codecritique 2.0.0 → 2.0.2

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "codecritique",
-  "version": "2.0.0",
+  "version": "2.0.2",
   "description": "AI-powered code review tool for any programming language",
   "type": "module",
   "main": "src/index.js",

package/src/project-analyzer.js

@@ -14,6 +14,7 @@ import * as llm from './llm.js';
 import { FILE_SELECTION_SYSTEM_PROMPT, PROJECT_SUMMARY_SYSTEM_PROMPT } from './prompt-cache.js';
 import { isDocumentationFile, isTestFile } from './utils/file-validation.js';
 import { verboseLog } from './utils/logging.js';
+import { escapeSqlString } from './utils/string-utils.js';
 
 // Consolidated file classification configuration
 const FILE_PATTERNS = {
@@ -377,6 +378,7 @@ export class ProjectAnalyzer {
 
     try {
       verboseLog({}, chalk.gray(`   📊 Using LanceDB hybrid search for project: ${projectPath}`));
+      const projectPathFilter = `project_path = '${escapeSqlString(projectPath)}'`;
 
       // Unified query function
       const queryFiles = async (config) => {
@@ -384,15 +386,11 @@ export class ProjectAnalyzer {
           let query = table.query().select(['path', 'name', 'content', 'type', 'language']);
 
           if (config.whereClause) {
-            query = query.where(`project_path = '${projectPath}' AND (${config.whereClause})`);
+            query = query.where(`${projectPathFilter} AND (${config.whereClause})`);
           }
           else if (config.terms) {
             // For term-based searches, query ALL files and sort by depth to prioritize shallow config files
-            const allFiles = await table
-              .query()
-              .select(['path', 'name', 'content', 'type', 'language'])
-              .where(`project_path = '${projectPath}'`)
-              .toArray(); // NO LIMIT - get all files
+            const allFiles = await table.query().select(['path', 'name', 'content', 'type', 'language']).where(projectPathFilter).toArray(); // NO LIMIT - get all files
 
             // Sort by path depth (shorter paths first) to prioritize config files
             allFiles.sort((a, b) => {
@@ -416,7 +414,7 @@ export class ProjectAnalyzer {
             });
           }
           else {
-            query = query.where(`project_path = '${projectPath}'`);
+            query = query.where(projectPathFilter);
           }
 
           return await query.limit(config.limit || 30).toArray();
@@ -655,7 +653,7 @@ Select files following the criteria in the system instructions.`;
       const records = await table
         .query()
         .select(['type', 'path', 'content_hash', 'project_path'])
-        .where(`project_path = '${projectPath.replace(/'/g, "''")}'`)
+        .where(`project_path = '${escapeSqlString(projectPath)}'`)
         .toArray();
 
       const hash = crypto.createHash('sha256');

package/src/project-analyzer.test.js

@@ -402,6 +402,14 @@ describe('ProjectAnalyzer', () => {
       expect(mockTable.query).toHaveBeenCalled();
     });
 
+    it('should escape project paths in key file queries', async () => {
+      const queryChain = mockTable.query();
+
+      await analyzer.mineKeyFilesFromEmbeddings("/mock/we're/project");
+
+      expect(queryChain.where).toHaveBeenCalledWith(expect.stringContaining("/mock/we''re/project"));
+    });
+
     it('should handle table optimization errors gracefully', async () => {
       mockTable.optimize.mockRejectedValue(new Error('legacy format'));
       mockTable.query().toArray.mockResolvedValue([]);

package/src/rag-analyzer.js

@@ -722,7 +722,7 @@ function prepareContextForLLM(filePath, content, language, finalCodeExamples, fi
     context: contextSections,
     codeExamples,
     guidelineSnippets,
-    customDocs: relevantCustomDocChunks || customDocs, // Use relevant chunks if available, fallback to full docs
+    customDocs: relevantCustomDocChunks?.length > 0 ? relevantCustomDocChunks : customDocs, // Use relevant chunks if available, fallback to full docs
     feedbackContext: generateFeedbackContext(dismissedPatterns), // Add feedback context for LLM
     projectSummary: projectSummary, // Add project architecture summary
     metadata: {
@@ -1729,7 +1729,12 @@ async function performHolisticPRAnalysis(options) {
           items: unifiedContext.prComments.slice(0, 10),
         },
       ],
-      customDocs: unifiedContext.customDocChunks || options.relevantCustomDocChunks || customDocs, // Use unified chunks first, then relevant chunks, then full docs
+      customDocs:
+        unifiedContext.customDocChunks?.length > 0
+          ? unifiedContext.customDocChunks
+          : options.relevantCustomDocChunks?.length > 0
+            ? options.relevantCustomDocChunks
+            : customDocs, // Use chunks only when available, then fall back to full docs
       projectSummary: projectSummary, // Add project architecture summary
       metadata: {
         hasCodeExamples: unifiedContext.codeExamples.length > 0,

package/src/rag-analyzer.test.js

@@ -762,6 +762,19 @@ describe('rag-analyzer', () => {
       expect(result.success).toBe(true);
     });
 
+    it('should fall back to full custom docs when no relevant chunks are found', async () => {
+      mockEmbeddingsSystem.getExistingCustomDocumentChunks.mockResolvedValue([
+        { id: 'existing', content: 'Existing indexed chunk', document_title: 'Indexed Doc', chunk_index: 0 },
+      ]);
+      mockEmbeddingsSystem.findRelevantCustomDocChunks.mockResolvedValue([]);
+
+      await runAnalysis('/test/file.js', {
+        customDocs: [{ content: 'Mandatory internal review rule', document_title: 'Internal Standards' }],
+      });
+
+      expect(llm.sendPromptToClaude).toHaveBeenCalledWith(expect.stringContaining('Mandatory internal review rule'), expect.any(Object));
+    });
+
     it('should log selected chunks when verbose', async () => {
       mockEmbeddingsSystem.findRelevantCustomDocChunks.mockResolvedValue([
         { id: 'chunk1', content: 'Guidelines', document_title: 'Guide', chunk_index: 0, similarity: 0.85 },

package/src/rag-review.js

@@ -230,7 +230,17 @@ async function reviewPullRequestWithCrossFileContext(filesToReview, options = {}
     verboseLog(verbose, chalk.gray(`Base branch: ${baseBranch}, Target branch: ${targetBranch}`));
 
     const prFiles = [];
-    for (const filePath of filesToReview) {
+    if (options.preloadedPRFiles?.length > 0) {
+      prFiles.push(
+        ...options.preloadedPRFiles.map((file) => ({
+          ...file,
+          baseBranch: file.baseBranch || baseBranch,
+          targetBranch: file.targetBranch || targetBranch,
+        }))
+      );
+    }
+
+    for (const filePath of options.preloadedPRFiles?.length > 0 ? [] : filesToReview) {
       try {
         // Check if the file should be processed before fetching its content from git
         if (!shouldProcessFile(filePath, '', options)) {
@@ -596,6 +606,7 @@ async function reviewPRChunk(chunk, sharedContext, options, chunkNumber, totalCh
 
   // Skip chunking decision for chunk reviews to prevent infinite recursion
   const skipChunkingOptions = { ...chunkOptions, skipChunking: true };
+  skipChunkingOptions.preloadedPRFiles = chunk.files;
 
   return await reviewPullRequestWithCrossFileContext(chunkFilePaths, skipChunkingOptions);
 }