llm_server/llm/openai/oai_to_vllm.py

@@ -61,3 +61,6 @@ def validate_oai(parameters):
         return format_oai_err(f"{parameters['top_p']} is greater than the maximum of 1 - 'top_p'")
     if parameters.get('top_p', 1) < 0:
         return format_oai_err(f"{parameters['top_p']} less than the minimum of 0 - 'top_p'")
+
+    if parameters.get('max_tokens', 2) < 1:
+        return format_oai_err(f"{parameters['max_tokens']} is less than the minimum of 1 - 'max_tokens'")

llm_server/routes/openai/chat_completions.py

@@ -27,6 +27,9 @@ def openai_chat_completions():
         handler = OpenAIRequestHandler(incoming_request=request, incoming_json=request_json_body)
         if not request_json_body.get('stream'):
             try:
+                invalid_oai_err_msg = validate_oai(request_json_body)
+                if invalid_oai_err_msg:
+                    return invalid_oai_err_msg
                 return handler.handle_request()
             except Exception:
                 traceback.print_exc()

llm_server/routes/openai/completions.py

@@ -44,6 +44,9 @@ def openai_completions():
             pass
 
         if not request_json_body.get('stream'):
+            invalid_oai_err_msg = validate_oai(request_json_body)
+            if invalid_oai_err_msg:
+                return invalid_oai_err_msg
             response, status_code = handler.handle_request(return_ok=False)
             if status_code == 429:
                 return handler.handle_ratelimited()

llm_server/routes/request_handler.py

@@ -171,7 +171,17 @@ class RequestHandler:
             else:
                 error_msg = error_msg.strip('.') + '.'
             backend_response = self.handle_error(error_msg)
-            log_prompt(self.client_ip, self.token, prompt, backend_response[0].data.decode('utf-8'), None, self.parameters, dict(self.request.headers), response_status_code, self.request.url, self.backend_url, is_error=True)
+            log_prompt(ip=self.client_ip,
+                       token=self.token,
+                       prompt=prompt,
+                       response=backend_response[0].data.decode('utf-8'),
+                       gen_time=None,
+                       parameters=self.parameters,
+                       headers=dict(self.request.headers),
+                       backend_response_code=response_status_code,
+                       request_url=self.request.url,
+                       backend_url=self.backend_url,
+                       is_error=True)
             return (False, None, None, 0), backend_response
 
         # ===============================================