llm_server/routes/openai/chat_completions.py

@@ -3,7 +3,6 @@ import time
 import traceback
 
 from flask import Response, jsonify, request
-from redis import Redis
 
 from llm_server.custom_redis import redis
 from . import openai_bp
@@ -98,7 +97,6 @@ def openai_chat_completions():
                     oai_string = generate_oai_string(30)
 
                     def generate():
-                        try:
                         response = generator(msg_to_backend, handler.backend_url)
                         generated_text = ''
                         partial_response = b''
@@ -147,13 +145,12 @@ def openai_chat_completions():
                             r_url,
                             handler.backend_url,
                         )
-                        finally:
-                            # The worker incremented it, we'll decrement it.
-                            decrement_ip_count(handler.client_ip, 'processing_ips')
-                            decr_active_workers(handler.selected_model, handler.backend_url)
-                            print('cleaned up')
 
                     return Response(generate(), mimetype='text/event-stream')
                 except Exception:
                     traceback.print_exc()
                     return 'INTERNAL SERVER', 500
+                finally:
+                    # The worker incremented it, we'll decrement it.
+                    decrement_ip_count(handler.client_ip, 'processing_ips')
+                    decr_active_workers(handler.selected_model, handler.backend_url)