openairev 0.2.2 → 0.2.4

package/bin/openairev.js

@@ -12,7 +12,7 @@ const program = new Command();
 program
   .name('openairev')
   .description('OpenAIRev — cross-model AI code reviewer')
-  .version('0.2.2');
+  .version('0.2.4');
 
 program
   .command('init')

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "openairev",
-  "version": "0.2.2",
+  "version": "0.2.4",
   "description": "Cross-model AI code reviewer — independent review for AI-assisted coding workflows",
   "type": "module",
   "bin": {

package/src/agents/codex.js

@@ -46,7 +46,12 @@ export class CodexAdapter {
           if (event.type === 'item.completed' && event.item?.type === 'agent_message') {
             agentMessage = event.item.text;
           }
-        } catch {
+          if (event.type === 'error' || event.type === 'turn.failed') {
+            const errMsg = event.message || event.error?.message || JSON.stringify(event);
+            throw new Error(`Codex error: ${errMsg}`);
+          }
+        } catch (e) {
+          if (e.message?.startsWith('Codex error:')) throw e;
           // skip non-JSON lines
         }
       }

package/src/cli/init.js

@@ -241,6 +241,7 @@ This project uses OpenAIRev for independent AI code review. When the user asks t
 
 - You are the **executor**. Set \`executor\` to \`"claude_code"\` when calling the tool.
 - A different AI model will review your code independently.
+- **IMPORTANT**: Pass only the diff for files you changed using \`diff_cmd\`, e.g. \`"git diff HEAD -- src/auth.ts src/routes.ts"\`. Do NOT let it auto-detect — the full repo diff may be too large.
 - When you receive review feedback, treat it as **peer review** — use your judgment, don't blindly apply every suggestion.
 - The review verdict includes \`critical_issues\`, \`repair_instructions\`, and a \`confidence\` score. Focus on high-confidence critical issues.
 ${marker}
@@ -300,6 +301,7 @@ This project uses OpenAIRev for independent AI code review. When the user asks t
 
 - You are the **executor**. Set \`executor\` to \`"codex"\` when calling the tool.
 - A different AI model will review your code independently.
+- **IMPORTANT**: Pass only the diff for files you changed using \`diff_cmd\`, e.g. \`"git diff HEAD -- src/auth.ts src/routes.ts"\`. Do NOT let it auto-detect — the full repo diff may be too large.
 - When you receive review feedback, treat it as **peer review** — use your judgment, don't blindly apply every suggestion.
 - The review verdict includes \`critical_issues\`, \`repair_instructions\`, and a \`confidence\` score. Focus on high-confidence critical issues.
 ${marker}

package/src/config/plan-verdict-schema.json

@@ -39,6 +39,6 @@
       "items": { "type": "string" }
     }
   },
-  "required": ["status", "critical_issues", "risk_level", "confidence"],
+  "required": ["status", "critical_issues", "missing_requirements", "sequencing_issues", "risks", "risk_level", "confidence", "repair_instructions", "false_positives_reconsidered"],
   "additionalProperties": false
 }

package/src/config/verdict-schema.json

@@ -39,6 +39,6 @@
       "items": { "type": "string" }
     }
   },
-  "required": ["status", "critical_issues", "risk_level", "confidence"],
+  "required": ["status", "critical_issues", "test_gaps", "requirement_mismatches", "rule_violations", "risk_level", "confidence", "repair_instructions", "false_positives_reconsidered"],
   "additionalProperties": false
 }

package/src/mcp/mcp-server.js

@@ -14,7 +14,7 @@ const config = loadConfig(cwd);
 
 const server = new McpServer({
   name: 'openairev',
-  version: '0.2.2',
+  version: '0.2.4',
 });
 
 server.tool(
@@ -22,17 +22,27 @@ server.tool(
   'TRIGGER: Use this tool when the user says "review", "review my code", "get a review", "check my changes", "openairev", or asks for independent/cross-model code review. Sends current code changes to a DIFFERENT AI model for independent review. Returns a structured verdict with critical issues, test gaps, risk level, confidence score, and repair instructions.',
   {
     executor: z.string().optional().describe('Which agent wrote the code (claude_code or codex). If you are Claude Code, set this to "claude_code". If you are Codex, set this to "codex".'),
-    diff: z.string().optional().describe('The diff or code to review. If omitted, auto-detects from git.'),
+    diff: z.string().optional().describe('The diff to review. IMPORTANT: Pass only the diff for files YOU changed, not the entire repo. Use `git diff HEAD -- file1 file2` to scope it. If omitted, auto-detects from git which may be too large.'),
+    diff_cmd: z.string().optional().describe('The git command used to get the diff, e.g. "git diff HEAD -- src/auth.ts src/routes.ts". If provided instead of diff, the server will run this command to get the diff.'),
     task_description: z.string().optional().describe('What the code is supposed to do. Used for requirement checking.'),
   },
-  async ({ executor, diff, task_description }) => {
+  async ({ executor, diff, diff_cmd, task_description }) => {
     const execAgent = executor || Object.keys(config.agents || {}).find(a => config.agents[a].available);
     const reviewerName = getReviewer(config, execAgent);
     if (!reviewerName) {
       return { content: [{ type: 'text', text: `No reviewer configured for executor "${execAgent}"` }] };
     }
 
-    const diffContent = diff || getDiff();
+    let diffContent = diff;
+    if (!diffContent && diff_cmd) {
+      try {
+        const { execSync } = await import('child_process');
+        diffContent = execSync(diff_cmd, { encoding: 'utf-8', maxBuffer: 10 * 1024 * 1024, cwd });
+      } catch (e) {
+        return { content: [{ type: 'text', text: `diff_cmd failed: ${e.message}` }] };
+      }
+    }
+    if (!diffContent) diffContent = getDiff();
     if (!diffContent?.trim()) {
       return { content: [{ type: 'text', text: 'No changes found to review.' }] };
     }

package/src/tools/git-tools.js

@@ -1,18 +1,34 @@
 import { execFileSync } from 'child_process';
 
+// Files to exclude from diffs — lock files, generated code, etc.
+const EXCLUDE_PATTERNS = [
+  '*.lock',
+  'package-lock.json',
+  'yarn.lock',
+  'pnpm-lock.yaml',
+  '*.generated.*',
+  'dist/*',
+  'build/*',
+  '.next/*',
+];
+
 /**
  * Get the current git diff. Tries staged first, then unstaged.
  * Returns empty string if no changes found.
+ * Uses minimal context (1 line) and excludes irrelevant files.
  */
-export function getDiff(ref) {
+export function getDiff(ref, { context = 1, excludes = EXCLUDE_PATTERNS } = {}) {
+  const excludeArgs = excludes.flatMap(p => ['--', `:!${p}`]);
+  const contextArgs = [`-U${context}`];
+
   if (ref) {
-    return gitExec(['diff', ref]);
+    return gitExec(['diff', ...contextArgs, ref, ...excludeArgs]);
   }
 
-  const staged = gitExec(['diff', '--cached']);
+  const staged = gitExec(['diff', '--cached', ...contextArgs, ...excludeArgs]);
   if (staged.trim()) return staged;
 
-  const unstaged = gitExec(['diff']);
+  const unstaged = gitExec(['diff', ...contextArgs, ...excludeArgs]);
   if (unstaged.trim()) return unstaged;
 
   return '';
@@ -20,7 +36,7 @@ export function getDiff(ref) {
 
 function gitExec(args) {
   try {
-    return execFileSync('git', args, { encoding: 'utf-8', maxBuffer: 5 * 1024 * 1024 });
+    return execFileSync('git', args, { encoding: 'utf-8', maxBuffer: 10 * 1024 * 1024 });
   } catch (e) {
     throw new Error(`git ${args[0]} failed: ${e.message}`);
   }