formatted the input prompt when it's passed to gpt and set it to none before ifs and else

app.py

@@ -59,6 +59,7 @@ def multimodal_and_generation(message, history):
     depending on the request from prompt, that prompt output will return here.
     """
     print(f"Message:\n{message}\nType:\n{type(message)}")
+    image_path = None
     if message["files"]:
         if type(message["files"][-1]) == dict:
             image_path = message["files"][-1]["path"]
@@ -70,20 +71,20 @@ def multimodal_and_generation(message, history):
             if type(hist[0]) == tuple:
                 image_path = hist[0][0] # item inside items for history
 
-    prompt = f"<|start_header_id|>user<|end_header_id|>\n\n<image>\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
-
     if image_path is None:
+        input_prompt = message["text"]
         # GPT Generation
         client = OpenAI(api_key=API_KEY)
         stream = client.chat.completions.create(
             model="gpt-3.5-turbo",
             messages=[{"role": "system", "content": "You are a helpful assistant called 'chimera'."},
-                      {"role": "user", "content": message}],
+                      {"role": "user", "content": input_prompt}],
             stream=True,
         )
         return stream
 
     else:
+        prompt = f"<|start_header_id|>user<|end_header_id|>\n\n<image>\n{message['text']}<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n"
         # Time to instance the llava
         image = Image.open(image_path)
         inputs = processor(prompt, image, return_tensors='pt').to(0, torch.float16)