@hone-ai/cli 1.8.0 → 1.9.0

package/hone-cli.js

@@ -362,10 +362,16 @@ program
     console.log('Running setup-ai-pipeline.sh v3.1...');
     console.log('');
 
+    // Auto-detect non-TTY (CI, piped, Claude Code) and add --non-interactive
+    const isNonInteractive = opts.nonInteractive || !process.stdin.isTTY;
+    if (isNonInteractive && !opts.nonInteractive) {
+      console.log('  (non-TTY detected — running in non-interactive mode)');
+    }
+
     const flags = [
       `--source "${path.join(tmpDir, 'enterprise-github')}"`,
-      opts.dryRun        ? '--dry-run'         : '',
-      opts.nonInteractive? '--non-interactive'  : '',
+      opts.dryRun      ? '--dry-run'         : '',
+      isNonInteractive ? '--non-interactive'  : '',
     ].filter(Boolean).join(' ');
 
     try {
@@ -4144,6 +4150,7 @@ program
   .option('--contracts', 'Run contract validation between pipeline agents')
   .option('--snapshot', 'Save current eval + contract results as regression baseline')
   .option('--regression', 'Compare current results against saved baseline (detect drift)')
+  .option('--judge', 'Run LLM-as-judge scenarios (requires ANTHROPIC_API_KEY, costs tokens)')
   .action(async (opts) => {
     const path = require('path');
     const fs = require('fs');
@@ -4205,7 +4212,84 @@ program
       process.exit(results.failed > 0 ? 1 : 0);
     }
 
-    // Scenario evaluation mode
+    // LLM-judge mode (HC-019i / #268)
+    if (opts.judge) {
+      const { loadScenarios, formatResults } = require('./lib/eval-runner');
+      const { runJudgeScenario } = require('./lib/eval-llm-judge');
+
+      const apiKey = process.env.ANTHROPIC_API_KEY;
+      if (!apiKey) {
+        console.error('ANTHROPIC_API_KEY required for --judge mode. Set: export ANTHROPIC_API_KEY=sk-ant-...');
+        process.exit(1);
+      }
+
+      const scenarios = loadScenarios({
+        evalDir, agent: opts.agent, tag: opts.tag, scenarioId: opts.scenario,
+        readFile: (p) => fs.readFileSync(p, 'utf8'),
+        listDir: (p) => fs.readdirSync(p), isDir: (p) => fs.statSync(p).isDirectory(),
+        parseYaml: (text) => yaml.load(text),
+      });
+
+      const judgeScenarios = scenarios.filter(s => s.grading?.mode === 'llm-judge');
+      if (judgeScenarios.length === 0) {
+        console.log('No llm-judge scenarios found. Add grading.mode: llm-judge to eval YAML files.');
+        process.exit(0);
+      }
+
+      // LLM call function using Anthropic API
+      async function callLLM(systemPrompt, userPrompt) {
+        const { data } = await axios.post('https://api.anthropic.com/v1/messages', {
+          model: 'claude-sonnet-4-20250514',
+          max_tokens: 2048,
+          system: systemPrompt,
+          messages: [{ role: 'user', content: userPrompt }],
+        }, {
+          headers: {
+            'x-api-key': apiKey,
+            'anthropic-version': '2023-06-01',
+            'content-type': 'application/json',
+          },
+          timeout: 60000,
+        });
+        return data.content?.[0]?.text || '';
+      }
+
+      console.log(`Running ${judgeScenarios.length} LLM-judge scenario(s)...`);
+      console.log('');
+
+      const results = [];
+      for (const scenario of judgeScenarios) {
+        const agentName = scenario.evalAgent || scenario.agent;
+        const promptText = AGENT_PROMPTS[agentName];
+        if (!promptText) {
+          results.push({ id: scenario.id, agent: agentName, result: 'error',
+            checks: 0, checks_passed: 0, failures: [{ type: 'missing_prompt', passed: false, detail: `agent "${agentName}" not found` }] });
+          continue;
+        }
+        try {
+          const result = await runJudgeScenario({ scenario, agentPrompt: promptText, callLLM });
+          results.push(result);
+        } catch (e) {
+          results.push({ id: scenario.id, agent: agentName, result: 'error',
+            checks: 0, checks_passed: 0, failures: [{ type: 'llm_error', passed: false, detail: e.message }] });
+        }
+
+        if (opts.failFast && results[results.length - 1].result !== 'pass') break;
+      }
+
+      const summary = {
+        total: results.length,
+        passed: results.filter(r => r.result === 'pass').length,
+        failed: results.filter(r => r.result === 'fail').length,
+        errors: results.filter(r => r.result === 'error').length,
+        scenarios: results,
+      };
+
+      console.log(formatResults(summary, opts.format));
+      process.exit(summary.failed + summary.errors > 0 ? 1 : 0);
+    }
+
+    // Scenario evaluation mode (deterministic)
     const { loadScenarios, runAllScenarios, formatResults } = require('./lib/eval-runner');
 
     if (!fs.existsSync(evalDir)) {
@@ -4429,6 +4513,216 @@ process.on('SIGINT', () => {
   process.exit(0);
 });
 
+// ── Release Review (pre-deployment holistic review) ──────────────────────────
+program
+  .command('release-review')
+  .description('Holistic code review of all changed files before deployment (runs Opus)')
+  .option('--base <branch>', 'Base branch to diff against', 'main')
+  .option('--format <fmt>', 'Output format: pretty or json', 'pretty')
+  .option('--dry-run', 'Show what would be reviewed without calling the LLM', false)
+  .option('--max-files <n>', 'Max source files to include in review', '40')
+  .action(async (opts) => {
+    const { execSync } = require('child_process');
+    const fs = require('fs');
+    const repoRoot = process.cwd();
+
+    // 1. Get changed files
+    let changedFiles;
+    try {
+      const raw = execSync(`git diff --name-only origin/${opts.base}...HEAD`, { encoding: 'utf8', cwd: repoRoot });
+      changedFiles = raw.trim().split('\n').filter(Boolean);
+    } catch {
+      try {
+        const raw = execSync('git diff --name-only HEAD~10', { encoding: 'utf8', cwd: repoRoot });
+        changedFiles = raw.trim().split('\n').filter(Boolean);
+      } catch {
+        console.error('Could not determine changed files. Run from a git repo.');
+        process.exit(1);
+      }
+    }
+
+    if (changedFiles.length === 0) {
+      console.log('No changed files found. Nothing to review.');
+      process.exit(0);
+    }
+
+    // 2. Filter to source files
+    const sourceExts = ['.js', '.ts', '.py', '.go', '.java', '.rb', '.rs', '.sql', '.yml', '.yaml'];
+    const sourceFiles = changedFiles.filter(f =>
+      sourceExts.some(ext => f.endsWith(ext)) &&
+      !f.includes('node_modules') && !f.includes('.test.') && !f.includes('/test/')
+    );
+
+    const maxFiles = parseInt(opts.maxFiles, 10) || 40;
+    const filesToReview = sourceFiles.slice(0, maxFiles);
+
+    console.log('');
+    console.log('Hone AI — Production Review');
+    console.log('================================');
+    console.log(`Base: ${opts.base}`);
+    console.log(`Changed files: ${changedFiles.length} total, ${sourceFiles.length} source, ${filesToReview.length} to review`);
+    console.log('');
+
+    if (opts.dryRun) {
+      console.log('Source files that would be reviewed:');
+      for (const f of filesToReview) console.log(`  ${f}`);
+      if (sourceFiles.length > maxFiles) console.log(`  ... and ${sourceFiles.length - maxFiles} more (increase --max-files)`);
+      process.exit(0);
+    }
+
+    // 3. Check for API key
+    const apiKey = process.env.ANTHROPIC_API_KEY;
+    if (!apiKey) {
+      console.error('ANTHROPIC_API_KEY not set. Required for production review (Opus model).');
+      console.error('Set it: export ANTHROPIC_API_KEY=sk-ant-...');
+      process.exit(1);
+    }
+
+    // 4. Build the diff content (truncated per-file to stay within context)
+    let diffContent;
+    try {
+      diffContent = execSync(`git diff origin/${opts.base}...HEAD -- ${filesToReview.map(f => `'${f}'`).join(' ')}`, {
+        encoding: 'utf8', cwd: repoRoot, maxBuffer: 10 * 1024 * 1024,
+      });
+    } catch {
+      try {
+        diffContent = execSync('git diff HEAD~10', { encoding: 'utf8', cwd: repoRoot, maxBuffer: 10 * 1024 * 1024 });
+      } catch (e) {
+        console.error(`Could not generate diff: ${e.message}`);
+        process.exit(1);
+      }
+    }
+
+    // Truncate if over 100k chars (~25k tokens) to stay within budget
+    const MAX_DIFF_CHARS = 100000;
+    if (diffContent.length > MAX_DIFF_CHARS) {
+      diffContent = diffContent.slice(0, MAX_DIFF_CHARS) + '\n\n[... diff truncated at 100k chars ...]';
+    }
+
+    // 5. Build the prompt
+    const systemPrompt = [
+      '# Production Reviewer — Pre-Deployment Gate',
+      '',
+      'You are reviewing ALL changed files holistically before deployment.',
+      'Your job is to catch cross-file issues that per-story reviews miss.',
+      '',
+      '## Check each file for:',
+      '1. SQL injection (parameterized queries?)',
+      '2. Tenant isolation (org_id in every query?)',
+      '3. Error handling (unhandled promises?)',
+      '4. Race conditions (concurrent access?)',
+      '5. Resource leaks (connections released?)',
+      '6. Security (secrets exposed? auth correct?)',
+      '7. Logic bugs (null handling? type coercion?)',
+      '8. Dead code (built but never wired?)',
+      '9. Performance (N+1? unbounded? large payloads?)',
+      '',
+      '## Check cross-file interactions:',
+      '- Do modules wire together correctly?',
+      '- Is auth middleware in correct order?',
+      '- Are exported functions imported somewhere?',
+      '- Do CLI messages reference options that exist?',
+      '- Are DB constraints consistent with app validation?',
+      '',
+      '## Rate each finding: CRITICAL / HIGH / MEDIUM / LOW',
+      '- CRITICAL: tenant isolation breach, data leak, security bypass',
+      '- HIGH: race condition, dead safety code, UX broken',
+      '- MEDIUM: performance waste, missing validation, resource leak',
+      '- LOW: dead code, cosmetic',
+      '',
+      '## Output as JSON:',
+      '```json',
+      '{',
+      '  "findings": [{ "severity": "...", "file": "...", "line": N, "issue": "...", "recommendation": "..." }],',
+      '  "crossFileChecks": [{ "check": "...", "status": "ok|issue", "detail": "..." }],',
+      '  "summary": { "critical": N, "high": N, "medium": N, "low": N },',
+      '  "recommendation": "DEPLOY | FIX_FIRST | DO_NOT_DEPLOY"',
+      '}',
+      '```',
+    ].join('\n');
+
+    const userPrompt = [
+      `## Files changed (${filesToReview.length} source files):`,
+      filesToReview.map(f => `- ${f}`).join('\n'),
+      '',
+      '## Full diff:',
+      '```diff',
+      diffContent,
+      '```',
+      '',
+      'Review ALL files holistically. Return findings as JSON.',
+    ].join('\n');
+
+    console.log('Calling Anthropic API (claude-opus-4-20250514)...');
+    console.log('');
+
+    // 6. Call Anthropic Messages API
+    try {
+      const { data } = await axios.post('https://api.anthropic.com/v1/messages', {
+        model: 'claude-opus-4-20250514',
+        max_tokens: 8192,
+        system: systemPrompt,
+        messages: [{ role: 'user', content: userPrompt }],
+      }, {
+        headers: {
+          'x-api-key': apiKey,
+          'anthropic-version': '2023-06-01',
+          'content-type': 'application/json',
+        },
+        timeout: 120000,
+      });
+
+      const responseText = data.content?.[0]?.text || '';
+
+      // 7. Parse and display results
+      if (opts.format === 'json') {
+        // Try to extract JSON from response
+        const jsonMatch = responseText.match(/\{[\s\S]*\}/);
+        if (jsonMatch) {
+          try {
+            const parsed = JSON.parse(jsonMatch[0]);
+            console.log(JSON.stringify({
+              base: opts.base,
+              totalFiles: changedFiles.length,
+              sourceFiles: sourceFiles.length,
+              reviewedFiles: filesToReview.length,
+              model: 'claude-opus-4-20250514',
+              inputTokens: data.usage?.input_tokens || 0,
+              outputTokens: data.usage?.output_tokens || 0,
+              ...parsed,
+            }, null, 2));
+          } catch {
+            console.log(JSON.stringify({ raw: responseText }, null, 2));
+          }
+        } else {
+          console.log(JSON.stringify({ raw: responseText }, null, 2));
+        }
+      } else {
+        console.log(responseText);
+      }
+
+      // 8. Exit code based on findings
+      const hasCritical = responseText.includes('"CRITICAL"') || responseText.includes('"critical"');
+      const recommendation = responseText.includes('DO_NOT_DEPLOY');
+      if (hasCritical || recommendation) {
+        console.log('');
+        console.log('CRITICAL issues found. Fix before deploying.');
+        process.exit(1);
+      }
+    } catch (e) {
+      const status = e.response?.status;
+      const msg = e.response?.data?.error?.message || e.message;
+      if (status === 401) {
+        console.error('Invalid ANTHROPIC_API_KEY. Check your key and try again.');
+      } else if (status === 429) {
+        console.error('Rate limited by Anthropic API. Try again shortly.');
+      } else {
+        console.error(`Production review failed: ${msg}`);
+      }
+      process.exit(1);
+    }
+  });
+
 // ── CLI setup ─────────────────────────────────────────────────────────────────
 program
   .name('hone')

package/lib/eval-contracts.js

@@ -56,6 +56,32 @@ const PIPELINE_CONTRACTS = [
     outputGate: 'step_3',
     metadataField: 'step_3.gate_result',
   },
+  {
+    agent: 'e2e-test-spec-writer',
+    step: '5a',
+    inputArtifact: 'step-4-implementation.md',
+    inputGate: 'step_4.gate_result',
+    outputArtifact: null,
+    outputGate: null,
+    metadataField: null,
+    extraChecks: [
+      { text: 'Playwright', check: 'playwright', detail: 'generates Playwright specs' },
+      { text: 'data-testid', check: 'data_testid', detail: 'uses data-testid selectors' },
+    ],
+  },
+  {
+    agent: 'e2e-qa-spec-healer',
+    step: 'independent',
+    inputArtifact: null,
+    inputGate: null,
+    outputArtifact: null,
+    outputGate: null,
+    metadataField: null,
+    extraChecks: [
+      { text: 'DIAGNOSIS', check: 'diagnosis', detail: 'provides diagnosis category' },
+      { text: 'Application bug', check: 'app_bug_handling', detail: 'handles application bugs (skip + file bug)' },
+    ],
+  },
   {
     agent: 'code-builder',
     step: 4,
@@ -106,6 +132,21 @@ const PIPELINE_CONTRACTS = [
       { text: 'test_strategy', check: 'test_strategy', detail: 'includes test_strategy in plan' },
     ],
   },
+  {
+    agent: 'release-reviewer',
+    step: 'independent',
+    inputArtifact: null,
+    inputGate: null,
+    outputArtifact: null,
+    outputGate: null,
+    metadataField: null,
+    extraChecks: [
+      { text: 'CRITICAL', check: 'severity_critical', detail: 'defines CRITICAL severity level' },
+      { text: 'cross-file', check: 'cross_file_review', detail: 'checks cross-file interactions' },
+      { text: 'tenant', check: 'tenant_isolation', detail: 'checks tenant isolation' },
+      { text: 'DEPLOY', check: 'deploy_recommendation', detail: 'provides deploy/no-deploy recommendation' },
+    ],
+  },
 ];
 
 /**

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hone-ai/cli",
-  "version": "1.8.0",
+  "version": "1.9.0",
   "description": "Hone AI — Enterprise SDLC Pipeline CLI",
   "main": "hone-cli.js",
   "bin": {