@rigstate/mcp 0.5.2 → 0.5.4

package/package.json

@@ -1,6 +1,6 @@
 {
     "name": "@rigstate/mcp",
-    "version": "0.5.2",
+    "version": "0.5.4",
     "description": "Rigstate MCP Server - Model Context Protocol for AI Editors",
     "type": "module",
     "main": "./dist/index.js",
@@ -18,6 +18,7 @@
         "@ai-sdk/google": "^3.0.13",
         "@google/generative-ai": "^0.24.1",
         "@modelcontextprotocol/sdk": "^1.0.0",
+        "@openrouter/ai-sdk-provider": "^1.5.4",
         "@rigstate/rules-engine": "0.1.0",
         "@supabase/supabase-js": "^2.39.0",
         "ai": "^6.0.5",

package/src/tools/query-brain.ts

@@ -34,23 +34,37 @@ architecture rules, decisions, and constraints.`,
     }
 });
 
-// Simple embedding generation using Google Gemini
+// Generate embedding using the preferred provider (OpenRouter or Google)
 async function generateQueryEmbedding(query: string): Promise<number[] | null> {
-    const apiKey = process.env.GOOGLE_GENERATIVE_AI_API_KEY;
-    if (!apiKey) {
-        console.warn('GOOGLE_GENERATIVE_AI_API_KEY not found, skipping vector search.');
+    const openRouterKey = process.env.OPENROUTER_API_KEY;
+    const googleKey = process.env.GOOGLE_GENERATIVE_AI_API_KEY;
+
+    if (!openRouterKey && !googleKey) {
+        console.warn('Neither OPENROUTER_API_KEY nor GOOGLE_GENERATIVE_AI_API_KEY found, skipping vector search.');
         return null;
     }
 
     try {
-        const { google } = await import('@ai-sdk/google');
         const { embed } = await import('ai');
 
-        const { embedding } = await embed({
-            model: google.embedding('text-embedding-004'),
-            value: query.replace(/\n/g, ' '),
-        });
-        return embedding;
+        if (openRouterKey) {
+            const { createOpenRouter } = await import('@openrouter/ai-sdk-provider');
+            const openrouter = createOpenRouter({ apiKey: openRouterKey });
+
+            // Use Gemini embedding via OpenRouter to maintain 768 dimensions
+            const { embedding } = await embed({
+                model: openrouter.embedding('google/text-embedding-004'),
+                value: query.replace(/\n/g, ' '),
+            });
+            return embedding;
+        } else {
+            const { google } = await import('@ai-sdk/google');
+            const { embedding } = await embed({
+                model: google.embedding('text-embedding-004'),
+                value: query.replace(/\n/g, ' '),
+            });
+            return embedding;
+        }
     } catch (error) {
         console.error('Failed to generate embedding for search:', error);
         return null;
@@ -84,8 +98,8 @@ export async function queryBrain(
     let memories: MemoryRecord[] = [];
 
     // Use the hybrid search RPC
-    console.log(`Searching brain for "${query}" (limit: ${limit}, threshold: ${threshold})`);
-    console.log(`Embedding present: ${!!embedding}`);
+    console.error(`Searching brain for "${query}" (limit: ${limit}, threshold: ${threshold})`);
+    console.error(`Embedding present: ${!!embedding}`);
 
     const { data: searchResults, error: searchError } = await supabase
         .rpc('hybrid_search_memories', {
@@ -118,7 +132,7 @@ export async function queryBrain(
             }));
         }
     } else if (searchResults) {
-        console.log(`Found ${searchResults.length} results from RPC`);
+        console.error(`Found ${searchResults.length} results from RPC`);
         memories = searchResults.map((m: any) => ({
             id: m.id,
             content: m.content,