c

llm_server/routes/ooba_request_handler.py

@@ -18,7 +18,7 @@ class OobaRequestHandler(RequestHandler):
         if self.offline:
             msg = f'{self.selected_model} is not a valid model choice.'
             print(msg)
-            return jsonify({'results': [{'text': format_sillytavern_err(msg)}]}), 200
+            self.handle_error(msg)
 
         request_valid, invalid_response = self.validate_request()
         if not request_valid:

llm_server/routes/request_handler.py

@@ -13,7 +13,6 @@ from llm_server.helpers import auto_set_base_client_api
 from llm_server.llm.oobabooga.ooba_backend import OobaboogaBackend
 from llm_server.llm.vllm.vllm_backend import VLLMBackend
 from llm_server.routes.auth import parse_token
-from llm_server.routes.helpers.client import format_sillytavern_err
 from llm_server.routes.helpers.http import require_api_key, validate_json
 from llm_server.routes.queue import priority_queue
 
@@ -47,15 +46,13 @@ class RequestHandler:
             self.offline = True
         else:
             self.offline = False
-
-        self.selected_model = self.cluster_backend_info['model']
-        self.backend = get_backend_handler(self.cluster_backend_info['mode'], self.backend_url)
-        self.parameters = None
-        self.used = False
-
-        if self.token and not self.token.startswith('SYSTEM__'):
-            # "recent_prompters" is only used for stats.
-            redis.zadd('recent_prompters', {self.client_ip: time.time()})
+            self.selected_model = self.cluster_backend_info['model']
+            self.backend = get_backend_handler(self.cluster_backend_info['mode'], self.backend_url)
+            self.parameters = None
+            self.used = False
+            if self.token and not self.token.startswith('SYSTEM__'):
+                # "recent_prompters" is only used for stats.
+                redis.zadd('recent_prompters', {self.client_ip: time.time()})
 
     def get_auth_token(self):
         if self.request_json_body.get('X-API-KEY'):