { ChatCatalogCase } from "./chat-catalog-contract.js";
import { chatCatalogScenarios } from "./chat-catalog-inventory.js";
import type { IdentityBoundaryCase } from "./chat-catalog-domains/identity-boundary-contract.js";
import { identityBoundaryScenarioIds } from "./chat-catalog-domains/identity-boundary-contract.js";
export type ChatCatalogDescriptor = Omit<ChatCatalogCase, "expectedCancelledWorkerRun" | "beforeCanonicalSend" | "dispatchSlackOtherThread" | "verifySlackSupportedTransfer" | "dispatchSlackUnsupportedFile" | "dispatchSlackQuery" | "runCompanionSurface" | "journeyKind" | "setup" | "act" | "observe" | "grade" | "cleanup" | "verifyResult" | "verifySetup" | "verifyQueryRejection"> & {
journeyKind?: "agent-chat" | "browser-boundary" | "scoped-external-assistant" | "native-slack-thread";
boundaryTarget?: "authenticated-harness" | "empty-installed-target";
modulePath: string;
exportName: string;
};
/** Pure discovery metadata. Importing the catalog never evaluates a domain module. */
export const chatCatalogCases: readonly ChatCatalogDescriptor[] = JSON.parse(
readFileSync(new URL("./chat-catalog-descriptors.json", import.meta.url), "utf8"),
);
const known = new Set(chatCatalogScenarios.map(row => row.id));
export function findChatCatalogCase(scenarioId: string): ChatCatalogDescriptor | undefined {
const matches = chatCatalogCases.filter(row => row.scenarioId === scenarioId);
if (matches.length > 1) throw new Error(`Duplicate chat implementation: ${scenarioId}`);
const descriptor = matches[0];
if (!descriptor) return;
if (descriptor.runPolicy?.kind === "source-settled-with-active-work" && descriptor.runPolicy.expectedWorkerCancellation !== undefined
&& (scenarioId !== "tasks.edit-properties" || descriptor.runPolicy.expectedWorkerCancellation !== "owned-task-reassignment"))
throw new Error("Worker reassignment cancellation is declared only for edit-properties");
if ((descriptor.companionSurface || descriptor.journeyKind === "scoped-external-assistant") && !validCompanionDeclaration(descriptor)) throw new Error("Only the declared scoped-client catalog case may select an external assistant");
if ((descriptor.slackIngress || descriptor.journeyKind === "native-slack-thread") && !validSlackIngressDeclaration(descriptor)) throw new Error("Only two declared Slack thread cases may use native ingress");
if (descriptor.canonicalBoundary && !validNativeStopDeclaration(descriptor)) throw new Error("Native Stop boundary is exclusive to the exact human post-stop scenario");
if (descriptor.browserTimezone && (scenarioId !== "routines.create-daily-brief" || descriptor.browserTimezone !== "America/Chicago"))
throw new Error("Browser timezone is declared only for the canonical Chicago weekday routine");
if (descriptor.respondingAgent && (scenarioId !== "cases.handoff-same-task" || Object.keys(descriptor.respondingAgent).length !== 4
|| descriptor.respondingAgent.kind !== "case-primary-owner" || descriptor.respondingAgent.name !== "Morgan" || descriptor.respondingAgent.role !== "engineer" || descriptor.respondingAgent.canAssignTasks !== true))
throw new Error("A non-CEO responding fixture is declared only for the canonical Morgan primary-owner handoff");
if (descriptor.installedSource && (descriptor.installedSource !== "published-global-npm" || !(installedChatScenarioIds as readonly string[]).includes(scenarioId)))
throw new Error("Installed ordinary Chat is allowlisted only for the two published-source operator scenarios");
if (!known.has(scenarioId) || !descriptor.oracleVersion || !descriptor.sourceFiles.length
|| !/^\.\/chat-catalog-domains\/[a-z0-9-]+\.js$/.test(descriptor.modulePath)
|| !descriptor.exportName || (!["browser-boundary", "scoped-external-assistant"].includes(descriptor.journeyKind ?? "agent-chat") && !validChatCatalogHumanCounts(descriptor)) || descriptor.expectedRunCount < 0 || descriptor.turnTimeoutMs <= 0 || descriptor.attemptTimeoutMs <= 0
|| (descriptor.minimumRunCount ?? descriptor.expectedRunCount) > descriptor.expectedRunCount)
throw new Error(`Incomplete bounds/provenance for ${scenarioId}`);
return descriptor;
}
/** Source-local Playwright imports need the real .ts URL; compiled consumers retain .js. */
export function chatCatalogModuleUrl(modulePath: string): string {
const source = new URL(modulePath.replace(/\.js$/, ".ts"), import.meta.url);
return existsSync(source) ? source.href : new URL(modulePath, import.meta.url).href;
}
export async function loadChatCatalogImplementation(
descriptor: ChatCatalogDescriptor,
importer: (specifier: string) => Promise<Record<string, unknown>> = specifier => import(specifier),
): Promise<ChatCatalogCase> {
if (descriptor.journeyKind === "browser-boundary") throw new Error("Browser boundary cannot enter ordinary Agent Chat dispatch");
// The sole dynamic import boundary: only the selected descriptor's module executes.
const module = await importer(chatCatalogModuleUrl(descriptor.modulePath));
const implementations = module[descriptor.exportName];
if (!Array.isArray(implementations)) throw new Error(`Missing domain export: ${descriptor.exportName}`);
const selected = implementations.filter((row: ChatCatalogCase) => row.scenarioId === descriptor.scenarioId);
if (selected.length !== 1) throw new Error(`Missing or duplicate runtime case: ${descriptor.scenarioId}`);
const implementation = selected[0] as ChatCatalogCase;
assertExpectedWorkerCancellation(implementation);
if (implementation.optionalCompanySecretBindingApproval === true
? !validChatCatalogHumanCounts(implementation) || implementation.scenarioId!=="secrets.configure-company-secret" || implementation.expectedRunCount!==3 || implementation.minimumRunCount!==2 || typeof implementation.expectedCompanySecretBindingApproval!=="function"
: implementation.expectedCompanySecretBindingApproval!==undefined) throw new Error("Invalid optional governed company approval declaration");
for (const key of ["sourceFiles", "oracleVersion", "expectedRunCount", "minimumRunCount", "expectedHumanMessageCount", "optionalCompanySecretBindingApproval", "setupHumanMessageCount", "turnTimeoutMs", "attemptTimeoutMs", "coverageGaps", "journeyKind", "authentication", "setupPolicy", "runPolicy", "queryOutcome", "queryBindings", "queryAttachments", "requiredCredentials", "externalInput", "installedSource", "browserTimezone", "respondingAgent", "companionSurface", "slackIngress", "canonicalBoundary"] as const) {
if (JSON.stringify(implementation[key]) !== JSON.stringify(descriptor[key])) throw new Error(`Runtime descriptor drift: ${descriptor.scenarioId}.${key}`);
}
if ((descriptor.companionSurface || descriptor.journeyKind === "scoped-external-assistant") && (!validCompanionDeclaration(implementation) || typeof implementation.runCompanionSurface !== "function" || implementation.act)) throw new Error("Missing actual finite scoped assistant driver");
if ((descriptor.slackIngress || descriptor.journeyKind === "native-slack-thread") && (!validSlackIngressDeclaration(implementation) || typeof implementation.dispatchSlackQuery !== "function" || implementation.act)) throw new Error("Missing finite actual Slack ingress driver");
if (descriptor.scenarioId === "chat-channels.channel-file" && descriptor.slackIngress && (typeof implementation.verifySlackSupportedTransfer !== "function" || typeof implementation.dispatchSlackUnsupportedFile !== "function")) throw new Error("Slack file case requires actual supported and unsupported transfer phases");
if ((descriptor.scenarioId !== "chat-channels.channel-file" || !descriptor.slackIngress) && (implementation.verifySlackSupportedTransfer || implementation.dispatchSlackUnsupportedFile)) throw new Error("Undeclared Slack file conditional");
const declaredOtherThread = descriptor.scenarioId === "chat-channels.reply-in-thread" && descriptor.slackIngress?.setupThread === "same-agent-distractor";
if (declaredOtherThread ? typeof implementation.dispatchSlackOtherThread !== "function" : implementation.dispatchSlackOtherThread !== undefined) throw new Error("Missing or undeclared Slack other-thread callback");
assertNativeStopCallback(implementation);
if (!descriptor.slackIngress && implementation.dispatchSlackQuery) throw new Error("Undeclared Slack ingress driver");
if (!descriptor.companionSurface && implementation.runCompanionSurface) throw new Error("Undeclared external assistant driver");
if ((descriptor.setupPolicy?.funding === "preserve-scenario" || descriptor.setupPolicy?.readiness === "preserve-scenario")
&& (!descriptor.setupPolicy.invariantCheckIds.length || typeof implementation.verifySetup !== "function"))
throw new Error("Scenario-specific setup policy requires independent prerequisite assertions");
if (descriptor.queryOutcome === "rejected" && typeof implementation.verifyQueryRejection !== "function")
throw new Error("Rejected chat query requires independent rejection evidence");
if ([implementation.setup, implementation.observe, implementation.grade, implementation.verifyResult].some(value => typeof value !== "function"))
throw new Error(`Incomplete executable case: ${descriptor.scenarioId}`);
return implementation;
}
/** Separate typed dispatch: no company, agent, or ChatFlowInput is fabricated. */
export async function loadIdentityBoundaryImplementation(descriptor: ChatCatalogDescriptor,
importer: (specifier: string) => Promise<Record<string, unknown>> = specifier => import(specifier)): Promise<IdentityBoundaryCase> {
if (descriptor.canonicalBoundary || descriptor.journeyKind !== "browser-boundary" || !(identityBoundaryScenarioIds as readonly string[]).includes(descriptor.scenarioId))
throw new Error("Only explicit catalog browser boundaries may bypass Agent Chat prerequisites");
if (descriptor.scenarioId === "agent-chat.chat-disabled" && (descriptor.boundaryTarget !== "authenticated-harness" || descriptor.authentication !== "human-session"))
throw new Error("Disabled Agent Chat requires the actual owned authenticated company target");
if (descriptor.expectedRunCount !== 0 || (descriptor.minimumRunCount ?? 0) !== 0 || descriptor.expectedHumanMessageCount !== 0 || (descriptor.setupHumanMessageCount ?? 0) !== 0)
throw new Error("Pre-chat boundaries require zero agent admissions and zero persisted chat queries");
const module = await importer(chatCatalogModuleUrl(descriptor.modulePath));
const rows = module[descriptor.exportName];
const selected = Array.isArray(rows) ? rows.filter(row => row.scenarioId === descriptor.scenarioId) : [];
if (selected.length !== 1) throw new Error("Missing or duplicate browser boundary implementation");
const implementation = selected[0] as IdentityBoundaryCase;
for (const key of ["sourceFiles", "oracleVersion", "expectedRunCount", "minimumRunCount", "expectedHumanMessageCount", "optionalCompanySecretBindingApproval", "setupHumanMessageCount", "turnTimeoutMs", "attemptTimeoutMs", "coverageGaps", "journeyKind", "boundaryTarget", "authentication", "setupPolicy", "runPolicy", "queryOutcome", "queryBindings", "queryAttachments", "requiredCredentials", "externalInput", "installedSource"] as const) {
if (JSON.stringify(implementation[key]) !== JSON.stringify(descriptor[key])) throw new Error(`Runtime boundary descriptor drift: ${descriptor.scenarioId}.${key}`);
}
if ([implementation.run, implementation.grade, implementation.cleanup].some(value => typeof value !== "function") || descriptor.queryBindings?.length || descriptor.queryAttachments?.length
|| !["authenticated-harness", "empty-installed-target"].includes(String(descriptor.boundaryTarget)))
throw new Error("Boundary requires executable outcome, independent grade and cleanup; query bindings are Agent Chat only");
return implementation;
}