app.py

import gradio as gr
import openai
import os # To securely load the API key
# No need to import torch, transformers, or peft for the external API call

# --- Configuration ---
# Your system prompt is now an instruction for the external model
SYSTEM_PROMPT = "You are a highly accurate math solver. Provide the final numerical answer to the user's problem. Use the required units (e.g., '40 cm^2') and round to two decimal places if needed. Do not show your work, steps, or formulas." 

# Initialize the OpenAI client using the environment variable
try:
    # This automatically looks for the OPENAI_API_KEY environment variable
    client = openai.OpenAI(api_key=os.environ.get("OPENAI_API_KEY"))
except Exception as e:
    print(f"Error initializing OpenAI client: {e}")
    # Fallback for local testing if key is not set as environment variable
    client = None

# ------------------------------------------------------------------
# 💡 EXPLICIT EXTERNAL API CALL FUNCTION
# ------------------------------------------------------------------
def call_external_api(prompt):
    """
    Calls the external OpenAI API to get the model's response.
    """
    if not client:
        return "Error: API Key not configured. Please set OPENAI_API_KEY environment variable."

    try:
        # Call the chat completions API
        response = client.chat.completions.create(
            model="gpt-3.5-turbo", # A fast and capable model
            messages=[
                {"role": "system", "content": SYSTEM_PROMPT},
                {"role": "user", "content": prompt}
            ],
            temperature=0.0 # Set to 0.0 for deterministic, accurate math answers
        )
        
        # Extract the text content from the response
        return response.choices[0].message.content.strip()
        
    except Exception as e:
        return f"API Call Error: Could not get a response from the external model. Details: {e}"

# --- Inference Function (The Main App Logic) ---

def generate_response(message, history):
    # We will pass the user's message directly to the external API
    # The system prompt is already defined in the API call
    
    # 1. Call the External Model (The API Call)
    # The external model handles all the complex generation and cleanup internally
    response = call_external_api(message)
    
    # 2. Return the clean response
    return response

# --- Gradio Interface ---

# We use gr.ChatInterface for a standard chatbot layout
demo = gr.ChatInterface(
    fn=generate_response,
    title=f"Reliable Math LLM (Powered by External API)",
    description="Ask a math problem! This uses a reliable external service for answers.",
)

if __name__ == "__main__":
    demo.launch()