@matware/e2e-runner 1.0.2 → 1.1.0

package/README.md

@@ -128,6 +128,14 @@ npx e2e-runner pool start                 # Start Chrome container
 npx e2e-runner pool stop                  # Stop Chrome container
 npx e2e-runner pool status                # Check pool health
 
+# Issue-to-test
+npx e2e-runner issue <url>                # Fetch issue details
+npx e2e-runner issue <url> --generate     # Generate test file via AI
+npx e2e-runner issue <url> --verify       # Generate + run + report
+
+# Dashboard
+npx e2e-runner dashboard                  # Start web dashboard
+
 # Other
 npx e2e-runner list                       # List available suites
 npx e2e-runner init                       # Scaffold project
@@ -150,6 +158,7 @@ npx e2e-runner init                       # Scaffold project
 | `--env <name>` | `default` | Environment profile |
 | `--pool-port <port>` | `3333` | Chrome pool port |
 | `--max-sessions <n>` | `10` | Max concurrent Chrome sessions |
+| `--project-name <name>` | dir name | Project display name for dashboard |
 
 ## Configuration
 
@@ -201,6 +210,9 @@ When `--env <name>` is set, the matching profile from `environments` overrides e
 | `TEST_TIMEOUT` | `testTimeout` |
 | `OUTPUT_FORMAT` | `outputFormat` |
 | `E2E_ENV` | `env` |
+| `PROJECT_NAME` | `projectName` |
+| `ANTHROPIC_API_KEY` | `anthropicApiKey` |
+| `ANTHROPIC_MODEL` | `anthropicModel` |
 
 ## Hooks
 
@@ -342,12 +354,14 @@ claude mcp add --transport stdio --scope user e2e-runner \
 | `e2e_list` | List available test suites with test names and counts |
 | `e2e_create_test` | Create a new test JSON file |
 | `e2e_pool_status` | Check Chrome pool availability and capacity |
-| `e2e_pool_start` | Start the Chrome pool Docker container |
-| `e2e_pool_stop` | Stop the Chrome pool |
+| `e2e_screenshot` | Retrieve a screenshot by its hash (e.g. `ss:a3f2b1c9`) |
+| `e2e_issue` | Fetch a GitHub/GitLab issue and generate E2E tests |
+
+> **Note:** Pool start/stop are only available via CLI (`e2e-runner pool start|stop`), not via MCP — restarting the pool kills all active sessions from other clients.
 
 All tools accept an optional `cwd` parameter (absolute path to the project root). Claude Code passes its current working directory so the MCP server resolves `e2e/tests/`, `e2e.config.js`, and `.e2e-pool/` relative to the correct project — even when switching between multiple projects in the same session.
 
-Once installed, Claude Code can run tests, analyze failures, create new test files, and manage the Chrome pool as part of its normal workflow. Just ask:
+Once installed, Claude Code can run tests, analyze failures, and create new test files as part of its normal workflow. Just ask:
 
 > "Run all E2E tests"
 > "Create a test that verifies the checkout flow"
@@ -360,6 +374,70 @@ claude mcp list
 # e2e-runner: ... - Connected
 ```
 
+## Issue-to-Test
+
+Turn GitHub and GitLab issues into executable E2E tests. Paste an issue URL and get runnable tests -- automatically.
+
+### How It Works
+
+1. **Fetch** -- Pulls issue details (title, body, labels) via `gh` or `glab` CLI
+2. **Generate** -- AI creates JSON test actions based on the issue description
+3. **Run** -- Optionally executes the tests immediately to verify if a bug is reproducible
+
+### Two Modes
+
+**Prompt mode** (default, no API key): Returns issue data + a structured prompt. Claude Code uses its own intelligence to create tests via `e2e_create_test` and run them.
+
+**Verify mode** (requires `ANTHROPIC_API_KEY`): Calls Claude API directly, generates tests, runs them, and reports whether the bug is confirmed or not reproducible.
+
+### CLI
+
+```bash
+# Fetch and display issue details
+e2e-runner issue https://github.com/owner/repo/issues/42
+
+# Generate a test file via Claude API
+e2e-runner issue https://github.com/owner/repo/issues/42 --generate
+# -> Creates e2e/tests/issue-42.json
+
+# Generate + run + report bug status
+e2e-runner issue https://github.com/owner/repo/issues/42 --verify
+# -> "BUG CONFIRMED" or "NOT REPRODUCIBLE"
+
+# Output AI prompt as JSON (for piping)
+e2e-runner issue https://github.com/owner/repo/issues/42 --prompt
+```
+
+### MCP
+
+In Claude Code, the `e2e_issue` tool handles everything:
+
+> "Fetch issue https://github.com/owner/repo/issues/42 and create E2E tests for it"
+
+Claude Code receives the issue data, generates appropriate test actions, saves them via `e2e_create_test`, and runs them with `e2e_run`.
+
+### Auth Requirements
+
+- **GitHub**: `gh` CLI authenticated (`gh auth login`)
+- **GitLab**: `glab` CLI authenticated (`glab auth login`)
+
+Provider is auto-detected from the URL. Self-hosted GitLab is supported via `glab` config.
+
+### Bug Verification Logic
+
+Generated tests assert the **correct** behavior. If the tests fail, the correct behavior doesn't work -- bug confirmed. If all tests pass, the bug is not reproducible.
+
+## Web Dashboard
+
+Real-time UI for running tests, viewing results, screenshots, and run history.
+
+```bash
+e2e-runner dashboard                  # Start on default port 8484
+e2e-runner dashboard --port 9090      # Custom port
+```
+
+Features: live test execution, screenshot viewer with copy-to-clipboard hashes (`ss:a3f2b1c9`), multi-project support via SQLite, run history with auto-pruning.
+
 ## Architecture
 
 ```
@@ -371,6 +449,12 @@ src/runner.js         Parallel test executor with retries and timeouts
 src/actions.js        Action engine: maps JSON actions to Puppeteer calls
 src/reporter.js       JSON reports, JUnit XML, console output
 src/mcp-server.js     MCP server: exposes tools for Claude Code
+src/mcp-tools.js      Shared MCP tool definitions and handlers
+src/dashboard.js      Web dashboard: HTTP server, REST API, WebSocket
+src/db.js             SQLite multi-project database
+src/issues.js         GitHub/GitLab issue fetching (gh/glab CLI)
+src/ai-generate.js    AI test generation (prompt builder + Claude API)
+src/verify.js         Bug verification orchestrator
 src/logger.js         ANSI colored logger
 src/index.js          Programmatic API (createRunner)
 templates/            Scaffolding templates for init command

package/bin/cli.js

@@ -13,6 +13,11 @@
  *   e2e-runner pool stop                  Stop the Chrome Pool
  *   e2e-runner pool status                Show pool status
  *   e2e-runner pool restart               Restart the pool
+ *   e2e-runner dashboard                   Start the web dashboard
+ *   e2e-runner issue <url>                Fetch issue and show details
+ *   e2e-runner issue <url> --generate     Generate test file via Claude API
+ *   e2e-runner issue <url> --verify       Generate + run + report bug status
+ *   e2e-runner issue <url> --prompt       Output the AI prompt (for piping)
  *   e2e-runner init                       Scaffold e2e/ in the current project
  *   e2e-runner --help                     Show help
  *   e2e-runner --version                  Show version
@@ -20,11 +25,16 @@
 
 import fs from 'fs';
 import path from 'path';
+import http from 'http';
 import { fileURLToPath } from 'url';
 import { loadConfig } from '../src/config.js';
 import { startPool, stopPool, restartPool, getPoolStatus, waitForPool } from '../src/pool.js';
 import { runTestsParallel, loadTestFile, loadTestSuite, loadAllSuites, listSuites } from '../src/runner.js';
-import { generateReport, saveReport, printReport } from '../src/reporter.js';
+import { generateReport, saveReport, printReport, persistRun } from '../src/reporter.js';
+import { startDashboard } from '../src/dashboard.js';
+import { fetchIssue } from '../src/issues.js';
+import { buildPrompt, generateTests, hasApiKey } from '../src/ai-generate.js';
+import { verifyIssue } from '../src/verify.js';
 import { log, colors as C } from '../src/logger.js';
 
 const __filename = fileURLToPath(import.meta.url);
@@ -59,6 +69,9 @@ function parseCLIConfig() {
   if (getFlag('--test-timeout')) cliArgs.testTimeout = parseInt(getFlag('--test-timeout'));
   if (getFlag('--output')) cliArgs.outputFormat = getFlag('--output');
   if (getFlag('--env')) cliArgs.env = getFlag('--env');
+  if (getFlag('--port')) cliArgs.dashboardPort = parseInt(getFlag('--port'));
+  if (getFlag('--dashboard-port')) cliArgs.dashboardPort = parseInt(getFlag('--dashboard-port'));
+  if (getFlag('--project-name')) cliArgs.projectName = getFlag('--project-name');
   return cliArgs;
 }
 
@@ -75,6 +88,14 @@ ${C.bold}Usage:${C.reset}
 
   e2e-runner list                       List available suites
 
+  e2e-runner dashboard                  Start the web dashboard
+  e2e-runner dashboard --port <port>    Custom port (default: 8484)
+
+  e2e-runner issue <url>                Fetch issue and show details
+  e2e-runner issue <url> --generate     Generate test file via Claude API
+  e2e-runner issue <url> --verify       Generate + run + report bug status
+  e2e-runner issue <url> --prompt       Output the AI prompt (for piping)
+
   e2e-runner pool start                 Start the Chrome Pool
   e2e-runner pool stop                  Stop the Chrome Pool
   e2e-runner pool status                Show pool status
@@ -96,6 +117,7 @@ ${C.bold}Options:${C.reset}
   --test-timeout <ms>      Per-test timeout (default: 60000)
   --output <format>        Report format: json, junit, both (default: json)
   --env <name>             Environment profile from config (default: default)
+  --project-name <name>    Project display name for dashboard (default: directory name)
 
 ${C.bold}Config:${C.reset}
   Looks for e2e.config.js or e2e.config.json in the current directory.
@@ -151,13 +173,35 @@ async function cmdRun() {
   const pressure = await waitForPool(config.poolUrl);
   log('✅', `Pool ready (${pressure.running}/${pressure.maxConcurrent} sessions, queued: ${pressure.queued})`);
 
+  // Wire up live progress to dashboard if running
+  let _lastBroadcast = null;
+  try {
+    const res = await fetch('http://127.0.0.1:' + (config.dashboardPort || 8484) + '/api/status');
+    if (res.ok) {
+      const dp = config.dashboardPort || 8484;
+      config.onProgress = (data) => {
+        const body = JSON.stringify(data);
+        _lastBroadcast = new Promise((resolve) => {
+          const req = http.request({ hostname: '127.0.0.1', port: dp, path: '/api/broadcast', method: 'POST', headers: { 'Content-Type': 'application/json', 'Content-Length': Buffer.byteLength(body) }, timeout: 1000 });
+          req.on('error', () => resolve());
+          req.on('close', () => resolve());
+          req.end(body);
+        });
+      };
+    }
+  } catch { /* dashboard not running */ }
+
   // Execute tests
   console.log('');
+  const suiteName = getFlag('--suite') || (hasFlag('--all') ? null : null);
   const results = await runTestsParallel(tests, config, hooks);
   const report = generateReport(results);
   saveReport(report, config.screenshotsDir, config);
+  persistRun(report, config, suiteName);
   printReport(report, config.screenshotsDir);
 
+  // Wait for the last dashboard broadcast (run:complete) to flush before exiting
+  if (_lastBroadcast) await _lastBroadcast;
   process.exit(report.summary.failed > 0 ? 1 : 0);
 }
 
@@ -288,6 +332,114 @@ ${C.bold}Next steps:${C.reset}
 `);
 }
 
+async function cmdDashboard() {
+  const cliArgs = parseCLIConfig();
+  const config = await loadConfig(cliArgs);
+
+  console.log(`\n${C.bold}${C.cyan}@matware/e2e-runner${C.reset} v${pkg.version}`);
+  console.log(`${C.dim}Starting dashboard on port ${config.dashboardPort}...${C.reset}\n`);
+
+  const handle = await startDashboard(config);
+
+  // Keep process alive until SIGINT/SIGTERM
+  const shutdown = () => {
+    console.log(`\n${C.dim}Shutting down dashboard...${C.reset}`);
+    handle.close();
+    process.exit(0);
+  };
+  process.on('SIGINT', shutdown);
+  process.on('SIGTERM', shutdown);
+}
+
+async function cmdIssue() {
+  const url = args[1];
+  if (!url || url.startsWith('--')) {
+    console.error(`${C.red}Usage: e2e-runner issue <url> [--generate|--verify|--prompt]${C.reset}`);
+    process.exit(1);
+  }
+
+  const cliArgs = parseCLIConfig();
+  const config = await loadConfig(cliArgs);
+
+  if (hasFlag('--prompt')) {
+    // Output AI prompt as JSON to stdout
+    const issue = fetchIssue(url);
+    const promptData = buildPrompt(issue, config);
+    console.log(JSON.stringify(promptData, null, 2));
+    return;
+  }
+
+  if (hasFlag('--verify')) {
+    // Generate + run + report
+    if (!hasApiKey(config)) {
+      console.error(`${C.red}ANTHROPIC_API_KEY is required for --verify mode.${C.reset}`);
+      process.exit(1);
+    }
+
+    console.log(`\n${C.bold}${C.cyan}@matware/e2e-runner${C.reset} v${pkg.version}`);
+    log('🔍', 'Fetching issue...');
+
+    const result = await verifyIssue(url, config);
+    const { issue, report, bugConfirmed } = result;
+
+    console.log('');
+    if (bugConfirmed) {
+      log('🐛', `${C.red}${C.bold}BUG CONFIRMED${C.reset} — ${issue.title}`);
+      log('', `${C.dim}${report.summary.failed} of ${report.summary.total} tests failed${C.reset}`);
+    } else {
+      log('✅', `${C.green}${C.bold}NOT REPRODUCIBLE${C.reset} — ${issue.title}`);
+      log('', `${C.dim}All ${report.summary.total} tests passed${C.reset}`);
+    }
+    console.log(`${C.dim}Issue: ${issue.url}${C.reset}\n`);
+
+    process.exit(bugConfirmed ? 1 : 0);
+  }
+
+  if (hasFlag('--generate')) {
+    // Generate test file via Claude API
+    if (!hasApiKey(config)) {
+      console.error(`${C.red}ANTHROPIC_API_KEY is required for --generate mode.${C.reset}`);
+      process.exit(1);
+    }
+
+    console.log(`\n${C.bold}${C.cyan}@matware/e2e-runner${C.reset} v${pkg.version}`);
+    log('🔍', 'Fetching issue...');
+
+    const issue = fetchIssue(url);
+    log('📋', `${C.cyan}${issue.title}${C.reset}`);
+    log('🤖', 'Generating tests via Claude API...');
+
+    const { tests, suiteName } = await generateTests(issue, config);
+
+    if (!fs.existsSync(config.testsDir)) {
+      fs.mkdirSync(config.testsDir, { recursive: true });
+    }
+    const filePath = path.join(config.testsDir, `${suiteName}.json`);
+    fs.writeFileSync(filePath, JSON.stringify(tests, null, 2) + '\n');
+
+    log('✅', `Created ${C.cyan}${filePath}${C.reset} (${tests.length} tests)`);
+    console.log(`${C.dim}Run with: e2e-runner run --suite ${suiteName}${C.reset}\n`);
+    return;
+  }
+
+  // Default: fetch and display issue
+  log('🔍', 'Fetching issue...');
+  const issue = fetchIssue(url);
+
+  console.log(`\n${C.bold}${issue.title}${C.reset}`);
+  console.log(`${C.dim}${'─'.repeat(50)}${C.reset}`);
+  console.log(`  Repo:    ${C.cyan}${issue.repo}${C.reset}`);
+  console.log(`  Number:  #${issue.number}`);
+  console.log(`  State:   ${issue.state === 'open' ? C.green : C.red}${issue.state}${C.reset}`);
+  console.log(`  Labels:  ${issue.labels.length ? issue.labels.join(', ') : C.dim + 'none' + C.reset}`);
+  console.log(`  URL:     ${C.dim}${issue.url}${C.reset}`);
+  if (issue.body) {
+    console.log(`\n${C.bold}Description:${C.reset}`);
+    console.log(issue.body.length > 500 ? issue.body.substring(0, 500) + '...' : issue.body);
+  }
+  console.log('');
+}
+
 // ==================== Main ====================
 
 async function main() {
@@ -316,6 +468,14 @@ async function main() {
       await cmdPool();
       break;
 
+    case 'dashboard':
+      await cmdDashboard();
+      break;
+
+    case 'issue':
+      await cmdIssue();
+      break;
+
     case 'init':
       cmdInit();
       break;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@matware/e2e-runner",
-  "version": "1.0.2",
+  "version": "1.1.0",
   "mcpName": "io.github.fastslack/e2e-runner",
   "description": "E2E test runner using Chrome Pool (browserless/chrome) with parallel execution",
   "type": "module",
@@ -25,7 +25,9 @@
     "browserless",
     "parallel",
     "mcp",
-    "claude-code"
+    "claude-code",
+    "github-issues",
+    "ai-testing"
   ],
   "author": "Matware",
   "license": "Apache-2.0",
@@ -36,6 +38,7 @@
   "homepage": "https://github.com/fastslack/mtw-e2e-runner#readme",
   "dependencies": {
     "@modelcontextprotocol/sdk": "^1.12.1",
+    "better-sqlite3": "^11.0.0",
     "puppeteer-core": "^24.0.0"
   },
   "engines": {

package/src/actions.js

@@ -7,6 +7,7 @@
  * The JS comes from team-authored JSON test files.
  */
 
+import path from 'path';
 import { log } from './logger.js';
 
 function sleep(ms) {
@@ -70,7 +71,9 @@ export async function executeAction(page, action, config) {
       if (!/\.(png|jpg|jpeg|webp)$/i.test(filename)) {
         filename += '.png';
       }
-      const filepath = `${screenshotsDir}/${filename}`;
+      // Sanitize: use only the basename to prevent path traversal
+      filename = path.basename(filename);
+      const filepath = path.join(screenshotsDir, filename);
       await page.screenshot({ path: filepath, fullPage: action.fullPage || false });
       return { screenshot: filepath };
     }
@@ -146,6 +149,19 @@ export async function executeAction(page, action, config) {
       await page.hover(selector);
       break;
 
+    case 'navigate': {
+      const navUrl = value.startsWith('http') ? value : `${baseUrl}${value}`;
+      // Navigate with a race: try page.goto but don't block more than 5s
+      // This handles SPAs where domcontentloaded may not fire on client-side routing
+      try {
+        await Promise.race([
+          page.goto(navUrl, { waitUntil: 'load', timeout: 30000 }),
+          sleep(5000),
+        ]);
+      } catch { /* navigation may still be loading */ }
+      break;
+    }
+
     case 'evaluate':
       // Intentional: runs JS in browser page context (from test JSON files)
       await page.evaluate(value);

package/src/ai-generate.js

@@ -0,0 +1,185 @@
+/**
+ * AI Test Generation — builds prompts and optionally calls Claude API
+ *
+ * Two modes:
+ *   1. buildPrompt() — Returns issue data + prompt for Claude Code (MCP mode, no API key)
+ *   2. generateTests() — Calls Claude API directly (CLI automation, requires ANTHROPIC_API_KEY)
+ */
+
+import fs from 'fs';
+import { listSuites } from './runner.js';
+
+const SYSTEM_PROMPT = `You are an E2E test generator for a JSON-driven browser test runner.
+
+You output ONLY valid JSON — no markdown fences, no explanation, no comments.
+
+The test format is:
+[
+  {
+    "name": "descriptive-test-name",
+    "actions": [
+      { "type": "goto", "value": "/path" },
+      { "type": "click", "selector": "#btn" },
+      { "type": "click", "text": "Button Label" },
+      { "type": "type", "selector": "input[name=email]", "value": "user@example.com" },
+      { "type": "wait", "selector": ".loaded" },
+      { "type": "wait", "text": "Expected text" },
+      { "type": "wait", "value": "2000" },
+      { "type": "assert_text", "text": "Expected text on page" },
+      { "type": "assert_url", "value": "/expected-path" },
+      { "type": "assert_visible", "selector": ".element" },
+      { "type": "assert_count", "selector": ".items", "value": "5" },
+      { "type": "screenshot", "value": "step-name.png" },
+      { "type": "select", "selector": "select#role", "value": "admin" },
+      { "type": "clear", "selector": "input" },
+      { "type": "press", "value": "Enter" },
+      { "type": "scroll", "selector": ".target" },
+      { "type": "hover", "selector": ".menu" },
+      { "type": "evaluate", "value": "document.title" }
+    ]
+  }
+]
+
+Rules:
+- Output a JSON array of test objects
+- Use only the action types listed above
+- "click" with "text" (no selector) finds buttons/links by visible text
+- "goto" values starting with "/" are relative to the app's base URL
+- Include a screenshot action before key assertions for debugging
+- For bug reports: write tests that assert the CORRECT behavior. If the test fails, the bug is confirmed
+- Keep test names descriptive and kebab-case
+- Prefer CSS selectors that are stable (data-testid, name, role) over fragile ones (nth-child, classes)
+- If the issue description is vague, create a reasonable test that covers the described scenario`;
+
+/**
+ * Returns a structured prompt + issue data for Claude Code to consume.
+ * Claude Code uses its own intelligence to create tests via e2e_create_test.
+ * No API key needed.
+ *
+ * @param {object} issue - Normalized issue from fetchIssue()
+ * @param {object} config - Loaded config
+ * @returns {object}
+ */
+export function buildPrompt(issue, config) {
+  let existingSuites = [];
+  try {
+    existingSuites = listSuites(config.testsDir).map(s => s.name);
+  } catch { /* no suites yet */ }
+
+  const prompt = `Based on the following issue, generate E2E test actions using the e2e_create_test tool.
+
+## Issue: ${issue.title}
+**Repo:** ${issue.repo}
+**Labels:** ${issue.labels.join(', ') || 'none'}
+**State:** ${issue.state}
+**URL:** ${issue.url}
+
+### Description
+${issue.body || 'No description provided.'}
+
+## Instructions
+1. Analyze the issue and determine what user flows to test
+2. Create one or more tests that verify the expected behavior
+3. For bug reports: assert the CORRECT behavior (test failure = bug confirmed)
+4. Use the \`e2e_create_test\` tool with suite name \`issue-${issue.number}\`
+5. After creating the test, use \`e2e_run\` with suite \`issue-${issue.number}\` to execute it
+
+Base URL: ${config.baseUrl}
+Existing suites: ${existingSuites.join(', ') || 'none'}`;
+
+  return {
+    issue: {
+      title: issue.title,
+      body: issue.body,
+      labels: issue.labels,
+      url: issue.url,
+      number: issue.number,
+      repo: issue.repo,
+      state: issue.state,
+    },
+    baseUrl: config.baseUrl,
+    prompt,
+    existingSuites,
+  };
+}
+
+/**
+ * Checks if the Anthropic API key is available.
+ * @returns {boolean}
+ */
+export function hasApiKey(config = {}) {
+  return !!(config.anthropicApiKey || process.env.ANTHROPIC_API_KEY);
+}
+
+/**
+ * Calls Claude API directly to generate E2E tests from an issue.
+ * Requires ANTHROPIC_API_KEY env var or config.anthropicApiKey.
+ *
+ * @param {object} issue - Normalized issue from fetchIssue()
+ * @param {object} config - Loaded config
+ * @returns {Promise<{ tests: object[], suiteName: string }>}
+ */
+export async function generateTests(issue, config) {
+  const apiKey = config.anthropicApiKey || process.env.ANTHROPIC_API_KEY;
+  if (!apiKey) {
+    throw new Error('ANTHROPIC_API_KEY is required for test generation. Set it as an environment variable or in config.');
+  }
+
+  const model = config.anthropicModel || 'claude-sonnet-4-5-20250929';
+  const suiteName = `issue-${issue.number}`;
+
+  const userMessage = `Generate E2E tests for this issue:
+
+Title: ${issue.title}
+Repo: ${issue.repo}
+Labels: ${issue.labels.join(', ') || 'none'}
+State: ${issue.state}
+
+Description:
+${issue.body || 'No description provided.'}
+
+Base URL: ${config.baseUrl}
+
+Output a JSON array of test objects. Nothing else.`;
+
+  const response = await fetch('https://api.anthropic.com/v1/messages', {
+    method: 'POST',
+    headers: {
+      'Content-Type': 'application/json',
+      'x-api-key': apiKey,
+      'anthropic-version': '2023-06-01',
+    },
+    body: JSON.stringify({
+      model,
+      max_tokens: 4096,
+      system: SYSTEM_PROMPT,
+      messages: [{ role: 'user', content: userMessage }],
+    }),
+  });
+
+  if (!response.ok) {
+    const body = await response.text();
+    throw new Error(`Claude API error (${response.status}): ${body}`);
+  }
+
+  const result = await response.json();
+  const text = result.content?.[0]?.text;
+  if (!text) {
+    throw new Error('Claude API returned empty response');
+  }
+
+  // Parse JSON — strip markdown fences if present
+  const cleaned = text.replace(/^```(?:json)?\s*\n?/m, '').replace(/\n?```\s*$/m, '').trim();
+  let tests;
+  try {
+    tests = JSON.parse(cleaned);
+  } catch (err) {
+    throw new Error(`Failed to parse generated tests as JSON: ${err.message}\n\nRaw output:\n${text}`);
+  }
+
+  if (!Array.isArray(tests)) {
+    throw new Error('Generated tests must be a JSON array');
+  }
+
+  return { tests, suiteName };
+}

package/src/config.js

@@ -30,6 +30,11 @@ const DEFAULTS = {
   outputFormat: 'json',
   env: 'default',
   hooks: { beforeAll: [], afterAll: [], beforeEach: [], afterEach: [] },
+  dashboardPort: 8484,
+  maxHistoryRuns: 100,
+  projectName: null,
+  anthropicApiKey: null,
+  anthropicModel: 'claude-sonnet-4-5-20250929',
 };
 
 function loadEnvVars() {
@@ -47,6 +52,9 @@ function loadEnvVars() {
   if (process.env.TEST_TIMEOUT) env.testTimeout = parseInt(process.env.TEST_TIMEOUT);
   if (process.env.OUTPUT_FORMAT) env.outputFormat = process.env.OUTPUT_FORMAT;
   if (process.env.E2E_ENV) env.env = process.env.E2E_ENV;
+  if (process.env.PROJECT_NAME) env.projectName = process.env.PROJECT_NAME;
+  if (process.env.ANTHROPIC_API_KEY) env.anthropicApiKey = process.env.ANTHROPIC_API_KEY;
+  if (process.env.ANTHROPIC_MODEL) env.anthropicModel = process.env.ANTHROPIC_MODEL;
   return env;
 }
 
@@ -100,5 +108,11 @@ export async function loadConfig(cliArgs = {}, cwd = null) {
     fs.mkdirSync(config.screenshotsDir, { recursive: true });
   }
 
+  // Stash cwd for project identity (used by db.js)
+  config._cwd = cwd;
+  if (!config.projectName) {
+    config.projectName = path.basename(cwd);
+  }
+
   return config;
 }