@hone-ai/cli 1.8.1 → 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)) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@hone-ai/cli",
-  "version": "1.8.1",
+  "version": "1.9.0",
   "description": "Hone AI — Enterprise SDLC Pipeline CLI",
   "main": "hone-cli.js",
   "bin": {