mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-13 07:12:00 +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>
99 lines
8.7 KiB
TypeScript
99 lines
8.7 KiB
TypeScript
import { createHash } from "node:crypto";
|
|
import type { Locator, Page } from "@playwright/test";
|
|
import type { AskUserQuestionsInteraction } from "../../packages/shared/src/types/issue.js";
|
|
import type { ChatCatalogContext } from "./chat-catalog-contract.js";
|
|
import type { InternalObservation, Entity } from "./chat-catalog-domains/internal-support.js";
|
|
|
|
interface QuestionInput {
|
|
origin: string; pageUrl: string; companyPrefix: string; sourceId: string; leadId: string; query: string;
|
|
observation: InternalObservation & { runs: Array<InternalObservation["runs"][number] & { finishedAt?: string | null }> };
|
|
comments: Entity[]; interactions: AskUserQuestionsInteraction[]; tasks: Entity[];
|
|
}
|
|
export function linkedQuestionBinding(input: QuestionInput) {
|
|
const o = input.observation, run = o.runs[0];
|
|
if (o.runs.length !== 1 || !run || run.companyId !== o.companyId || run.agentId !== input.leadId || run.runtimeMode !== "native" || run.status !== "succeeded"
|
|
|| run.contextSnapshot?.issueId !== input.sourceId || input.tasks.length !== 2 || new Set(input.tasks.map(task => task.id)).size !== 2
|
|
|| input.tasks.some(task => task.companyId !== o.companyId || !o.tasks.some(saved => saved.id === task.id && saved.companyId === task.companyId))) return;
|
|
const humans = input.comments.filter(comment => comment.companyId === o.companyId && comment.issueId === input.sourceId && !comment.authorAgentId
|
|
&& !!comment.authorUserId && comment.body === input.query);
|
|
if (humans.length !== 1 || o.humanQueries.length !== 1 || o.humanQueries[0] !== input.query) return;
|
|
const human = humans[0]!, start = Date.parse(run.startedAt ?? ""), end = Date.parse(run.finishedAt ?? "");
|
|
if (!Number.isFinite(start) || !Number.isFinite(end)) return;
|
|
const candidates = input.interactions.filter(card => card.kind === "ask_user_questions" && card.status === "pending" && !card.dismissedAt && !card.resolvedAt
|
|
&& card.companyId === o.companyId && card.issueId === input.sourceId && card.createdByAgentId === input.leadId && !card.createdByUserId
|
|
&& card.sourceRunId === run.id && !card.addresseeAgentId && (!card.addresseeUserId || card.addresseeUserId === human.authorUserId)
|
|
&& ["anyone", "human_only", "not_creator"].includes(card.effectiveResolverPolicy)
|
|
&& (card.originCommentIds?.includes(human.id) || card.sourceCommentId === human.id)
|
|
&& (card.originCommentIds ?? []).every(id => input.comments.some(comment => comment.id === id && comment.companyId === o.companyId && comment.issueId === input.sourceId))
|
|
&& (!card.sourceCommentId || card.sourceCommentId === human.id)
|
|
&& Date.parse(String(card.createdAt)) >= start && Date.parse(String(card.createdAt)) <= end
|
|
&& o.activity.some(event => event.companyId === o.companyId && event.action === "issue.thread_interaction_created"
|
|
&& (event as Entity & { entityType?: string }).entityType === "issue" && event.entityId === input.sourceId
|
|
&& event.actorType === "agent" && event.actorId === input.leadId && event.agentId === input.leadId
|
|
&& (event as Entity & { runId?: string }).runId === run.id && event.details?.interactionId === card.id && event.details?.interactionKind === "ask_user_questions"));
|
|
if (candidates.length !== 1) return;
|
|
const card = candidates[0]!, set = card.payload.questionSet, question = set?.questions[0];
|
|
if (set?.schema !== "paperclip.question_set.v1" || !set.description || !question?.id || !question.prompt) return;
|
|
return { origin: input.origin, pageUrl: input.pageUrl, companyPrefix: input.companyPrefix, tasks: input.tasks.map(task => ({ id: task.id, identifier: task.identifier })), sourceId: input.sourceId, anchorCommentId: human.id, interactionId: card.id, questionId: question.id,
|
|
prompt: question.prompt, sourceRunId: run.id, leadId: input.leadId,
|
|
payloadDigest: createHash("sha256").update(JSON.stringify(card.payload)).digest("hex") };
|
|
}
|
|
export function readLinkedQuestionDom(input: NonNullable<ReturnType<typeof linkedQuestionBinding>>) {
|
|
const visible = (element: Element) => {
|
|
for (let node: Element | null = element; node; node = node.parentElement) {
|
|
if (node.hasAttribute("hidden") || node.hasAttribute("inert") || node.getAttribute("aria-hidden") === "true") return false;
|
|
const style = window.getComputedStyle(node);
|
|
if (style.display === "none" || ["hidden", "collapse"].includes(style.visibility) || style.opacity === "0") return false;
|
|
}
|
|
return element.getClientRects().length > 0;
|
|
};
|
|
const selector = '[data-testid="task-chat-thread"]', clean = (value: string) => value.replace(/\s+/g, " ").trim();
|
|
const roots = Array.from(document.querySelectorAll(selector)).filter(root => visible(root) && Array.from(root.querySelectorAll("[id]")).some(node =>
|
|
node.id === `comment-${input.anchorCommentId}` && node.closest(selector) === root));
|
|
if (roots.length !== 1) return;
|
|
const root = roots[0]!, same = (node: Element) => node.closest(selector) === root;
|
|
const button = Array.from(root.querySelectorAll('[data-testid="task-chat-unanswered-question"]')).filter(node =>
|
|
node.id === `interaction-${input.interactionId}` && same(node) && visible(node) && node.getAttribute("aria-label") === `Answer question: ${input.prompt}`);
|
|
if (button.length !== 1) return;
|
|
const takeovers = Array.from(root.querySelectorAll('[data-testid="task-chat-composer-takeover"]')).filter(node => visible(node) && same(node));
|
|
if (takeovers.length !== 1) return;
|
|
const takeover = takeovers[0]!, prompts = Array.from(takeover.querySelectorAll("[id]")).filter(node => node.id === `${input.interactionId}-${input.questionId}-prompt`
|
|
&& visible(node) && same(node) && clean(node.textContent ?? "") === clean(input.prompt));
|
|
const contexts = Array.from(takeover.querySelectorAll('[role="region"][aria-label="Question context"]')).filter(node => visible(node) && same(node));
|
|
if (prompts.length !== 1 || contexts.length !== 1) return;
|
|
const links = Array.from(contexts[0]!.querySelectorAll("a[href]")).filter(node => visible(node) && same(node)).map(node => node.getAttribute("href")!);
|
|
const qualified = links.flatMap(href => {
|
|
try {
|
|
const url = new URL(href, input.origin), page = new URL(input.pageUrl);
|
|
const parts = url.pathname.split("/").filter(Boolean).map(decodeURIComponent);
|
|
const unprefixed = parts.length === 2 && parts[0] === "issues";
|
|
if (url.origin !== new URL(input.origin).origin || url.search || url.hash
|
|
|| !(parts.length === 3 && parts[0] === input.companyPrefix && parts[1] === "issues" || unprefixed
|
|
&& page.origin === url.origin && page.pathname.split("/").filter(Boolean)[0] === input.companyPrefix)) return [];
|
|
const id = parts[unprefixed ? 1 : 2], task = input.tasks.find(task => [task.id, task.identifier].includes(id));
|
|
return task ? [{ href, taskId: task.id }] : [];
|
|
} catch { return []; }
|
|
});
|
|
if (new Set(qualified.map(link => link.taskId)).size !== 2) return;
|
|
return { ...input, links: [...new Set(qualified.map(link => link.href))], text: prompts[0]!.textContent ?? "" };
|
|
}
|
|
export type LinkedQuestionEvidence = NonNullable<ReturnType<typeof readLinkedQuestionDom>>;
|
|
export type ClarificationObservation = InternalObservation & { linkedQuestionEvidence?: LinkedQuestionEvidence };
|
|
export async function observeLinkedTaskQuestion(ctx: ChatCatalogContext, state: Record<string, unknown>, observation: InternalObservation): Promise<ClarificationObservation> {
|
|
const source = ctx.issue(), prefix = ctx.input.fixtures.company.issuePrefix;
|
|
if (!source || !prefix || ctx.scenario.id !== "tasks.ambiguous-task") return observation;
|
|
const [comments, interactions] = await Promise.all([
|
|
ctx.input.api.get<Entity[]>(`/api/issues/${source.id}/comments?order=asc`),
|
|
ctx.input.api.get<AskUserQuestionsInteraction[]>(`/api/issues/${source.id}/interactions`),
|
|
]);
|
|
const binding = linkedQuestionBinding({ origin: observation.origin, pageUrl: ctx.input.page.url(), companyPrefix: prefix, sourceId: source.id, leadId: observation.leadId,
|
|
query: ctx.scenario.query, observation, comments, interactions, tasks: [state.task as Entity, state.otherTask as Entity] });
|
|
if (!binding) return observation;
|
|
const evidence = await ctx.input.page.evaluate(readLinkedQuestionDom, binding);
|
|
return evidence ? { ...observation, reply: [observation.reply, evidence.text].filter(Boolean).join("\n"), links: [...new Set([...observation.links, ...evidence.links])], linkedQuestionEvidence: evidence } : observation;
|
|
}
|
|
export function linkedQuestionTakeover(page: Page, evidence: LinkedQuestionEvidence): Locator {
|
|
return page.getByTestId("task-chat-thread").filter({ has: page.locator(`[id=${JSON.stringify(`comment-${evidence.anchorCommentId}`)}]`) })
|
|
.getByTestId("task-chat-composer-takeover").filter({ has: page.locator(`[id=${JSON.stringify(`${evidence.interactionId}-${evidence.questionId}-prompt`)}]`) });
|
|
}
|