Remove OpenAI client, using it from langfuse.

utils.py

@@ -1,13 +1,12 @@
 import json
 import os
-from typing import Dict, Generator, List, Tuple
+from typing import Dict, Generator, List
 
 import streamlit as st
 import yaml
-from langfuse.decorators import observe
+from langfuse.decorators import langfuse_context, observe
 from langfuse.openai import openai
 from loguru import logger
-from openai import OpenAI
 from openai.types.chat import ChatCompletion
 from pydantic import BaseModel
 from qdrant_client import QdrantClient
@@ -74,7 +73,7 @@ def load_config(yaml_file_path: str = "./config.yaml") -> Config:
 
 
 @st.cache_resource
-def initialize_clients() -> Tuple[OpenAI, QdrantClient]:
+def initialize_clients() -> QdrantClient:
     """
     Initializes and returns the clients for OpenAI and Qdrant services.
 
@@ -90,18 +89,14 @@ def initialize_clients() -> Tuple[OpenAI, QdrantClient]:
         qdrant_api_key = os.environ["QDRANT_API_KEY"]
         qdrant_client = QdrantClient(url=qdrant_url, api_key=qdrant_api_key)
 
-        # Retrieve OpenAI client configuration from environment variables
-        openai_api_key = os.environ["OPENAI_API_KEY"]
-        openai_client = OpenAI(api_key=openai_api_key)
-
-        return openai_client, qdrant_client
+        return qdrant_client
     except KeyError as e:
         error_msg = f"Missing environment variable: {str(e)}"
         logger.error(error_msg)
         raise EnvironmentError(error_msg)
 
 
-@observe()
+@observe(as_type="generation")
 def call_llm(
     model: str,
     temperature: float,
@@ -132,14 +127,13 @@ def call_llm(
 
 @observe()
 def generate_response(
-    query: str, openai_client: OpenAI, qdrant_client: QdrantClient, config: Config
+    query: str, qdrant_client: QdrantClient, config: Config
 ) -> Generator[str, None, None]:
     """
     Generates a response for a given user query using a combination of semantic search and a chat model.
 
     Args:
     - query (str): The user's query string.
-    - openai_client (OpenAI): Client to interact with OpenAI's API.
     - qdrant_client (QdrantClient): Client to interact with Qdrant's API.
     - config (Config): Configuration settings for API interaction and response handling.
 
@@ -162,10 +156,10 @@ def generate_response(
         )
         collections = json.loads(response.choices[0].message.content)["response"]
         logger.info(f"Query routed to collections: {collections}")
+        langfuse_context.update_current_trace(tags=collections)
 
         # Embed the user query using the specified model in the configuration
         embedding_response = embed_text(
-            client=openai_client,
             text=query,
             model=config.openai.embeddings.model,
         )
@@ -191,6 +185,8 @@ def generate_response(
             if part is not None:
                 yield part
 
+        langfuse_context.flush()
+
     except Exception as e:
         logger.error(f"An error occurred while generating the response: {str(e)}")
         yield "Sorry, an error occurred while processing your request."