Update app.py

app.py

@@ -81,7 +81,7 @@ def generate_text(prompt, model_path, text_generation_parameters, headers):
         print(output_list)
     return(stringlist)
 
-model_path = "Colby/StarCoder-3B-WoW-JSON"
+model_path = "Colby/StarCoder-1B-WoW-JSON"
 parameters = {
     "max_new_tokens": 250,
     "return_full_text": False,
@@ -137,10 +137,10 @@ def merlin_chat(message, history):
                     context += entsum + '\n\n'
     context
     system_msg = {
-        'role': 'system', 'content': context + f'\n\nThe following is a conversation about {max_topic}.'
+        'role': 'system', 'content': context
     }
     user_msg = {'role': 'user', 'content': message}
-    prompt = "[" + json.dumps(system_msg) + chat_json + json.dumps(user_msg) + "{'role': 'assistant, 'content': '*recalls \""
+    prompt = "[" + json.dumps(system_msg) + chat_json + json.dumps(user_msg) + "{'role': 'assistant, 'content': '"
     for attempt in range(3):
         # result = model(prompt, max_new_tokens=250, return_full_text=False, handle_long_generation="hole")
         result = generate_text(prompt, model_path, parameters, headers)
@@ -157,7 +157,9 @@ def merlin_chat(message, history):
         if end<=0:
             continue
         cleanStr = cleanStr[:end]
-        message = json.loads(cleanStr)
+        messageStr = prompt + cleanStr + "]"
+        messages = json.loads(cleanStr)
+        message = messages[-1]
         if message['role'] != 'assistant':
             continue
         return message['content']