mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-14 17:22:00 +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>
59 lines
7.0 KiB
TypeScript
59 lines
7.0 KiB
TypeScript
import { execFile } from "node:child_process";
|
||
import { promisify } from "node:util";
|
||
import type { Agent } from "../../../packages/shared/src/types/agent.js";
|
||
import type { IssueThreadInteraction } from "../../../packages/shared/src/types/issue.js";
|
||
import { expect } from "@playwright/test";
|
||
import { checks } from "./internal-support.js";
|
||
import { waitForReply } from "./identity-support.js";
|
||
import { interactionSurface, runtimeDefinition, runtimeObservation, observeRuntime } from "./runtime-support.js";
|
||
import { runtimePrivateSpec, selectedRuntimeInputs } from "./runtime-private.js";
|
||
import { createOwnedDaytonaTarget, deleteOwnedDaytonaTarget, type OwnedDaytonaTarget } from "./runtime-daytona-target.js";
|
||
import { privateClaudeOperation } from "./runtime-claude-private.js";
|
||
export function cloudChoiceMatches(reply: string) {
|
||
return /Claude\s*(?:Code)?/i.test(reply) && /(?:cloud|hosted|remote)\s+(?:computer|machine|environment|execution|runtime)|running (?:in|on) (?:the )?cloud/i.test(reply)
|
||
&& /which|do you mean|did you mean|choose|prefer|clarif|whether/i.test(reply) && /engineer|hire|role|developer/i.test(reply) && reply.length < 1200
|
||
&& !/claude_local|paperclip_runner|adapterType|acpxAgent|supply.*(?:adapter|internal).*id/i.test(reply)
|
||
&& !/(?:already|I(?:['’]ve| have))\s+(?:created|hired|selected|assigned)|(?:cloud code)\s+(?:definitely|always|clearly)\s+means/i.test(reply);
|
||
}
|
||
export function gradeCloudChoice(value: unknown) {
|
||
const c = checks(), o = runtimeObservation(value); c.add("actual-two-available-context", !!o, "Both actually qualified possibilities and independent no-commit evidence are required."); if (!o) return c.result();
|
||
const s = o.state;
|
||
c.add("concrete-clarification-retains-role", cloudChoiceMatches(`${o.reply}\n${String(s.visibleChoiceText ?? "")}`) && s.questionVisible === true, "The actual human sees a short concrete Claude Code versus cloud-computer choice retaining the requested engineer role, with no internal adapter vocabulary.");
|
||
c.add("waits-before-consequential-commit", s.claudeAvailable === true && s.cloudAvailable === true && s.noNewEmployee === true && s.originalBindingsUnchanged === true && s.sourceWaiting === true
|
||
&& o.runs.length === 1 && o.runs[0]?.agentId === o.leadId && o.runs[0].status === "succeeded" && o.tasks.length === 0 && o.humanQueries.length === 1,
|
||
"Real available choices remain unselected: no employee, harness/computer binding or delegated task is committed before the person's answer.");
|
||
return c.result();
|
||
}
|
||
export const runtimeCloudChoiceCases = [runtimeDefinition("agents.ambiguous-cloud-code", async ctx => {
|
||
const target = await createOwnedDaytonaTarget(ctx);
|
||
try {
|
||
const selected = selectedRuntimeInputs(ctx, "agents.ambiguous-cloud-code"), companyId = ctx.input.fixtures.company.id;
|
||
ctx.input.privateSecrets!.push(selected.credentials.ANTHROPIC_API_KEY!);
|
||
const localAdapters = await ctx.input.api.get<Array<{ type: string; available?: boolean; supportedRunners?: string[] }>>(`/api/adapters?companyId=${companyId}&environmentId=${ctx.input.fixtures.environment.id}`);
|
||
if (!localAdapters.some(row => row.type === "claude_local" && row.supportedRunners?.includes("legacy"))) throw new Error("Setup blocker: actual local Claude Code availability is missing; a cloud-only target does not establish the ambiguity context");
|
||
const models = await ctx.input.api.get<Array<{ id: string }>>(`/api/companies/${companyId}/adapters/claude_local/models`);
|
||
if (!models.length || !models[0]?.id) throw new Error("Actual authorized Claude model discovery returned no available model");
|
||
const executable = (await promisify(execFile)("which", ["claude"], { timeout: 5000, maxBuffer: 4096 })).stdout.trim();
|
||
const account = await privateClaudeOperation({ operation: "prepare-account", companyId, environmentId: ctx.input.fixtures.environment.id, credential: selected.credentials.ANTHROPIC_API_KEY!, harnessExecutable: executable, model: models[0].id, baseURL: ctx.input.api.baseURL, cookies: await ctx.input.page.context().cookies(ctx.input.api.baseURL) });
|
||
if (account.probeStatus !== "pass") throw new Error("Actual selected Claude command/account/model probe did not establish Claude Code availability");
|
||
return { cloudTarget: target, cloudAvailable: true, claudeAvailable: true, beforeAgents: await ctx.input.api.get<Agent[]>(`/api/companies/${companyId}/agents`), providerProbes: [...target.providerProbes, { phase: "claude-availability", operation: account.operation, publicProbeCount: 2, status: account.probeStatus, usage: "unavailable", cost: "unavailable" }] };
|
||
} catch (error) { await deleteOwnedDaytonaTarget(ctx, target); throw error; }
|
||
}, gradeCloudChoice, {
|
||
...runtimePrivateSpec("agents.ambiguous-cloud-code"), authentication: "human-session", expectedRunCount: 1, attemptTimeoutMs: 900000,
|
||
sourceFiles: ["chat-catalog-domains/runtime-cloud-choice.ts", "chat-catalog-domains/runtime-daytona-target.ts", "chat-catalog-domains/runtime-private.ts", "chat-catalog-domains/runtime-claude-private.ts", "chat-catalog-domains/runtime-claude-private-worker.ts", "chat-catalog-domains/runtime-support.ts", "chat-catalog-domains/external-private-browser.ts", "chat-catalog-domains/external-private-browser-worker.ts", "live-fixtures.ts", "../../packages/plugins/sandbox-providers/daytona/src/plugin.ts", "chat-scenarios/agents.yaml"],
|
||
async act(ctx, state) {
|
||
await waitForReply(ctx); const source = ctx.issue()!, interactions = await ctx.input.api.get<IssueThreadInteraction[]>(`/api/issues/${source.id}/interactions`);
|
||
const questions = interactions.filter(row => row.kind === "ask_user_questions" && row.status === "pending");
|
||
if (questions.length) {
|
||
const runs = await ctx.allRuns();
|
||
if (questions.some(row => row.sourceRunId !== runs[0]?.id || row.issueId !== source.id)) throw new Error("Clarification card is not attributed to the actual canonical source run");
|
||
await expect(ctx.input.page.getByTestId("task-chat-unanswered-question").first()).toBeVisible();
|
||
state.visibleChoiceText = await interactionSurface(ctx.input.page, questions[0]!).first().innerText();
|
||
}
|
||
else await expect(ctx.input.page.getByTestId("task-chat-composer-input")).toBeVisible();
|
||
state.questionVisible = questions.length > 0 || /\?|which|do you mean/i.test((await observeRuntime(ctx, state)).reply); state.sourceWaiting = source.conversationState === "waiting";
|
||
},
|
||
async observe(ctx, state) { const o = await observeRuntime(ctx, state), prior = state.beforeAgents as Agent[]; state.noNewEmployee = o.agents.length === prior.length && prior.every(before => o.agents.some(row => row.id === before.id)); state.originalBindingsUnchanged = prior.every(before => { const after = o.agents.find(row => row.id === before.id); return after?.adapterType === before.adapterType && JSON.stringify(after.adapterConfig) === JSON.stringify(before.adapterConfig) && after.defaultEnvironmentId === before.defaultEnvironmentId; }); return o; },
|
||
async cleanup(ctx, state) { if (state.cloudTarget) await deleteOwnedDaytonaTarget(ctx, state.cloudTarget as OwnedDaytonaTarget); },
|
||
})];
|