app.py

import os
import json
import google.generativeai as genai
import gradio as gr

# Configure the Gemini API
genai.configure(api_key=os.environ["GEMINI_API_KEY"])

# Generation configuration
generation_config = {
    "temperature": 1,
    "top_p": 0.95,
    "top_k": 64,
    "max_output_tokens": 8192,
    "response_mime_type": "text/plain",
}

# Define the model
model = genai.GenerativeModel(
    model_name="gemini-1.5-flash",
    generation_config=generation_config,
)

# Function to generate substeps recursively
def generate_substeps(reactants, products):
    prompt = f"Given reactants: {reactants} and products: {products}, break down the reaction mechanism into simple steps and provide in JSON format."
    
    chat_session = model.start_chat(history=[])
    response = chat_session.send_message(prompt)
    
    # Extract the JSON output
    try:
        steps = json.loads(response.text)
    except json.JSONDecodeError:
        steps = {"error": "Failed to decode JSON from Gemini response."}
    
    # Recursively break down each step if possible
    for step in steps:
        step_reactants = steps[step][0]  # reactants in this step
        step_products = steps[step][1]   # products in this step
        reason = steps[step][2]          # reason in this step
        
        if reason != "found through experiment":
            substeps = generate_substeps(step_reactants, step_products)
            steps[step].append({"substeps": substeps})
    
    return steps

# Gradio interface
def process_reaction(reactants, products):
    steps = generate_substeps(reactants, products)
    return json.dumps(steps, indent=4)

# Create the Gradio interface
iface = gr.Interface(
    fn=process_reaction,
    inputs=[gr.Textbox(label="Reactants (comma-separated)"), gr.Textbox(label="Products (comma-separated)")],
    outputs="json",
    title="Chemistry Rationalizer",
    description="Break down a reaction mechanism into recursive steps."
)

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