@o-lang/semantic-doc-search 1.0.10 → 1.0.12

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@o-lang/semantic-doc-search",
-  "version": "1.0.10",
+  "version": "1.0.12",
   "description": "O-lang Semantic Document Search Resolver with hybrid search, embeddings, rerank, and streaming.",
   "main": "src/index.js",
   "type": "commonjs",

package/src/adapters/pgvectorAdapter.js

@@ -11,57 +11,44 @@ class PgVectorAdapter {
       password: config.DB_PASSWORD || process.env.DB_PASSWORD,
       database: config.DB_NAME || process.env.DB_NAME || 'olang',
     });
-    
-    // Test connection
-    this.testConnection();
   }
 
-  async testConnection() {
-    try {
-      await this.pool.query('SELECT 1');
-      console.log('✅ PgVector adapter connected successfully');
-    } catch (error) {
-      console.error('❌ PgVector connection failed:', error.message);
-      throw error;
-    }
+  // Convert JavaScript array to PostgreSQL array format
+  // [1.0, 2.0, 3.0] -> {1.0,2.0,3.0}
+  arrayToPgArray(arr) {
+    return `{${arr.join(',')}}`;
   }
 
   async upsert({ id, vector, content, source, metadata = {} }) {
-    // Convert vector to proper format for pgvector
-    const vectorStr = `[${vector.join(',')}]`;
-    
-    const query = `
-      INSERT INTO doc_embeddings (id, embedding, content, source, metadata, updated_at)
-      VALUES ($1, $2::vector, $3, $4, $5, NOW())
-      ON CONFLICT (id) 
-      DO UPDATE SET 
-        embedding = $2::vector,
-        content = $3,
-        source = $4,
-        metadata = $5,
-        updated_at = NOW()`;
+    // Convert to PostgreSQL array format (NOT JSON)
+    const pgVector = this.arrayToPgArray(vector);
     
-    await this.pool.query(query, [id, vectorStr, content, source, JSON.stringify(metadata)]);
+    await this.pool.query(
+      `INSERT INTO doc_embeddings (id, embedding, content, source, metadata)
+       VALUES ($1, $2::vector, $3, $4, $5::jsonb)
+       ON CONFLICT (id) DO UPDATE
+       SET embedding = $2::vector, content = $3, source = $4, metadata = $5::jsonb, updated_at = NOW()`,
+      [id, pgVector, content, source, JSON.stringify(metadata)]
+    );
   }
 
   async query(vector, topK = 5) {
-    const vectorStr = `[${vector.join(',')}]`;
-    
-    const query = `
-      SELECT id, content, source, metadata,
-             1 - (embedding <=> $1::vector) AS score
-      FROM doc_embeddings
-      ORDER BY embedding <=> $1::vector
-      LIMIT $2`;
+    const pgVector = this.arrayToPgArray(vector);
     
-    const res = await this.pool.query(query, [vectorStr, topK]);
+    const res = await this.pool.query(
+      `SELECT id, content, source, metadata,
+              1 - (embedding <=> $1::vector) AS score
+       FROM doc_embeddings
+       ORDER BY embedding <=> $1::vector
+       LIMIT $2`,
+      [pgVector, topK]
+    );
     
-    // Return in the format your resolver expects
     return res.rows.map(row => ({
       id: row.id,
       content: row.content,
       source: row.source,
-      metadata: row.metadata,
+      meta: row.metadata,
       score: parseFloat(row.score)
     }));
   }

package/src/index.js

@@ -257,12 +257,50 @@ async function loadAllDocuments(context) {
   return documents;
 }
 
-// ✅ PGVECTOR SEARCH FUNCTION (Using imported adapter)
+// 🔥 AUTO-MIGRATION HELPER FUNCTIONS
+async function checkPgVectorHasData(pgVectorAdapter) {
+  try {
+    const result = await pgVectorAdapter.pool.query('SELECT COUNT(*) FROM doc_embeddings');
+    return parseInt(result.rows[0].count) > 0;
+  } catch (error) {
+    // Table doesn't exist or other error - treat as empty
+    return false;
+  }
+}
+
+async function migrateDocumentsToPgVector(docRoot, pgVectorAdapter, embedder) {
+  const baseDir = safeResolve(process.cwd(), docRoot);
+  if (!fs.existsSync(baseDir)) {
+    console.log('📁 No docs directory found, skipping migration');
+    return;
+  }
+
+  const files = fs.readdirSync(baseDir).filter(f => f.endsWith(".txt") || f.endsWith(".md"));
+  console.log(`🔄 Migrating ${files.length} documents to pgvector...`);
+  
+  for (const file of files) {
+    try {
+      const content = fs.readFileSync(path.join(baseDir, file), "utf8");
+      const vector = await embedder.embed(content);
+      
+      await pgVectorAdapter.upsert({
+        id: file,
+        vector: vector,
+        content: content,
+        source: `file:${file}`
+      });
+      console.log(`✅ Migrated ${file}`);
+    } catch (error) {
+      console.warn(`⚠️ Failed to migrate ${file}: ${error.message}`);
+    }
+  }
+}
+
+// ✅ PGVECTOR SEARCH FUNCTION WITH AUTO-MIGRATION
 async function performPgVectorSearch(query, context = {}) {
   const options = context.options || {};
   const topK = options.topK || 5;
   
-  // Check for POSTGRES_URL in context or environment
   const postgresUrl = context.POSTGRES_URL || process.env.POSTGRES_URL;
   if (!postgresUrl) {
     return { 
@@ -282,6 +320,16 @@ async function performPgVectorSearch(query, context = {}) {
   });
 
   try {
+    // 🔥 AUTO-MIGRATION LOGIC
+    if (context.migrate_on_demand && context.doc_root) {
+      const hasData = await checkPgVectorHasData(pgVectorAdapter);
+      if (!hasData) {
+        console.log('🔄 Auto-migrating documents to pgvector (first run)...');
+        await migrateDocumentsToPgVector(context.doc_root, pgVectorAdapter, embedder);
+        console.log('✅ Migration completed');
+      }
+    }
+
     const queryVector = await embedder.embed(query);
     const docs = await pgVectorAdapter.query(queryVector, topK);
 
@@ -465,7 +513,7 @@ async function performHybridDocQA(query, context = {}) {
 async function performDocQA(query, context = {}) {
   // 🔍 AUTO-DETECT MODE BASED ON CONTEXT
   
-  // Mode 1: pgvector mode (if PostgreSQL URL provided in context or env)
+  // Mode 1: pgvector mode (if PostgreSQL URL provided)
   const postgresUrl = context.POSTGRES_URL || process.env.POSTGRES_URL;
   if (postgresUrl) {
     console.log('🔍 Using pgvector search mode');