evaluation for rag systems update, script for evaluating against questions

services/rag/langchain/vector_storage.py

@@ -25,6 +25,10 @@ OLLAMA_EMBEDDING_MODEL = os.getenv("OLLAMA_EMBEDDING_MODEL", "nomic-embed-text")
 OPENAI_EMBEDDING_MODEL = os.getenv("OPENAI_EMBEDDING_MODEL", "text-embedding-ada-002")
 OPENAI_EMBEDDING_BASE_URL = os.getenv("OPENAI_EMBEDDING_BASE_URL")
 OPENAI_EMBEDDING_API_KEY = os.getenv("OPENAI_EMBEDDING_API_KEY")
+EMBEDDING_REQUEST_TIMEOUT_SECONDS = float(
+    os.getenv("EMBEDDING_REQUEST_TIMEOUT_SECONDS", "30")
+)
+EMBEDDING_MAX_RETRIES = int(os.getenv("EMBEDDING_MAX_RETRIES", "0"))
 
 
 def initialize_vector_store(
@@ -53,6 +57,8 @@ def initialize_vector_store(
             model=OPENAI_EMBEDDING_MODEL,
             openai_api_base=OPENAI_EMBEDDING_BASE_URL,
             openai_api_key=OPENAI_EMBEDDING_API_KEY,
+            request_timeout=EMBEDDING_REQUEST_TIMEOUT_SECONDS,
+            max_retries=EMBEDDING_MAX_RETRIES,
         )
     elif EMBEDDING_STRATEGY == "none":
         embeddings = None
@@ -63,6 +69,8 @@ def initialize_vector_store(
         embeddings = OllamaEmbeddings(
             model=OLLAMA_EMBEDDING_MODEL,
             base_url="http://localhost:11434",  # Default Ollama URL
+            sync_client_kwargs={"timeout": EMBEDDING_REQUEST_TIMEOUT_SECONDS},
+            async_client_kwargs={"timeout": EMBEDDING_REQUEST_TIMEOUT_SECONDS},
         )
 
     # Check if collection exists and create if needed