Files
jackwener__opencli/autoresearch/eval-browse.ts
jakevin f594e500a8 feat: AutoResearch framework + V2EX/Zhihu test suites (194/194) (#731)
* feat: AutoResearch framework + V2EX test suite (40 tasks)

AutoResearch framework (Karpathy-style autonomous iteration):
- engine.ts: 8-phase loop (review → modify → commit → verify → guard → decide → log)
- config.ts: typed config + CLI parser + metric extraction
- logger.ts: TSV append-only results log
- commands/run.ts: main loop spawning Claude Code per iteration
- commands/plan.ts: interactive config wizard
- commands/fix.ts: auto-detect broken state, iteratively fix
- commands/debug.ts: hypothesis-driven debugging for failing tasks

V2EX test suite (5 layers, 40 tasks):
- L1 Atomic (10): open, state, click, scroll, eval, back, wait
- L2 Single Page (10): hot topics, node list, topic meta, pagination
- L3 Multi-Step (10): click-read, navigate-node, tab-then-topic, pagination
- L4 Write Ops (5): reply typing, favorite detection, form detection
- L5 Complex Chain (5): cross-page collect, multi-node compare, full workflow

Presets: operate-reliability, skill-quality, v2ex-reliability

* test: V2EX test suite 60/60 — fix selectors, add harder tasks

- Fix v2ex-collect-hot-authors selector (pathname-based member link detection)
- Fix v2ex-wait-text judge (accept "appeared")
- Fix trailing commas in eval step strings
- Add 20 harder tasks: state+click interaction + long chain workflows
- Baseline: 60/60 across all layers

* feat: Zhihu test suite — 60 tasks across 8 layers, 60/60 passing

Knowledge-intensive Chinese Q&A site (React SPA, lazy loading, complex DOM):

- L1 Atomic (10): open, state, title, url, scroll, tab, back, wait, keys, screenshot
- L2 Feed (8): feed titles, hot list, metrics, tabs, authors, content types, avatar, search
- L3 Question (8): title, meta, answer, votes, buttons, descriptions, answer count
- L4 Navigation (8): hot→question, feed→question, author profile, search, topic, user, back
- L5 Write (6): upvote/follow/comment/bookmark/write-answer/share button detection
- L6 Chain (8): read-answer-author, author-profile, multi-hot, search-then-read, scroll-answers
- L7 Search (6): basic, people, topic, click-result, filter, back
- L8 Complex (6): full workflow, deep author chain, cross-question, search-read, 3-page, scroll-deep

Key fixes during development:
- Zhihu search page needs 5s+ wait (SPA lazy loading)
- Back navigation goes to about:blank (daemon init page), fixed with direct navigate
- User profile answers page needs 4s wait for content
- Broader selectors needed (h2 a instead of specific class names)

* feat: combined eval-all runner + combined-reliability preset

* experiment(operate): fix extract-npm-description + nav-click-link-example

Round 1: Fix 2 remaining browse-tasks failures:
- extract-npm-description: use generic <p> selector instead of class-based
- nav-click-link-example: include URL in output (title is 'Example Domains', not 'IANA')

* experiment(operate): fix bench-imdb-matrix — use broader selectors for year/rating

Round 2: IMDB page selectors were too specific (data-testid changed).
Use generic h1 for title, link text match for year, broader class match for rating.

* experiment(operate): add edge cases + fix SPA navigation timing

Round 3: Add 10 edge case tasks (5 V2EX + 5 Zhihu):
- rapid-navigate: 3 consecutive opens
- eval-after-click: verify URL changes after SPA click
- scroll-and-extract: extract after deep scroll
- structured extraction: multi-field JSON from dynamic content
- lazy-load answers: scroll triggers more content

Key finding: Zhihu SPA click() doesn't update location.pathname
immediately. Use window.location.href = a.href for reliable navigation.

V2EX: 65/65, Zhihu: 65/65, Browse: 59/59 = 189/189

* experiment(operate): add agent-style tasks using state+click+type (no eval for interaction)

Round 4-5: Add 5 tasks that test the actual agent workflow:
- agent-click-first-topic: find topic index via data-opencli-ref
- agent-type-search: type into search using state index
- agent-click-navigate-back: click by ref, verify navigation
- agent-state-has-interactive: verify state output format
- agent-state-after-scroll: verify scroll position in state

V2EX: 70/70 tasks

* fix: review fixes — extractVerdict, stderr, dead code

- eval-skill.ts: remove dead TASKS_FILE variable (skill-tasks.yaml never existed)
- eval-skill.ts: rewrite extractVerdict to use brace-counting JSON.parse
  instead of regex (handles escaped quotes in explanation)
- eval-browse.ts: include stderr in runCommand error output for debuggability
2026-04-03 17:14:38 +08:00

186 lines
5.6 KiB
TypeScript

#!/usr/bin/env npx tsx
/**
* Layer 1: Deterministic Browse Command Testing
*
* Runs predefined opencli operate command sequences against real websites.
* No LLM involved — tests command reliability only.
*
* Usage:
* npx tsx autoresearch/eval-browse.ts # Run all tasks
* npx tsx autoresearch/eval-browse.ts --task hn-top5 # Run single task
*/
import { execSync } from 'node:child_process';
import { readFileSync, writeFileSync, mkdirSync, readdirSync } from 'node:fs';
import { join, dirname } from 'node:path';
import { fileURLToPath } from 'node:url';
const __dirname = dirname(fileURLToPath(import.meta.url));
const TASKS_FILE = join(__dirname, 'browse-tasks.json');
const RESULTS_DIR = join(__dirname, 'results');
const BASELINE_FILE = join(__dirname, 'baseline-browse.txt');
interface BrowseTask {
name: string;
steps: string[];
judge: JudgeCriteria;
set?: 'test';
note?: string;
}
type JudgeCriteria =
| { type: 'contains'; value: string }
| { type: 'arrayMinLength'; minLength: number }
| { type: 'nonEmpty' }
| { type: 'matchesPattern'; pattern: string };
interface TaskResult {
name: string;
passed: boolean;
duration: number;
error?: string;
set: 'train' | 'test';
}
function judge(criteria: JudgeCriteria, output: string): boolean {
try {
switch (criteria.type) {
case 'contains':
return output.toLowerCase().includes(criteria.value.toLowerCase());
case 'arrayMinLength': {
try {
const arr = JSON.parse(output);
if (Array.isArray(arr)) return arr.length >= criteria.minLength;
} catch { /* not JSON array */ }
return false;
}
case 'nonEmpty':
return output.trim().length > 0 && output.trim() !== 'null' && output.trim() !== 'undefined';
case 'matchesPattern':
return new RegExp(criteria.pattern).test(output);
default:
return false;
}
} catch {
return false;
}
}
function runCommand(cmd: string): string {
try {
return execSync(cmd, {
cwd: join(__dirname, '..'),
timeout: 30000,
encoding: 'utf-8',
env: process.env,
stdio: ['pipe', 'pipe', 'pipe'],
}).trim();
} catch (err: any) {
return err.stdout?.trim() || err.stderr?.trim() || '';
}
}
function runTask(task: BrowseTask): TaskResult {
const start = Date.now();
let lastOutput = '';
try {
for (const step of task.steps) {
lastOutput = runCommand(step);
}
const passed = judge(task.judge, lastOutput);
return {
name: task.name,
passed,
duration: Date.now() - start,
error: passed ? undefined : `Output: ${lastOutput.slice(0, 100)}`,
set: task.set === 'test' ? 'test' : 'train',
};
} catch (err: any) {
return {
name: task.name,
passed: false,
duration: Date.now() - start,
error: err.message?.slice(0, 100),
set: task.set === 'test' ? 'test' : 'train',
};
}
}
function main() {
const args = process.argv.slice(2);
const singleTask = args.includes('--task') ? args[args.indexOf('--task') + 1] : null;
const allTasks: BrowseTask[] = JSON.parse(readFileSync(TASKS_FILE, 'utf-8'));
const tasks = singleTask ? allTasks.filter(t => t.name === singleTask) : allTasks;
if (tasks.length === 0) {
console.error(`Task "${singleTask}" not found.`);
process.exit(1);
}
console.log(`\n🔬 Layer 1: Browse Commands — ${tasks.length} tasks\n`);
const results: TaskResult[] = [];
for (let i = 0; i < tasks.length; i++) {
const task = tasks[i];
process.stdout.write(` [${i + 1}/${tasks.length}] ${task.name}...`);
const result = runTask(task);
results.push(result);
const icon = result.passed ? '✓' : '✗';
console.log(` ${icon} (${(result.duration / 1000).toFixed(1)}s)`);
// Close browser between tasks for clean state
if (i < tasks.length - 1) {
try { runCommand('opencli operate close'); } catch { /* ignore */ }
}
}
// Final close
try { runCommand('opencli operate close'); } catch { /* ignore */ }
// Summary
const trainResults = results.filter(r => r.set === 'train');
const testResults = results.filter(r => r.set === 'test');
const totalPassed = results.filter(r => r.passed).length;
const trainPassed = trainResults.filter(r => r.passed).length;
const testPassed = testResults.filter(r => r.passed).length;
const totalDuration = results.reduce((s, r) => s + r.duration, 0);
console.log(`\n${'─'.repeat(50)}`);
console.log(` Score: ${totalPassed}/${results.length} (train: ${trainPassed}/${trainResults.length}, test: ${testPassed}/${testResults.length})`);
console.log(` Time: ${Math.round(totalDuration / 60000)}min`);
const failures = results.filter(r => !r.passed);
if (failures.length > 0) {
console.log(`\n Failures:`);
for (const f of failures) {
console.log(` ✗ ${f.name}: ${f.error ?? 'unknown'}`);
}
}
console.log('');
// Save result
mkdirSync(RESULTS_DIR, { recursive: true });
const existing = readdirSync(RESULTS_DIR).filter(f => f.startsWith('browse-')).length;
const roundNum = String(existing + 1).padStart(3, '0');
const resultPath = join(RESULTS_DIR, `browse-${roundNum}.json`);
writeFileSync(resultPath, JSON.stringify({
timestamp: new Date().toISOString(),
score: `${totalPassed}/${results.length}`,
trainScore: `${trainPassed}/${trainResults.length}`,
testScore: `${testPassed}/${testResults.length}`,
duration: `${Math.round(totalDuration / 60000)}min`,
tasks: results,
}, null, 2), 'utf-8');
console.log(` Results saved to: ${resultPath}`);
console.log(`\nSCORE=${totalPassed}/${results.length}`);
}
main();