@comfanion/workflow 4.36.43 → 4.36.44

package/bin/cli.js

@@ -124,6 +124,7 @@ program
       install_vectorizer: true,   // Vectorizer ON by default
       vectorizer_enabled: true,
       vectorizer_auto_index: true,
+      vectorizer_model: 'Xenova/bge-small-en-v1.5',  // Default: balanced (quality + speed)
       project_name: path.basename(process.cwd())
     };
     
@@ -146,6 +147,7 @@ program
           // Parse vectorizer settings
           const vectorizerEnabledMatch = existingContent.match(/vectorizer:[\s\S]*?enabled:\s*(true|false)/);
           const vectorizerAutoIndexMatch = existingContent.match(/vectorizer:[\s\S]*?auto_index:\s*(true|false)/);
+          const vectorizerModelMatch = existingContent.match(/vectorizer:[\s\S]*?model:\s*["']?([^"'\n]+)["']?/);
           
           if (nameMatch) config.user_name = nameMatch[1];
           if (langMatch) config.communication_language = langMatch[1];
@@ -155,6 +157,7 @@ program
           if (jiraProjMatch) config.jira_project = jiraProjMatch[1];
           if (vectorizerEnabledMatch) config.vectorizer_enabled = vectorizerEnabledMatch[1] === 'true';
           if (vectorizerAutoIndexMatch) config.vectorizer_auto_index = vectorizerAutoIndexMatch[1] === 'true';
+          if (vectorizerModelMatch) config.vectorizer_model = vectorizerModelMatch[1].trim();
           
           isUpdate = true;
         } catch (e) {
@@ -249,6 +252,27 @@ program
           message: 'Install vectorizer? (semantic code search, ~100MB)',
           default: true
         },
+        {
+          type: 'list',
+          name: 'vectorizer_model',
+          message: 'Embedding model for semantic search:',
+          when: (answers) => answers.install_vectorizer,
+          choices: [
+            { 
+              name: 'MiniLM-L6 (Fast)    - ~10 files/10sec, 384 dims, good quality', 
+              value: 'Xenova/all-MiniLM-L6-v2' 
+            },
+            { 
+              name: 'BGE-small (Balanced) - ~9 files/10sec, 384 dims, better quality', 
+              value: 'Xenova/bge-small-en-v1.5' 
+            },
+            { 
+              name: 'BGE-base (Quality)   - ~3 files/10sec, 768 dims, best quality', 
+              value: 'Xenova/bge-base-en-v1.5' 
+            }
+          ],
+          default: 'Xenova/bge-small-en-v1.5'
+        },
         {
           type: 'confirm',
           name: 'vectorizer_auto_index',
@@ -449,6 +473,23 @@ program
         .replace(/(# Auto-index files.*\n\s+auto_index:)\s*(true|false)/, 
           `$1 ${config.vectorizer_auto_index}`);
       
+      // Add/update vectorizer model
+      if (config.vectorizer_model) {
+        if (configContent.includes('model:') && configContent.match(/vectorizer:[\s\S]*?model:/)) {
+          // Update existing model setting
+          configContent = configContent.replace(
+            /(vectorizer:[\s\S]*?)model:\s*["']?[^"'\n]+["']?/,
+            `$1model: "${config.vectorizer_model}"`
+          );
+        } else {
+          // Add model setting after auto_index
+          configContent = configContent.replace(
+            /(auto_index:\s*(true|false))/,
+            `$1\n  \n  # Embedding model for semantic search\n  # Options: Xenova/all-MiniLM-L6-v2 (fast), Xenova/bge-base-en-v1.5 (quality)\n  model: "${config.vectorizer_model}"`
+          );
+        }
+      }
+      
       await fs.writeFile(configPath, configContent);
       
       // Create docs structure (always)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@comfanion/workflow",
-  "version": "4.36.43",
+  "version": "4.36.44",
   "description": "Initialize OpenCode Workflow system for AI-assisted development with semantic code search",
   "type": "module",
   "bin": {

package/src/build-info.json

@@ -1,6 +1,6 @@
 {
-  "version": "4.36.43",
-  "buildDate": "2026-01-24T23:17:57.477Z",
+  "version": "4.36.44",
+  "buildDate": "2026-01-25T01:16:46.417Z",
   "files": [
     "config.yaml",
     "FLOW.yaml",

package/src/vectorizer/index.js

@@ -45,6 +45,7 @@ const DEFAULT_PRESETS = {
 // Will be populated from config.yaml if available
 let INDEX_PRESETS = { ...DEFAULT_PRESETS };
 let GLOBAL_IGNORE = [];
+let EMBEDDING_MODEL = 'Xenova/all-MiniLM-L6-v2';  // Default: fast model
 
 /**
  * Load index configuration from config.yaml
@@ -61,6 +62,13 @@ async function loadConfig(projectRoot) {
     
     const section = vectorizerMatch[1];
     
+    // Parse embedding model
+    const modelMatch = section.match(/^\s{2}model:\s*["']?([^"'\n]+)["']?/m);
+    if (modelMatch) {
+      EMBEDDING_MODEL = modelMatch[1].trim();
+      if (DEBUG) console.log('[vectorizer] Using model from config:', EMBEDDING_MODEL);
+    }
+    
     // Parse global exclude
     const excludeMatch = section.match(/^\s{2}exclude:\s*\n((?:\s{4}-\s+.+\n?)*)/m);
     if (excludeMatch) {
@@ -148,11 +156,14 @@ class CodebaseIndexer {
 
   async loadModel() {
     if (!this.model) {
-      if (DEBUG) console.log('[vectorizer] Loading embedding model...');
-      this.model = await pipeline('feature-extraction', 'Xenova/all-MiniLM-L6-v2', {
+      if (DEBUG) console.log(`[vectorizer] Loading embedding model: ${EMBEDDING_MODEL}...`);
+      // Model options:
+      // - Xenova/all-MiniLM-L6-v2: fast, 384 dims, ~10 files/10sec
+      // - Xenova/bge-base-en-v1.5: quality, 768 dims, ~3 files/10sec
+      this.model = await pipeline('feature-extraction', EMBEDDING_MODEL, {
         progress_callback: DEBUG ? undefined : null  // Suppress progress bar unless DEBUG
       });
-      if (DEBUG) console.log('[vectorizer] Model loaded');
+      if (DEBUG) console.log(`[vectorizer] Model loaded: ${EMBEDDING_MODEL}`);
     }
     return this.model;
   }
@@ -485,6 +496,7 @@ class CodebaseIndexer {
     return { 
       indexName: this.indexName,
       description: preset?.description || 'Custom index',
+      model: EMBEDDING_MODEL,
       fileCount, 
       chunkCount 
     };
@@ -553,4 +565,9 @@ class CodebaseIndexer {
   }
 }
 
-export { CodebaseIndexer, INDEX_PRESETS };
+// Getter for current embedding model (after config loaded)
+function getEmbeddingModel() {
+  return EMBEDDING_MODEL;
+}
+
+export { CodebaseIndexer, INDEX_PRESETS, getEmbeddingModel };