evaluation for rag systems update, script for evaluating against questions

services/rag/langchain/agent.py

@@ -20,6 +20,9 @@ from vector_storage import initialize_vector_store
 # Load environment variables
 load_dotenv()
 
+CHAT_REQUEST_TIMEOUT_SECONDS = float(os.getenv("CHAT_REQUEST_TIMEOUT_SECONDS", "45"))
+CHAT_MAX_RETRIES = int(os.getenv("CHAT_MAX_RETRIES", "0"))
+
 
 def get_llm_model_info(
     llm_model: Optional[str] = None,
@@ -149,10 +152,12 @@ def create_chat_agent(
             openai_api_base=base_url_or_api_base,
             openai_api_key=api_key,
             temperature=0.1,
+            request_timeout=CHAT_REQUEST_TIMEOUT_SECONDS,
+            max_retries=CHAT_MAX_RETRIES,
         )
 
         logger.info(
-            f"Using OpenAI-compatible model: {model_name} via {base_url_or_api_base}"
+            f"Using OpenAI-compatible model: {model_name} via {base_url_or_api_base}, timeout={CHAT_REQUEST_TIMEOUT_SECONDS}s, retries={CHAT_MAX_RETRIES}"
         )
     else:  # Default to ollama
         # Initialize the Ollama chat model
@@ -160,9 +165,13 @@ def create_chat_agent(
             model=model_name,
             base_url=base_url_or_api_base,  # Default Ollama URL
             temperature=0.1,
+            sync_client_kwargs={"timeout": CHAT_REQUEST_TIMEOUT_SECONDS},
+            async_client_kwargs={"timeout": CHAT_REQUEST_TIMEOUT_SECONDS},
         )
 
-        logger.info(f"Using Ollama model: {model_name}")
+        logger.info(
+            f"Using Ollama model: {model_name}, timeout={CHAT_REQUEST_TIMEOUT_SECONDS}s"
+        )
 
     # Create the document retrieval tool
     retrieval_tool = DocumentRetrievalTool()