openai error message cleanup

llm_server/llm/openai/oai_to_vllm.py

@@ -90,14 +90,7 @@ def return_invalid_model_err(requested_model: str):
         msg = f"The model `{requested_model}` does not exist"
     else:
         msg = "The requested model does not exist"
-    return jsonify({
+    return_oai_invalid_request_error(msg)
-        "error": {
-            "message": msg,
-            "type": "invalid_request_error",
-            "param": None,
-            "code": "model_not_found"
-        }
-    }), 404
 
 
 def return_oai_internal_server_error():
@@ -109,3 +102,14 @@ def return_oai_internal_server_error():
             "code": "internal_error"
         }
     }), 500
+
+
+def return_oai_invalid_request_error(msg: str = None):
+    return jsonify({
+        "error": {
+            "message": msg,
+            "type": "invalid_request_error",
+            "param": None,
+            "code": "model_not_found"
+        }
+    }), 404

llm_server/routes/openai/__init__.py

@@ -28,7 +28,7 @@ def handle_error(e):
     """
 
     _logger.error(f'OAI returning error: {e}')
-    return_oai_internal_server_error()
+    return return_oai_internal_server_error()
 
 
 from .models import openai_list_models

llm_server/routes/openai/chat_completions.py

@@ -32,8 +32,7 @@ def openai_chat_completions(model_name=None):
     else:
         handler = OpenAIRequestHandler(incoming_request=request, incoming_json=request_json_body, selected_model=model_name)
         if handler.offline:
-            # return return_invalid_model_err(model_name)
+            return return_oai_internal_server_error()
-            return_oai_internal_server_error()
 
         if not request_json_body.get('stream'):
             try:

llm_server/routes/openai_request_handler.py

@@ -14,7 +14,7 @@ from llm_server.custom_redis import redis
 from llm_server.database.database import is_api_key_moderated
 from llm_server.database.log_to_db import log_to_db
 from llm_server.llm import get_token_count
-from llm_server.llm.openai.oai_to_vllm import oai_to_vllm, validate_oai, return_invalid_model_err, return_oai_internal_server_error
+from llm_server.llm.openai.oai_to_vllm import oai_to_vllm, validate_oai, return_oai_internal_server_error, return_oai_invalid_request_error
 from llm_server.llm.openai.transform import ANTI_CONTINUATION_RE, ANTI_RESPONSE_RE, generate_oai_string, transform_messages_to_prompt, trim_messages_to_fit
 from llm_server.logging import create_logger
 from llm_server.routes.request_handler import RequestHandler
@@ -31,10 +31,7 @@ class OpenAIRequestHandler(RequestHandler):
     def handle_request(self) -> Tuple[flask.Response, int]:
         assert not self.used
         if self.offline:
-            # msg = return_invalid_model_err(self.selected_model)
+            return return_oai_internal_server_error()
-            # _logger.error(f'OAI is offline: {msg}')
-            # return self.handle_error(msg)
-            return_oai_internal_server_error()
 
         if GlobalConfig.get().openai_silent_trim:
             oai_messages = trim_messages_to_fit(self.request.json['messages'], self.cluster_backend_info['model_config']['max_position_embeddings'], self.backend_url)
@@ -111,14 +108,7 @@ class OpenAIRequestHandler(RequestHandler):
 
     def handle_error(self, error_msg: str, error_type: str = 'error') -> Tuple[flask.Response, int]:
         _logger.error(f'OAI Error: {error_msg}')
-        return jsonify({
+        return return_oai_invalid_request_error()
-            "error": {
-                "message": "Invalid request, check your parameters and try again.",
-                "type": "invalid_request_error",
-                "param": None,
-                "code": None
-            }
-        }), 400
 
     def build_openai_response(self, prompt, response, model=None):
         # Seperate the user's prompt from the context