Update app.py

app.py

@@ -5,13 +5,13 @@ import gradio as gr
 # Environment and Model/Client Initialization
 # ------------------------------------------------------------------------------
 try:
-    # Assume we’re in Google Colab (or another local environment with PyTorch)
+    # Assume we’re in Google Colab or another local environment with PyTorch
     from google.colab import userdata
     HF_TOKEN = userdata.get('HF_TOKEN')
     import torch
     from transformers import AutoTokenizer, AutoModelForCausalLM
 
-    # Small performance tweak for similar input sizes.
+    # Performance tweak
     torch.backends.cudnn.benchmark = True
 
     model_name = "HuggingFaceH4/zephyr-7b-beta"
@@ -33,7 +33,6 @@ except ImportError:
     from transformers import AutoTokenizer
 
     tokenizer = AutoTokenizer.from_pretrained(model_name)
-    # If an HF_TOKEN is provided and valid, it can be passed; otherwise, omit it.
     hf_token = os.getenv("HF_TOKEN", None)
     if hf_token:
         client = InferenceClient(model_name, token=hf_token)
@@ -70,8 +69,8 @@ Start the conversation by expressing your current feelings or challenges from th
 def build_prompt(history: list[tuple[str, str]], system_message: str, message: str, max_response_words: int) -> str:
     """
     Build a text prompt (for local inference) that starts with the system message,
-    includes conversation history with "Doctor:" and "Patient:" lines, and ends with
-    a new "Doctor:" line prompting the patient.
+    includes conversation history with "Doctor:" and "Patient:" labels,
+    and ends with a new "Doctor:" line prompting the patient.
     """
     prompt = system_message.format(max_response_words=max_response_words) + "\n"
     for user_msg, assistant_msg in history:
@@ -81,7 +80,6 @@ def build_prompt(history: list[tuple[str, str]], system_message: str, message: s
     prompt += f"Doctor: {message}\nPatient: "
     return prompt
 
-
 def build_messages(history: list[tuple[str, str]], system_message: str, message: str, max_response_words: int):
     """
     Build a messages list (for InferenceClient) using OpenAI-style formatting.
@@ -96,7 +94,6 @@ def build_messages(history: list[tuple[str, str]], system_message: str, message:
     messages.append({"role": "user", "content": f"Doctor: {message}\nPatient:"})
     return messages
 
-
 def truncate_response(text: str, max_words: int) -> str:
     """
     Truncate the response text to the specified maximum number of words.
@@ -106,7 +103,6 @@ def truncate_response(text: str, max_words: int) -> str:
         return " ".join(words[:max_words]) + "..."
     return text
 
-
 # ------------------------------------------------------------------------------
 # Response Function
 # ------------------------------------------------------------------------------
@@ -120,7 +116,7 @@ def respond(
     max_response_words: int,
 ):
     """
-    Generate a response. For local inference, use the model.generate() on a prompt.
+    Generate a response. For local inference, use model.generate() on a prompt.
     For non-local inference, use client.chat_completion() with streaming tokens.
     """
     if inference_mode == "local":
@@ -142,7 +138,7 @@ def respond(
         messages = build_messages(history, system_message, message, max_response_words)
         response = ""
         try:
-            # Use streaming chat_completion
+            # Generate response using streaming chat_completion
             for chunk in client.chat_completion(
                 messages,
                 max_tokens=max_tokens,
@@ -150,7 +146,6 @@ def respond(
                 temperature=temperature,
                 top_p=top_p,
             ):
-                # The chunk returns a dictionary; get the token from the delta.
                 token = chunk.choices[0].delta.get("content", "")
                 response += token
             truncated_response = truncate_response(response, max_response_words)
@@ -159,7 +154,6 @@ def respond(
             print(f"An error occurred: {e}")
             return "I'm sorry, I encountered an error. Please try again."
 
-
 # ------------------------------------------------------------------------------
 # Optional Initial Message and Gradio Interface
 # ------------------------------------------------------------------------------
@@ -167,7 +161,7 @@ initial_user_message = (
     "I’m sorry you’ve been feeling overwhelmed. Could you tell me more about your arguments with your partner and how that’s affecting you?"
 )
 
-# Use 'type="messages"' for the chatbot to avoid deprecated tuple warnings.
+# Remove chatbot_kwargs (unsupported in the current ChatInterface) to avoid error.
 demo = gr.ChatInterface(
     fn=respond,
     additional_inputs=[
@@ -179,9 +173,7 @@ demo = gr.ChatInterface(
     ],
     title="Patient Interview Practice Chatbot",
     description="Simulate a patient interview. You (the user) act as the doctor, and the chatbot replies with the patient's perspective only.",
-    chatbot_kwargs={"type": "messages"},
 )
 
 if __name__ == "__main__":
-    # In Spaces, do not set share=True.
     demo.launch()