mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-15 01:52:55 +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>
65 lines
10 KiB
TypeScript
65 lines
10 KiB
TypeScript
import { expect } from "@playwright/test";
|
|
import type { ToolProfileEffectiveSummary, ToolRunDecisionLookup, ToolConnection } from "../../../packages/shared/src/types/tool-access.js";
|
|
import { definition, root, checks, externalObservation, observeExternal, verifyRenderedReply } from "./external-support.js";
|
|
import { externalRequirements, selectedExternalBinding } from "./external-fixtures.js";
|
|
import { readNotionFixturePages, connectNamedNotion, type NotionPageOracle } from "./external-notion-provider.js";
|
|
import { notionFactSpan, notionDateMentioned, notionPageIdentityMatches } from "./external-notion-semantics.js";
|
|
import { identityWorker, waitForReply } from "./identity-support.js";
|
|
import { sendChatMessage } from "../chat-flow.js";
|
|
import type { ChatCatalogContext } from "../chat-catalog-contract.js";
|
|
async function decisions(ctx: ChatCatalogContext) { return Promise.all((await ctx.allRuns()).map(run => ctx.input.api.get<ToolRunDecisionLookup>(`${root(ctx)}/tools/runs/${run.id}/decisions`))); }
|
|
function successfulRead(rows: ToolRunDecisionLookup[], state: Record<string, unknown>, runId?: string) { return rows.some(row => (!runId || row.runId === runId) && row.decisions.some(decision => decision.invocation.runId === row.runId && decision.invocation.connectionId === state.connectionId && decision.invocation.agentId === state.ceoId && decision.invocation.issueId === state.sourceIssueId && (state.readEntryIds as string[]).includes(decision.invocation.catalogEntryId ?? "") && decision.invocation.status === "succeeded")); }
|
|
export function gradeNotionAssignment(value: unknown) {
|
|
const c = checks(), o = externalObservation(value); c.add("complete-named-notion-assignment-evidence", !!o, "Actual official Notion workspace, Morgan-only assignment, human grant and continued model read are required."); if (!o) return c.result();
|
|
c.add("usable-account-assigned-only-to-morgan", o.state.namedProvider === "notion" && o.state.workspaceMatched === true && o.state.healthyBefore === true && o.state.morganHadReadBefore === true && o.state.ceoHadReadBefore === false && o.state.morganId !== o.leadId, "Official provider reads and public effective access independently establish a usable account unavailable to this CEO.");
|
|
c.add("honest-assignment-handoff", o.state.beforeGrantReadCount === 0 && o.state.initialAccessExplanation === true && o.state.initialResearchClaim === false && o.state.humanManagementVisible === true, "The original phase identifies assignment as a prerequisite without borrowing Morgan's authority or inventing a summary.");
|
|
c.add("same-account-authorized-assignment", o.state.authorizedGrant === true && o.activity.some(row => row.entityId === o.state.connectionId && row.action === "tool_app.finished" && row.actorType === "user") && o.state.operatorChannel === "public-browser-agent-selection" && o.state.sameConnectionCount === true && o.state.morganRetainedRead === true && o.state.ceoHasReadAfter === true && o.state.fixturesUnchanged === true, "The authorized human adds this CEO to the existing connection without replacing Morgan or creating a duplicate account.");
|
|
const page = (o.state.pages as NotionPageOracle[] | undefined)?.[0], replies = o.phaseReplies.filter(row => row.runId !== o.state.initialRunId), body = replies.map(row => row.body).join("\n");
|
|
c.add("phase-bound-current-workspace-read", Array.isArray(o.state.decisions) && replies.some(reply => o.runs.some(run => run.id === reply.runId && run.status === "succeeded") && successfulRead(o.state.decisions as ToolRunDecisionLookup[], o.state, reply.runId))
|
|
&& !!page && !!notionFactSpan(body, page.fact) && notionDateMentioned(body, page.editedAt) && o.links.some(link => notionPageIdentityMatches(link, page.id)), "The resumed source task performs its own actual named-provider read and grounds its answer in the allowlisted current page body, date and exact source identity.");
|
|
return c.result();
|
|
}
|
|
export const notionAssignmentExternalCases = [definition("connection-setup.notion-agent-unassigned", async ctx => {
|
|
const binding = selectedExternalBinding(ctx), pages = await readNotionFixturePages(ctx, binding), morgan = await identityWorker(ctx);
|
|
const connection = await connectNamedNotion(ctx, binding, [morgan.id]);
|
|
const saved = await ctx.input.api.get<ToolConnection>(`/api/tool-connections/${connection.connectionId}`);
|
|
const effective = async (id: string) => ctx.input.api.get<ToolProfileEffectiveSummary>(`${root(ctx)}/tools/profiles/effective/agents/${id}`);
|
|
const morganAccess = await effective(morgan.id), ceoAccess = await effective(ctx.input.fixtures.agent.id);
|
|
const hasRead = (row: ToolProfileEffectiveSummary) => connection.readEntryIds.some(id => row.allowedTools.some(tool => tool.id === id));
|
|
if (saved.status !== "active" || !saved.enabled || saved.healthStatus !== "ok" || !hasRead(morganAccess) || hasRead(ceoAccess)) throw new Error("Setup blocker: actual official Notion connection must be healthy and assigned to Morgan only");
|
|
const count = (await ctx.input.api.get<{ connections: ToolConnection[] }>(`${root(ctx)}/tools/connections`)).connections.length;
|
|
await ctx.input.api.post(`${root(ctx)}/chats/${ctx.input.fixtures.agent.id}`); await ctx.refreshIssue();
|
|
await ctx.input.api.patch(`/api/issues/${ctx.issue()!.id}`, { description: `The intended Launch workspace is on the existing Launch Notion account (${connection.connectionId}). It is already usable by Morgan (${morgan.id}); the CEO is not assigned. The authorized human may assign the existing account to this CEO. Keep Morgan's access.` });
|
|
return { ...connection, pages, ceoId: ctx.input.fixtures.agent.id, morganId: morgan.id, namedProvider: "notion", workspaceMatched: true, healthyBefore: true, morganHadReadBefore: true, ceoHadReadBefore: false, connectionCountBefore: count, sourceIssueId: ctx.issue()!.id };
|
|
}, gradeNotionAssignment, {
|
|
authentication: "human-session", expectedRunCount: 2, expectedHumanMessageCount: 2, attemptTimeoutMs: 480_000,
|
|
externalInput: externalRequirements["connection-setup.notion-agent-unassigned"],
|
|
sourceFiles: ["chat-catalog-domains/external-notion-assignment.ts", "chat-catalog-domains/external-notion-provider.ts", "chat-catalog-domains/external-notion-semantics.ts", "chat-catalog-domains/external-fixtures.ts", "chat-catalog-domains/external-support.ts", "chat-catalog-domains/internal-support.ts", "chat-catalog-domains/identity-support.ts", "chat-flow.ts", "chat-scenarios/connection-setup.yaml"],
|
|
async act(ctx, state) {
|
|
await waitForReply(ctx); const o = await observeExternal(ctx, state); state.initialRunId = o.phaseReplies.at(-1)?.runId;
|
|
const rows = await decisions(ctx); state.beforeGrantReadCount = rows.reduce((count, row) => count + row.decisions.filter(decision => decision.invocation.connectionId === state.connectionId && decision.invocation.status === "succeeded").length, 0);
|
|
state.initialAccessExplanation = /assign|access|permission|not.*available|not.*connect/i.test(o.reply) && /notion|morgan/i.test(o.reply);
|
|
const page = (state.pages as NotionPageOracle[])[0]!; state.initialResearchClaim = !!notionFactSpan(o.reply, page.fact);
|
|
const prefix = `/${ctx.input.fixtures.company.issuePrefix}/apps/${state.connectionId}`;
|
|
const href = o.links.find(link => { try { const url = new URL(link, o.origin); return url.origin === o.origin && (url.pathname === prefix || url.pathname.startsWith(`${prefix}/`)); } catch { return false; } });
|
|
if (!href) throw new Error("Chat did not provide the existing account's supported access-management handoff");
|
|
const source = ctx.input.page.url(); await ctx.input.page.goto(new URL(href, o.origin).href, { waitUntil: "domcontentloaded" }); await expect(ctx.input.page.getByRole("main")).toContainText("Launch Notion"); state.humanManagementVisible = true;
|
|
if (!ctx.input.page.url().endsWith("/permissions")) await ctx.input.page.goto(`${prefix}/permissions`, {waitUntil:"domcontentloaded"});
|
|
await ctx.input.page.getByRole("button", {name:"1 agent selected",exact:true}).click();
|
|
const completed = ctx.input.page.waitForResponse(response => new URL(response.url()).pathname === `${root(ctx)}/tools/apps/${state.connectionId}/finish` && response.request().method() === "POST" && response.ok());
|
|
await ctx.input.page.getByRole("checkbox", {name:`Allow ${ctx.input.fixtures.agent.name}`,exact:true}).check();
|
|
const response = await completed, request = response.request().postDataJSON() as {access:{agentIds?:string[]}};
|
|
state.authorizedGrant = Array.isArray(request.access.agentIds) && request.access.agentIds.length === 2 && request.access.agentIds.includes(String(state.morganId)) && request.access.agentIds.includes(String(state.ceoId)); state.operatorChannel = "public-browser-agent-selection";
|
|
await ctx.input.page.goto(source, { waitUntil: "domcontentloaded" }); await sendChatMessage(ctx.input.page, "I assigned the existing Notion account to you and kept Morgan's access. Please continue the original launch summary here.");
|
|
},
|
|
async observe(ctx, state) {
|
|
const pages = await readNotionFixturePages(ctx), before = state.pages as NotionPageOracle[];
|
|
state.fixturesUnchanged = pages.length === before.length && pages.every((page, i) => page.id === before[i]?.id && page.editedAt === before[i]?.editedAt && JSON.stringify(page.fact) === JSON.stringify(before[i]?.fact));
|
|
const connections = await ctx.input.api.get<{ connections: ToolConnection[] }>(`${root(ctx)}/tools/connections`); state.sameConnectionCount = connections.connections.length === state.connectionCountBefore && connections.connections.some(row => row.id === state.connectionId);
|
|
for (const [id, field] of [[state.ceoId, "ceoHasReadAfter"], [state.morganId, "morganRetainedRead"]] as const) { const effective = await ctx.input.api.get<ToolProfileEffectiveSummary>(`${root(ctx)}/tools/profiles/effective/agents/${id}`); state[field] = (state.readEntryIds as string[]).some(read => effective.allowedTools.some(tool => tool.id === read)); }
|
|
state.decisions = await decisions(ctx); return observeExternal(ctx, state);
|
|
},
|
|
async verifyResult(ctx, state, value) { await verifyRenderedReply(ctx, value, String(state.initialRunId)); },
|
|
async cleanup(ctx, state) { if (typeof state.connectionId === "string") await ctx.input.api.delete(`/api/tool-connections/${state.connectionId}`, { allowNotFound: true }); },
|
|
})];
|