@realtimex/realtimex-alchemy 1.0.49 → 1.0.51

package/dist/CHANGELOG.md

@@ -5,6 +5,20 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [1.0.51] - 2026-01-27
+
+### Added
+- **Vector Storage**: Migrated embedding storage from local SDK to Supabase `pgvector`. This enables cloud-native similarity search and persistent memory across devices.
+- **Database**: Added `alchemy_vectors` table and HNSW indexes for high-performance semantic retrieval.
+
+### Improved
+- **Performance**: Implemented server-side similarity search via `match_vectors` RPC function, reducing latency for RAG and deduplication.
+
+## [1.0.50] - 2026-01-26
+
+### Improved
+- **Setup Wizard**: Added a smart check to automatically skip the migration step if the database is already initialized (`init_state` detected), streamlining the experience for existing projects.
+
 ## [1.0.49] - 2026-01-26
 
 ### Added

package/dist/api/services/AlchemistService.js

@@ -410,14 +410,14 @@ export class AlchemistService {
                     .eq('id', signal.id);
                 return;
             }
-            // Store embedding in RealTimeX vector storage
+            // Store embedding in Supabase pgvector storage
             await embeddingService.storeSignalEmbedding(signal.id, embedding, {
                 title: signal.title,
                 summary: signal.summary,
                 url: signal.url,
                 category: signal.category,
                 userId
-            });
+            }, supabase);
             // Update signal metadata
             await supabase
                 .from('signals')

package/dist/api/services/ChatService.js

@@ -64,7 +64,7 @@ export class ChatService {
             let sources = [];
             // 3. Retrieve Context (if embedding checks out)
             if (queryEmbedding) {
-                const similar = await embeddingService.findSimilarSignals(queryEmbedding, userId, 0.55, // Lowered threshold for better recall
+                const similar = await embeddingService.findSimilarSignals(queryEmbedding, userId, supabase, 0.55, // Lowered threshold for better recall
                 10 // Increased Top K
                 );
                 console.log(`[ChatService] RAG Retrieval: Found ${similar.length} signals for query: "${content}"`);

package/dist/api/services/DeduplicationService.js

@@ -18,7 +18,7 @@ export class DeduplicationService {
         try {
             // 1. Semantic Check (if embedding exists)
             if (embedding && embedding.length > 0) {
-                const similar = await embeddingService.findSimilarSignals(embedding, userId, this.SIMILARITY_THRESHOLD, 5 // Check top 5 matches
+                const similar = await embeddingService.findSimilarSignals(embedding, userId, supabase, this.SIMILARITY_THRESHOLD, 5 // Check top 5 matches
                 );
                 if (similar.length > 0) {
                     const bestMatch = similar[0];

package/dist/api/services/EmbeddingService.js

@@ -1,11 +1,11 @@
 import { SDKService } from './SDKService.js';
 /**
- * Embedding Service using RealTimeX SDK
- * Provides simplified interface for embedding generation and vector operations
+ * Embedding Service
+ * Uses RealTimeX SDK for embedding generation
+ * Uses Supabase pgvector for vector storage and similarity search
  * Gracefully degrades if SDK is not available
  */
 export class EmbeddingService {
-    WORKSPACE_ID = 'alchemy-signals';
     SIMILARITY_THRESHOLD = 0.85;
     /**
      * Generate embedding for a single text
@@ -60,22 +60,29 @@ export class EmbeddingService {
         }
     }
     /**
-     * Store signal embedding in RealTimeX vector storage
+     * Store signal embedding in Supabase pgvector storage
      * @param signalId - Unique signal ID
      * @param embedding - Embedding vector
      * @param metadata - Signal metadata
+     * @param supabase - Supabase client
      */
-    async storeSignalEmbedding(signalId, embedding, metadata) {
+    async storeSignalEmbedding(signalId, embedding, metadata, supabase) {
         try {
-            const sdk = SDKService.getSDK();
-            if (!sdk) {
-                throw new Error('SDK not available');
+            // Format embedding as pgvector string
+            const embeddingStr = `[${embedding.join(',')}]`;
+            const { error } = await supabase
+                .from('alchemy_vectors')
+                .upsert({
+                signal_id: signalId,
+                user_id: metadata.userId,
+                embedding: embeddingStr,
+                model: 'text-embedding-3-small'
+            }, {
+                onConflict: 'signal_id,model'
+            });
+            if (error) {
+                throw error;
             }
-            await sdk.llm.vectors.upsert([{
-                    id: signalId,
-                    vector: embedding,
-                    metadata
-                }], { workspaceId: this.WORKSPACE_ID });
             console.log('[EmbeddingService] Stored embedding for signal:', signalId);
         }
         catch (error) {
@@ -84,32 +91,39 @@ export class EmbeddingService {
         }
     }
     /**
-     * Find similar signals using semantic search
+     * Find similar signals using semantic search via Supabase pgvector
      * @param queryEmbedding - Query embedding vector
      * @param userId - User ID for filtering
+     * @param supabase - Supabase client
      * @param threshold - Similarity threshold (0-1)
      * @param limit - Max results
      * @returns Array of similar signals
      */
-    async findSimilarSignals(queryEmbedding, userId, threshold = this.SIMILARITY_THRESHOLD, limit = 10) {
+    async findSimilarSignals(queryEmbedding, userId, supabase, threshold = this.SIMILARITY_THRESHOLD, limit = 10) {
         try {
-            const sdk = SDKService.getSDK();
-            if (!sdk) {
+            // Format embedding as pgvector string
+            const embeddingStr = `[${queryEmbedding.join(',')}]`;
+            const { data, error } = await supabase.rpc('match_vectors', {
+                query_embedding: embeddingStr,
+                match_threshold: threshold,
+                match_count: limit,
+                target_user_id: userId
+            });
+            if (error) {
+                console.error('[EmbeddingService] Similarity search RPC error:', error.message);
                 return [];
             }
-            const response = await sdk.llm.vectors.query(queryEmbedding, {
-                topK: limit,
-                workspaceId: this.WORKSPACE_ID
-            });
-            // Filter by similarity threshold and user
-            const results = response.results || [];
-            return results
-                .filter((r) => r.metadata?.userId === userId)
-                .filter((r) => r.score >= threshold)
-                .map((r) => ({
-                id: r.id,
-                score: r.score,
-                metadata: r.metadata || {}
+            // Map results to expected format
+            return (data || []).map((r) => ({
+                id: r.signal_id,
+                score: r.similarity,
+                metadata: {
+                    title: r.title,
+                    summary: r.summary,
+                    url: r.url,
+                    category: r.category,
+                    userId
+                }
             }));
         }
         catch (error) {
@@ -120,12 +134,18 @@ export class EmbeddingService {
     /**
      * Delete all embeddings for a user
      * @param userId - User ID
+     * @param supabase - Supabase client
      */
-    async deleteUserEmbeddings(userId) {
+    async deleteUserEmbeddings(userId, supabase) {
         try {
-            // Note: Current SDK only supports deleteAll
-            // In future, we may need user-specific workspaces
-            console.warn('[EmbeddingService] User-specific deletion not yet supported');
+            const { error } = await supabase
+                .from('alchemy_vectors')
+                .delete()
+                .eq('user_id', userId);
+            if (error) {
+                throw error;
+            }
+            console.log('[EmbeddingService] Deleted all embeddings for user:', userId);
         }
         catch (error) {
             console.error('[EmbeddingService] Deletion failed:', error.message);

package/dist/api/services/SDKService.js

@@ -147,7 +147,6 @@ export class SDKService {
             throw new Error('RealTimeX SDK not linked. Cannot get app data dir.');
         }
         try {
-            // @ts-ignore - SDK method for getting app data directory
             return await sdk.getAppDataDir();
         }
         catch (error) {