Always return only the new tokens in generation functions

extensions/api/blocking_api.py

@@ -43,7 +43,7 @@ class Handler(BaseHTTPRequestHandler):
 
             response = json.dumps({
                 'results': [{
-                    'text': answer[len(prompt):]
+                    'text': answer
                 }]
             })
             self.wfile.write(response.encode('utf-8'))

extensions/api/streaming_api.py

@@ -29,7 +29,7 @@ async def _handle_connection(websocket, path):
             prompt, generate_params, stopping_strings=stopping_strings, is_chat=False)
 
         # As we stream, only send the new bytes.
-        skip_index = len(prompt)
+        skip_index = 0
         message_num = 0
 
         for a in generator: