adjust a few final things

llm_server/database/database.py

@@ -120,7 +120,7 @@ def weighted_average_column_for_model(table_name, column_name, model_name, backe
             results = cursor.fetchall()
         except Exception:
             traceback.print_exc()
-            return -1
+            return None
 
         total_weight = 0
         weighted_sum = 0

llm_server/routes/v1/generate_stream.py

@@ -14,7 +14,7 @@ from ...stream import sock
 
 # TODO: have workers process streaming requests
 
-@sock.route('/api/v1/stream')  # TODO: use blueprint route???
+@sock.route('/api/v1/stream')
 def stream(ws):
     if not opts.enable_streaming:
         # TODO: return a formatted ST error message
@@ -70,13 +70,12 @@ def stream(ws):
                     if partial_response.endswith(b'\x00'):
                         json_str = partial_response[:-1].decode()  # Remove the null character and decode the byte string to a string
                         json_obj = json.loads(json_str)
-                        if len(json_obj['text'][0].split(input_prompt + generated_text)) > 2:
+                        try:
+                            new = json_obj['text'][0].split(input_prompt + generated_text)[1]
+                        except IndexError:
                             # ????
-                            print(json_obj)
                             continue
 
-                        new = json_obj['text'][0].split(input_prompt + generated_text)[1]
-
                         ws.send(json.dumps({
                             'event': 'text_stream',
                             'message_num': message_num,

llm_server/threads.py

@@ -47,14 +47,14 @@ class MainBackgroundThread(Thread):
             # exclude_zeros=True filters out rows where an error message was returned. Previously, if there was an error, 0
             # was entered into the column. The new code enters null instead but we need to be backwards compatible for now.
             average_generation_elapsed_sec = weighted_average_column_for_model('prompts', 'generation_time', opts.running_model, opts.mode, opts.backend_url, exclude_zeros=True) or 0
-            if average_generation_elapsed_sec > -1:
+            if average_generation_elapsed_sec:  # returns None on exception
                 redis.set('average_generation_elapsed_sec', average_generation_elapsed_sec)
 
             # overall = average_column_for_model('prompts', 'generation_time', opts.running_model)
             # print(f'Weighted: {average_generation_elapsed_sec}, overall: {overall}')
 
             average_output_tokens = weighted_average_column_for_model('prompts', 'response_tokens', opts.running_model, opts.mode, opts.backend_url, exclude_zeros=True) or 0
-            if average_generation_elapsed_sec > -1:
+            if average_generation_elapsed_sec:
                 redis.set('average_output_tokens', average_output_tokens)
 
             # overall = average_column_for_model('prompts', 'response_tokens', opts.running_model)