Switch to cost-optimized paid models: Gemini 2.5 Flash Lite + text-embedding-3-small

requirements.txt

@@ -19,7 +19,4 @@ pytest>=7.0.0
 
 # AI Chatbot dependencies (OpenRouter API)
 requests>=2.28.0
-python-dotenv>=1.0.0
-
-# Local embeddings (free, no API required)
-sentence-transformers>=2.2.0
+python-dotenv>=1.0.0

todo.md

@@ -171,13 +171,14 @@
 - [x] Update `requirements.txt`: remove `google-genai`, add `requests>=2.28.0`
 - [x] Environment variable: `GEMINI_API_KEY` → `OPENROUTER_API_KEY`
 
-## Completed: Switch to Free Models
-
-- [x] Evaluate OpenRouter free models for chatbot use case
-- [x] Switch chat model: `google/gemini-2.0-flash-001` → `qwen/qwen3-next-80b-a3b-instruct:free`
-- [x] Implement local embeddings using `sentence-transformers` (all-MiniLM-L6-v2)
-  - 384-dimensional embeddings, runs locally (free)
-  - Lazy-loaded to avoid slow startup
-- [x] Remove OpenRouter embedding dependency (no free embedding models available)
-- [x] Add `sentence-transformers>=2.2.0` to requirements.txt
-- [x] Clear old embeddings cache (different dimensions: 1536 → 384)
+## Completed: Switch to Paid OpenRouter Models (Cost-Optimized)
+
+- [x] Evaluate OpenRouter models for chatbot use case (cost vs quality)
+- [x] Switch chat model: `google/gemini-2.5-flash-lite`
+  - $0.10/$0.40 per 1M tokens (input/output)
+  - 1M context window, 318 tok/s, multimodal
+- [x] Switch embedding model: `openai/text-embedding-3-small`
+  - $0.02 per 1M tokens
+  - 1536 dimensions, high quality
+- [x] Remove local `sentence-transformers` dependency (simpler, no TF conflicts)
+- [x] Estimated cost: ~$1.50/month for moderate usage

utils/openrouter_client.py

@@ -1,8 +1,8 @@
 """
 OpenRouter API Client
 
-Wrapper for OpenRouter API providing text generation capabilities
-for the AI chatbot feature. Embeddings use local sentence-transformers (free).
+Wrapper for OpenRouter API providing text generation and embedding capabilities
+for the AI chatbot feature.
 
 Uses the OpenAI-compatible API via requests.
 """
@@ -17,83 +17,11 @@ from typing import List, Dict, Optional
 # =============================================================================
 # Change these to switch models across the entire application
 
-DEFAULT_CHAT_MODEL = "qwen/qwen3-next-80b-a3b-instruct:free"
-
-# Local embedding model (runs locally, completely free)
-# all-MiniLM-L6-v2: 384 dimensions, fast, good quality for semantic search
-LOCAL_EMBEDDING_MODEL = "all-MiniLM-L6-v2"
-
-# =============================================================================
-
-
-# =============================================================================
-# LOCAL EMBEDDING SERVICE
-# =============================================================================
-
-class LocalEmbeddingService:
-    """Local embedding service using sentence-transformers (free, no API required)."""
-    
-    _instance = None
-    _model = None
-    
-    def __new__(cls):
-        if cls._instance is None:
-            cls._instance = super().__new__(cls)
-        return cls._instance
-    
-    def _load_model(self):
-        """Lazy load the embedding model."""
-        if self._model is None:
-            try:
-                from sentence_transformers import SentenceTransformer
-                print(f"Loading local embedding model: {LOCAL_EMBEDDING_MODEL}...")
-                self._model = SentenceTransformer(LOCAL_EMBEDDING_MODEL)
-                print("Local embedding model loaded successfully.")
-            except ImportError:
-                print("sentence-transformers not installed. Run: pip install sentence-transformers")
-                return None
-            except Exception as e:
-                print(f"Error loading embedding model: {e}")
-                return None
-        return self._model
-    
-    def get_embedding(self, text: str) -> Optional[List[float]]:
-        """
-        Get embedding vector for text using local model.
-        
-        Args:
-            text: Text to embed
-            
-        Returns:
-            Embedding vector as list of floats, or None if failed
-        """
-        model = self._load_model()
-        if model is None:
-            return None
-        
-        try:
-            embedding = model.encode(text, convert_to_numpy=True)
-            return embedding.tolist()
-        except Exception as e:
-            print(f"Local embedding error: {e}")
-            return None
-    
-    def get_query_embedding(self, query: str) -> Optional[List[float]]:
-        """Get embedding for a query (same as document embedding for this model)."""
-        return self.get_embedding(query)
-
-
-# Singleton instance for local embeddings
-_local_embedding_service: Optional[LocalEmbeddingService] = None
-
-
-def _get_local_embedding_service() -> LocalEmbeddingService:
-    """Get or create the local embedding service."""
-    global _local_embedding_service
-    if _local_embedding_service is None:
-        _local_embedding_service = LocalEmbeddingService()
-    return _local_embedding_service
+# Chat model: Gemini 2.5 Flash Lite - $0.10/$0.40 per 1M tokens, 1M context
+DEFAULT_CHAT_MODEL = "google/gemini-2.5-flash-lite"
 
+# Embedding model: text-embedding-3-small - $0.02 per 1M tokens, 1536 dimensions
+DEFAULT_EMBEDDING_MODEL = "openai/text-embedding-3-small"
 
 # =============================================================================
 
@@ -280,9 +208,7 @@ class OpenRouterClient:
     
     def get_embedding(self, text: str) -> Optional[List[float]]:
         """
-        Get embedding vector for text using local sentence-transformers model.
-        
-        Note: This uses a local model (free) instead of OpenRouter API.
+        Get embedding vector for text using OpenRouter Embedding API.
         
         Args:
             text: Text to embed
@@ -290,12 +216,33 @@ class OpenRouterClient:
         Returns:
             Embedding vector as list of floats, or None if failed
         """
-        service = _get_local_embedding_service()
-        return service.get_embedding(text)
+        if not self.is_available:
+            return None
+        
+        try:
+            response = requests.post(
+                f"{OPENROUTER_BASE_URL}/embeddings",
+                headers=self._headers,
+                json={
+                    "model": DEFAULT_EMBEDDING_MODEL,
+                    "input": text
+                },
+                timeout=30
+            )
+            response.raise_for_status()
+            
+            data = response.json()
+            return data["data"][0]["embedding"]
+        except Exception as e:
+            print(f"Embedding error: {e}")
+            return None
     
     def get_query_embedding(self, query: str) -> Optional[List[float]]:
         """
-        Get embedding vector for a query using local model.
+        Get embedding vector for a query.
+        
+        Note: OpenRouter doesn't have separate task types for embeddings,
+        so this calls the same endpoint as get_embedding.
         
         Args:
             query: Query text to embed
@@ -303,8 +250,7 @@ class OpenRouterClient:
         Returns:
             Embedding vector as list of floats, or None if failed
         """
-        service = _get_local_embedding_service()
-        return service.get_query_embedding(query)
+        return self.get_embedding(query)
 
 
 # Singleton instance