Files
PaperClipAI/tests/runner-e2e/chat-catalog-domains/external-slack-thread.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

93 lines
16 KiB
TypeScript
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
import {createHash} from "node:crypto";
import {readFile} from "node:fs/promises";
import {expect} from "@playwright/test";
import type {ChatCatalogContext,ChatSlackIngressReceipt} from "../chat-catalog-contract.js";
import type {ChatActivityItem} from "../../../packages/shared/src/types/chat-channels.js";
import type {IssueAttachment} from "../../../packages/shared/src/types/issue.js";
import type {IssueWorkProduct} from "../../../packages/shared/src/types/work-product.js";
import {pollUntil} from "../api.js";
import {definition,observeExternal,externalObservation,checks,root,verifyRenderedReply,type ExternalObservation} from "./external-support.js";
import {project,type Entity} from "./internal-support.js";
import {identityWorker} from "./identity-support.js";
import {prepareSlackFixture,slackCompanionFields,type SlackMessage} from "./external-slack-fixture.js";
import {renderSlackLaunchArtifact,hasSlackLaunchRendering,type SlackLaunchRendering} from "./external-slack-render.js";
const hash=(value:string|Buffer)=>createHash("sha256").update(value).digest("hex");
interface Message {id:string;user:string;text:string;files:string[]}
const projection=(rows:SlackMessage[]):Message[]=>rows.map(row=>({id:row.ts,user:row.user??"",text:row.text??"",files:(row.files??[]).map(file=>file.id).sort()}));
interface ThreadEvidence {source:ChatSlackIngressReceipt;other:ChatSlackIngressReceipt&{runId:string};before:Message[];after:Message[];messages:Message[];duplicate:{canonicalDeliveryCaptured:boolean;duplicateReplayed:boolean};activity:ChatActivityItem[];attachments:IssueAttachment[];products:IssueWorkProduct[];files:Array<{id:string;hash:string;body:string;rendered?:SlackLaunchRendering}>}
interface Observation extends ExternalObservation {thread:ThreadEvidence}
type Fixture=Awaited<ReturnType<typeof prepareSlackFixture>>;
const privateFixtures=new WeakMap<ChatCatalogContext,{fixture:Fixture;otherRoot?:string;before?:Message[];lastRead?:number;rendered?:Map<string,SlackLaunchRendering>;lastThreads?:{after:Message[];messages:Message[]}}>();
function owned(ctx:ChatCatalogContext){const value=privateFixtures.get(ctx);if(!value)throw new Error("Actual Slack fixture is not prepared");return value;}
function completed(text:string){return !/\b(?:when|once|until|after)\b[^.!?\n]{0,100}\b(?:ready|completed|finished|built|delivered|done)\b/i.test(text)&&/\b(?:ready|completed|finished|built|delivered|done)\b/i.test(text)&&!/\b(?:not|isn['’]t|is not|hasn['’]t|have not)\s+(?:yet\s+)?(?:ready|completed|finished|built|delivered|done)\b|\b(?:will|would|should)\s+(?:be\s+)?(?:ready|completed|finished|built|delivered|done)\b/i.test(text);}
export function gradeSlackThread(value:unknown){
const c=checks(),o=externalObservation(value) as Observation|undefined,t=o?.thread;
c.add("actual-provider-thread-evidence",!!o&&!!t&&!!t.source?.provider&&!!t.other?.provider&&[t.before,t.after,t.messages,t.activity,t.attachments,t.products,t.files].every(Array.isArray),"Actual Slack messages, original and other-source receipts, registered result bytes and public lifecycle activity are required.");if(!o||!t||!t.source?.provider||!t.other?.provider||![t.before,t.after,t.messages,t.activity,t.attachments,t.products,t.files].every(Array.isArray))return c.result();
const worker=o.state.worker as Entity|undefined,source=t.source,other=t.other;
const tasks=o.tasks.filter(task=>task.id!==source.issueId&&task.id!==other.issueId&&task.assigneeAgentId===worker?.id&&task.companyId===o.companyId);
const target=tasks[0],workerRuns=target?o.runs.filter(run=>run.agentId===worker?.id&&run.contextSnapshot?.issueId===target.id):[];
const sourceRuns=o.runs.filter(run=>run.contextSnapshot?.issueId===source.issueId),otherRuns=o.runs.filter(run=>run.contextSnapshot?.issueId===other.issueId);
c.add("exact-real-human-two-thread-context",source.query===o.state.originalQuery&&source.endpointId===other.endpointId&&source.issueId!==other.issueId&&source.provider.threadId!==other.provider.threadId&&source.provider.workspaceId===other.provider.workspaceId&&source.provider.channelId===other.provider.channelId&&source.linkedPaperclipUserId===other.linkedPaperclipUserId&&source.provider.senderUserId===other.provider.senderUserId&&otherRuns.length===1&&otherRuns[0]?.id===other.runId&&otherRuns[0]?.agentId===o.leadId&&otherRuns[0]?.status==="succeeded","The genuine linked sender has two distinct conversations under the same actual assigned agent; the setup source remains separately attributed.");
c.add("one-attributed-Morgan-build",!!worker&&tasks.length===1&&!!target&&target.status==="done"&&target.projectId===(o.state.project as Entity|undefined)?.id&&sourceRuns.length>=1&&workerRuns.length===1&&workerRuns[0]?.status==="succeeded"&&o.runs.every(run=>run.companyId===o.companyId&&run.runtimeMode==="native"&&run.status==="succeeded"&&(run.contextSnapshot?.issueId===source.issueId&&run.agentId===o.leadId||run.id===other.runId||run.contextSnapshot?.issueId===target.id&&run.agentId===worker.id))&&o.activity.some(row=>row.action==="issue.created"&&row.entityId===target.id&&row.actorType==="agent"&&row.agentId===o.leadId),"Exactly one actual Morgan task is built under normal policy, with independently attributed task creation and source/worker native executions; no unrelated run is accepted.");
const file=t.attachments.find(file=>file.issueId===target?.id&&file.originatingRunId===workerRuns[0]?.id&&t.products.some(product=>product.issueId===target?.id&&product.createdByRunId===workerRuns[0]?.id&&product.type==="artifact"&&["active","ready_for_review","approved"].includes(product.status)&&product.metadata?.attachmentId===file.id)&&t.files.some(actual=>actual.id===file.id&&actual.hash===file.sha256&&hash(actual.body)===file.sha256&&/<html\b/i.test(actual.body)&&/<h1\b/i.test(actual.body)&&/Acorn\s+Notes/i.test(actual.body)&&!!actual.rendered&&actual.rendered.sha256===file.sha256&&hasSlackLaunchRendering(actual.rendered)));
c.add("inspectable-real-launch-page",!!file,"A real worker-attributed registered HTML result has independently retrieved matching bytes and the requested launch identity.");
const reference=target?.identifier,completionReplies=o.phaseReplies.filter(reply=>sourceRuns.some(run=>run.id===reply.runId)&&!!reference&&reply.body.includes(reference)&&completed(reply.body));
const delivered=t.messages.filter(message=>message.user===o.state.botUserId&&!!reference&&message.text.includes(reference)&&completed(message.text));
c.add("completion-published-back-to-original-thread",completionReplies.length>=1&&delivered.length===1&&t.activity.some(row=>row.kind==="publication"&&row.status==="published"),"Actual successful source completion and exactly one independently read app-bot completion identify the same built task in the original Slack thread.");
c.add("other-thread-receives-no-private-content",t.before.length>0&&JSON.stringify(t.before)===JSON.stringify(t.after),"The other live conversation's actual message IDs, authors, text and file references remain unchanged after the canonical request.");
c.add("genuine-duplicate-is-idempotent",t.duplicate.canonicalDeliveryCaptured&&t.duplicate.duplicateReplayed&&t.activity.some(row=>row.kind==="delivery"&&row.status==="processed"&&/duplicate.*ignored/i.test(row.summary))&&tasks.length===1&&delivered.length===1,"The exact genuine signed event was replayed once; the product records duplicate suppression without another task or outgoing completion.");return c.result();
}
export const slackThreadExternalCases=[definition("chat-channels.reply-in-thread",async ctx=>{
const worker=await identityWorker(ctx,"Morgan"),p=await project(ctx,"Launch",{description:"Acorn Notes launch page: introduce the calm research notebook and offer a join-the-list signup link. Produce an inspectable static HTML launch page in the authorized workspace.",workspace:{name:"Launch workspace",sourceType:"local_path",cwd:ctx.input.workspacePath,isPrimary:true}});
const fixture=await prepareSlackFixture(ctx);privateFixtures.set(ctx,{fixture});
try{await ctx.input.api.patch(`/api/chat-endpoints/${fixture.endpoint.id}`,{executionDefaults:{projectId:p.id}});return {worker,project:p,originalQuery:ctx.scenario.query,botUserId:fixture.binding.prerequisites.botUserId};}catch(error){await fixture.cleanup();privateFixtures.delete(ctx);throw error;}
},gradeSlackThread,{
authentication:"human-session",journeyKind:"native-slack-thread",slackIngress:{kind:"native-slack-thread",maximumProviderCalls:20,ingressTimeoutMs:120000,setupThread:"same-agent-distractor"},expectedRunCount:4,minimumRunCount:2,expectedHumanMessageCount:1,turnTimeoutMs:120000,attemptTimeoutMs:600000,
externalInput:{provider:"slack",fields:slackCompanionFields},
sourceFiles:["chat-catalog-domains/external-slack-thread.ts","chat-catalog-domains/external-slack-fixture.ts","chat-catalog-domains/external-slack-provider.ts","chat-catalog-domains/external-slack-relay.ts","chat-catalog-domains/external-slack-render.ts","chat-catalog-domains/runtime-rendered-artifact.ts","chat-catalog-domains/external-fixtures.ts","chat-catalog-domains/external-support.ts","chat-catalog-domains/internal-support.ts","chat-catalog-domains/identity-support.ts","chat-slack-ingress.ts","../../packages/shared/src/types/chat-channels.ts","../../server/src/routes/chat-channels.ts","../../server/src/services/chat-channels.ts","../../server/src/services/chat-channels/inbound/authorization.ts","../../server/src/services/chat-channels/inbound/wakeups.ts","chat-scenarios/chat-channels.yaml"],
async dispatchSlackOtherThread(ctx,_state,{query,signal}){
const local=owned(ctx),f=local.fixture,start=f.provider.receipts.length;signal.throwIfAborted();
const ts=await f.createDistractor(query);local.otherRoot=ts;
const receipt=await f.resolveIngress(query,ts,f.provider.receipts.length-start,ts);
const ready=await pollUntil({label:"actual other Slack source's sole completed native",deadlineAt:Date.now()+100000,intervalMs:1000,load:()=>ctx.allRuns(),accept:runs=>runs.length===1&&runs[0]?.runtimeMode==="native"&&runs[0]?.status==="succeeded"&&runs[0]?.agentId===ctx.input.fixtures.agent.id&&runs[0]?.contextSnapshot?.issueId===receipt.issueId,reject:runs=>runs.length>1||runs.some(run=>["failed","cancelled","timed_out"].includes(run.status))?"Other-thread setup admitted extra or failed native work":undefined});
local.before=projection(await f.readThread(ts));signal.throwIfAborted();return {...receipt,runId:ready[0]!.id,providerCalls:f.provider.receipts.length-start};
},
async dispatchSlackQuery(ctx,_state,{query,signal}){
if(query!==ctx.scenario.query)throw new Error("Slack canonical bytes changed");const f=owned(ctx).fixture,start=f.provider.receipts.length;
f.relay.expectCanonical(query,f.threadId);
const result=await f.provider.call<{ok:true;channel:string;ts:string;message:{user:string;text:string;thread_ts:string}}>("sender","chat.postMessage",{channel:f.binding.prerequisites.channelId,thread_ts:f.threadId,text:query,unfurl_links:false,unfurl_media:false},signal);
if(result.channel!==f.binding.prerequisites.channelId||result.message.user!==f.binding.prerequisites.senderUserId||result.message.text!==query||result.message.thread_ts!==f.threadId)throw new Error("Actual Slack canonical publication did not bind exact human/thread/query");
await pollUntil({label:"genuine canonical signed Slack delivery capture",deadlineAt:Date.now()+30000,load:async()=>f.relay.receipt(),accept:row=>row.canonicalDeliveryCaptured});
const duplicate=await f.relay.replayCanonicalOnce();if(duplicate.status!==200)throw new Error("Actual signed duplicate was not acknowledged");
return f.resolveIngress(query,result.ts,f.provider.receipts.length-start);
},
async observe(ctx,state){
const o=await observeExternal(ctx,state),local=owned(ctx),f=local.fixture,source=state.slackIngressReceipt as ChatSlackIngressReceipt|undefined,other=state.slackOtherThreadSetupReceipt as (ChatSlackIngressReceipt&{runId:string})|undefined;
if(!source||!other)return {...o,thread:undefined};
const tasks=o.tasks.filter(row=>row.assigneeAgentId===(state.worker as Entity).id&&row.id!==source.issueId&&row.id!==other.issueId),target=tasks.length===1?tasks[0]:undefined;
// Provider egress reads wait for actual native settlement; no periodic Slack
// scan runs while the worker is building. Final/reload checks re-read bytes.
if(o.runs.length&&o.runs.every(run=>run.status==="succeeded")&&target?.status==="done"&&(!local.lastRead||Date.now()-local.lastRead>2000)){
local.lastThreads={after:projection(await f.readThread(local.otherRoot!)),messages:projection(await f.readThread(f.threadId))};local.lastRead=Date.now();
}
const attachments=target?await ctx.input.api.get<IssueAttachment[]>(`/api/issues/${target.id}/attachments`):[],products=target?await ctx.input.api.get<IssueWorkProduct[]>(`/api/issues/${target.id}/work-products`):[],files:ThreadEvidence["files"]=[];
for(const file of attachments.filter(row=>/\.html?$/i.test(row.originalFilename??"")).slice(0,10)){if(file.byteSize>262144)throw new Error("Slack launch result exceeds bounded artifact inspection");const bytes=await ctx.input.api.getAttachmentContent(file.id);const body=bytes.toString("utf8"),digest=hash(bytes);local.rendered??=new Map();let rendered=local.rendered.get(digest);if(!rendered){rendered=await renderSlackLaunchArtifact(ctx,body);local.rendered.set(digest,rendered);}files.push({id:file.id,hash:digest,body,rendered});}
const activity=await ctx.input.api.get<ChatActivityItem[]>(`/api/chat-endpoints/${f.endpoint.id}/activity`);
return {...o,thread:{source,other,before:local.before??[],after:local.lastThreads?.after??[],messages:local.lastThreads?.messages??[],duplicate:f.relay.receipt(),activity,attachments,products,files}} satisfies Observation;
},
async verifyResult(ctx,_state,value){
const o=value as Observation;await verifyRenderedReply(ctx,value);
const target=o.tasks.find(row=>row.assigneeAgentId===(o.state.worker as Entity).id&&row.status==="done"),href=o.links.find(link=>{try{const url=new URL(link,o.origin);return url.origin===o.origin&&[target?.id,target?.identifier].some(id=>!!id&&url.pathname.endsWith(`/issues/${id}`));}catch{return false;}});
if(!target||!href)throw new Error("Actual source completion must link its built Morgan task");
await ctx.input.page.locator(`[id^="comment-"] a[href=${JSON.stringify(href)}]`).first().click();await expect(ctx.input.page.getByText(target.title!,{exact:true}).first()).toBeVisible();
const file=o.thread.attachments.find(row=>o.thread.files.some(actual=>actual.id===row.id&&actual.hash===row.sha256));if(!file?.contentPath)throw new Error("Actual launch result lacks a supported public content route");
let retrieved:Buffer|undefined;const page=await ctx.input.page.context().newPage();try{
const pending=page.waitForEvent("download",{timeout:5000}).catch(()=>undefined);
await page.goto(new URL(file.contentPath,o.origin).href,{waitUntil:"domcontentloaded"}).catch(error=>{if(!/download|ERR_ABORTED/i.test(String(error)))throw error;});
const download=await pending;if(download){const path=await download.path();if(path)retrieved=await readFile(path);if(!retrieved||hash(retrieved)!==file.sha256)throw new Error("Actual browser launch-page download differs from the registered bytes");}else{await expect(page.locator("h1")).toBeVisible();retrieved=await ctx.input.api.getAttachmentContent(file.id);if(hash(retrieved)!==file.sha256)throw new Error("Public rendered result changed its registered bytes");}
}finally{await page.close();}
if(!retrieved||!hasSlackLaunchRendering(await renderSlackLaunchArtifact(ctx,retrieved.toString("utf8"))))throw new Error("Actual publicly retrieved launch result is not visibly usable");
},
async cleanup(ctx){const local=privateFixtures.get(ctx);try{await local?.fixture.cleanup();}finally{privateFixtures.delete(ctx);}},
})];