mirror of
https://github.com/paperclipai/paperclip.git
synced 2026-11-13 15:38:22 +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>
169 lines
8.0 KiB
TypeScript
169 lines
8.0 KiB
TypeScript
import { expect, type Locator, type Page } from "@playwright/test";
|
|
|
|
/** Own a browser event and its initiating action in the same error boundary. */
|
|
export async function withBrowserEvent<T>(observe: () => Promise<T>, action: () => Promise<unknown>): Promise<T> {
|
|
// Attach both rejection handlers before the action can fail or the event can
|
|
// time out. A synchronous action failure must own the pending event too.
|
|
const observed = observe(), initiated = Promise.resolve().then(action);
|
|
try {
|
|
const [event] = await Promise.all([observed, initiated]);
|
|
return event;
|
|
} catch (error) {
|
|
// A failed observation must not let a late click mutate owned resources
|
|
// after normal cleanup has begun. Keep the first failure and settle both.
|
|
await Promise.allSettled([observed, initiated]);
|
|
throw error;
|
|
}
|
|
}
|
|
|
|
/** Rich editors interpret pasted Markdown; fill inserts literal paragraph text. */
|
|
export async function fillTaskPrompt(editor: Locator, prompt: string): Promise<void> {
|
|
if (!/^ {0,3}(?:`{3,}|~{3,})/m.test(prompt)
|
|
|| await editor.evaluate(element => element.tagName === "TEXTAREA")) {
|
|
await editor.fill(prompt);
|
|
return;
|
|
}
|
|
await editor.fill("");
|
|
await editor.focus();
|
|
await editor.evaluate((element, text) => {
|
|
const clipboardData = new DataTransfer();
|
|
clipboardData.setData("text/plain", text);
|
|
const event = new ClipboardEvent("paste", { clipboardData, bubbles: true, cancelable: true });
|
|
element.dispatchEvent(event);
|
|
if (!event.defaultPrevented) throw new Error("Task editor did not accept the Markdown paste");
|
|
}, prompt);
|
|
}
|
|
|
|
export async function createTaskThroughUi(input: {
|
|
page: Page;
|
|
origin?: string;
|
|
issuePrefix: string;
|
|
agentName: string;
|
|
title: string;
|
|
prompt: string;
|
|
workMode: "standard" | "planning" | "ask";
|
|
projectName?: string;
|
|
requireExplicitTitle?: boolean;
|
|
attachments?: readonly string[];
|
|
beforeDispatch?: () => void | Promise<void>;
|
|
}) {
|
|
// Search's public creation action exposes the explicit title field. Strict
|
|
// native-operation fixtures use it so automatic task naming is not an extra
|
|
// provider mutation before the operation whose permission is under test.
|
|
const issuesUrl = input.requireExplicitTitle
|
|
? `/${encodeURIComponent(input.issuePrefix)}/search?scope=issues&q=${encodeURIComponent(`"${input.title}"`)}`
|
|
: `/${encodeURIComponent(input.issuePrefix)}/issues`;
|
|
const newTask = input.requireExplicitTitle
|
|
? input.page.getByRole("button", { name: "Create task from this query", exact: true })
|
|
: input.page.getByRole("button", { name: /^New task$/i }).first();
|
|
let bootstrapError: unknown;
|
|
for (let bootstrapAttempt = 1; bootstrapAttempt <= 3; bootstrapAttempt += 1) {
|
|
try {
|
|
await input.page.goto(input.origin ? new URL(issuesUrl, input.origin).href : issuesUrl, {
|
|
waitUntil: "domcontentloaded",
|
|
timeout: 30_000,
|
|
});
|
|
await newTask.waitFor({ state: "visible", timeout: 20_000 });
|
|
bootstrapError = undefined;
|
|
break;
|
|
} catch (error) {
|
|
bootstrapError = error;
|
|
if (bootstrapAttempt < 3) await input.page.waitForTimeout(1_000);
|
|
}
|
|
}
|
|
if (bootstrapError) {
|
|
throw new Error(
|
|
`Browser bootstrap failed before task creation: ${bootstrapError instanceof Error ? bootstrapError.message : String(bootstrapError)}`,
|
|
{ cause: bootstrapError },
|
|
);
|
|
}
|
|
await newTask.click();
|
|
const dialog = input.page.getByRole("dialog", { name: "New task", exact: true });
|
|
await expect(dialog).toBeVisible();
|
|
const titleInput = dialog.getByRole("textbox", { name: "Task title", exact: true });
|
|
if (input.requireExplicitTitle) {
|
|
await expect(titleInput).toBeVisible();
|
|
await titleInput.fill(input.title);
|
|
} else if (await titleInput.isVisible()) await titleInput.fill(input.title);
|
|
await fillTaskPrompt(dialog.getByRole("textbox", { name: "editable markdown", exact: true }), input.prompt);
|
|
if (input.workMode !== "standard") {
|
|
await dialog.getByRole("button", { name: "Add to composer", exact: true }).click();
|
|
await input.page.getByTestId(input.workMode === "planning" ? "composer-add-plan" : "composer-add-ask").click();
|
|
}
|
|
await dialog
|
|
.getByRole("button", { name: "Select assignee", exact: true })
|
|
.click();
|
|
await input.page.getByRole("searchbox", { name: "Search assignees", exact: true }).fill(input.agentName);
|
|
await input.page.getByRole("option").filter({ has: input.page.getByText(input.agentName, { exact: true }) }).click();
|
|
await expect(input.page.getByRole("searchbox", { name: "Search assignees", exact: true })).toBeHidden();
|
|
if (input.projectName) {
|
|
// Selecting the assignee advances focus to this selector and opens it.
|
|
// Focus is idempotent here; clicking would toggle an already-open popover
|
|
// closed before the search field can be filled.
|
|
const projectTrigger = dialog.getByRole("group", { name: "Task project and worktrees", exact: true }).getByRole("button").first();
|
|
await projectTrigger.focus();
|
|
const projectSearch = input.page.getByPlaceholder(/^Search projects(?:…|\.\.\.)?$/i);
|
|
await expect(projectSearch).toBeVisible({ timeout: 2_000 }).catch(async () => {
|
|
if (await projectTrigger.getAttribute("aria-expanded") !== "true") await projectTrigger.click();
|
|
await expect(projectSearch).toBeVisible();
|
|
});
|
|
await projectSearch.fill(input.projectName);
|
|
await input.page.getByText(input.projectName, { exact: true }).last().click();
|
|
}
|
|
if (input.attachments?.length) {
|
|
const chooser = await withBrowserEvent(() => input.page.waitForEvent("filechooser"), async () => {
|
|
const upload = dialog.getByRole("button", { name: "Upload", exact: true });
|
|
if (await upload.count()) await upload.click();
|
|
else {
|
|
await dialog.getByRole("button", { name: "Add to composer", exact: true }).click();
|
|
await input.page.getByRole("menuitem", { name: /^Files and images/ }).click();
|
|
}
|
|
});
|
|
await chooser.setFiles([...input.attachments]);
|
|
// Upload finishes as part of task creation; the dialog retains the selected files.
|
|
}
|
|
// The prompt-only composer generates a provisional title which the provider
|
|
// can rename immediately. Bind the fixture to the actual creation response.
|
|
await input.beforeDispatch?.();
|
|
const submittedAtMs = Date.now();
|
|
const [response] = await Promise.all([
|
|
input.page.waitForResponse(response => response.request().method() === "POST"
|
|
&& /^\/api\/companies\/[^/]+\/issues$/.test(new URL(response.url()).pathname)),
|
|
dialog.getByRole("button", { name: "Create task", exact: true }).click(),
|
|
]);
|
|
expect(response.status()).toBe(201);
|
|
const issue = await response.json() as { id: string; companyId: string; title: string; titleNeedsGeneration: boolean };
|
|
expect(new URL(response.url()).pathname).toBe(`/api/companies/${issue.companyId}/issues`);
|
|
expect(issue.id).toEqual(expect.any(String));
|
|
expect(issue.id.length).toBeGreaterThan(0);
|
|
if (input.requireExplicitTitle) {
|
|
expect(issue.title).toBe(input.title);
|
|
expect(issue.titleNeedsGeneration).toBe(false);
|
|
}
|
|
// Creation precedes uploads; allow them to finish before navigation.
|
|
await expect(dialog).not.toBeVisible({ timeout: 30_000 });
|
|
return { submittedAtMs, issueId: issue.id };
|
|
}
|
|
|
|
export async function submitTaskReply(
|
|
page: Page,
|
|
body: string,
|
|
remainingBudget?: () => number,
|
|
): Promise<number> {
|
|
const composer = page.getByTestId("task-chat-composer-input").last();
|
|
const options = () => {
|
|
const remaining = remainingBudget!();
|
|
if (!Number.isFinite(remaining) || remaining <= 0) throw new Error("Task reply original observation budget exhausted");
|
|
return { timeout: Math.min(30_000, remaining) };
|
|
};
|
|
await expect(composer).toBeVisible(remainingBudget ? options() : { timeout: 30_000 });
|
|
const editor = composer.locator('[contenteditable="true"], textarea').first();
|
|
if (remainingBudget) await editor.fill(body, options());
|
|
else await editor.fill(body);
|
|
const submittedAtMs = Date.now();
|
|
const send = page.getByTestId("task-chat-composer-send").last();
|
|
if (remainingBudget) await send.click(options());
|
|
else await send.click();
|
|
return submittedAtMs;
|
|
}
|