app.py updated

app.py

@@ -65,8 +65,8 @@ class Config:
     QDRANT_API_KEY = os.getenv("QDRANT_API_KEY")
     COLLECTION_NAME = os.getenv("COLLECTION_NAME", "documents")
     EMBEDDING_MODEL = os.getenv("EMBEDDING_MODEL", "sentence-transformers/all-MiniLM-L6-v2")
-    TOP_K = int(os.getenv("TOP_K", "5"))
-    SIMILARITY_THRESHOLD = float(os.getenv("SIMILARITY_THRESHOLD", "0.7"))
+    TOP_K = int(os.getenv("TOP_K", "10"))  # Increased from 5
+    SIMILARITY_THRESHOLD = float(os.getenv("SIMILARITY_THRESHOLD", "0.1"))  # Lowered from 0.7
     DEVICE = os.getenv("DEVICE", "cuda" if torch.cuda.is_available() else "cpu")
 
 class ApplicationState:
@@ -299,9 +299,13 @@ class DocumentManager:
         try:
             await self._ensure_collection_exists()
             
+            print(f"Document Search - Query: '{query}', Limit: {limit}, Min Score: {min_score}")
+            
             # Generate query embedding
             query_embedding = await self.embedding_service.get_query_embedding(query)
             
+            print(f"Document Search - Generated embedding vector of size: {len(query_embedding)}")
+            
             # Search in Qdrant
             search_results = await self.qdrant_client.search(
                 collection_name=self.collection_name,
@@ -310,22 +314,29 @@ class DocumentManager:
                 score_threshold=min_score
             )
             
+            print(f"Document Search - Qdrant returned {len(search_results)} results")
+            
             # Format results
             results = []
-            for result in search_results:
+            for i, result in enumerate(search_results):
+                content = result.payload.get("content", result.payload.get("chunk_text", ""))
+                print(f"Document Search - Result {i+1}: Score={result.score:.4f}, Content preview: {content[:100]}...")
+                
                 results.append({
                     "score": result.score,
-                    "text": result.payload.get("content", result.payload.get("chunk_text", "")),
+                    "text": content,
                     "file_path": result.payload.get("file_path", ""),
                     "document_id": result.payload.get("document_id", ""),
                     "chunk_index": result.payload.get("chunk_index", 0)
                 })
             
-            print(f"✓ Found {len(results)} results for query: '{query}'")
+            print(f"✓ Document Search - Found {len(results)} results for query: '{query}'")
             return results
             
         except Exception as e:
             print(f"Error searching: {e}")
+            import traceback
+            traceback.print_exc()
             return []
     
     async def list_documents(self) -> List[Dict[str, Any]]:
@@ -409,13 +420,31 @@ class RAGService:
                 print("Error: Document manager is not initialized")
                 return []
             
+            # Use a lower similarity threshold for RAG to get more results
+            # Try multiple thresholds if needed
+            min_score = 0.1  # Lower threshold for RAG
+            
+            print(f"RAG Search - Query: '{query}', Limit: {top_k}, Min Score: {min_score}")
+            
             # Use the document manager's search functionality
             results = await app_state.document_manager.search_documents(
                 query=query,
                 limit=top_k,
-                min_score=Config.SIMILARITY_THRESHOLD
+                min_score=min_score
             )
             
+            print(f"RAG Search - Found {len(results)} results")
+            
+            # If no results with low threshold, try even lower
+            if not results:
+                print("No results with min_score=0.1, trying with min_score=0.0")
+                results = await app_state.document_manager.search_documents(
+                    query=query,
+                    limit=top_k,
+                    min_score=0.0
+                )
+                print(f"RAG Search - Found {len(results)} results with min_score=0.0")
+            
             return results
             
         except Exception as e: