app.py

import gradio as gr
from inference import evo_chat_predict, get_model_config, get_gpt_response
import pandas as pd
import csv
import os
from datetime import datetime

feedback_log = []

with gr.Blocks(theme=gr.themes.Base(), css="body { background-color: #0f0f0f; color: #f5f5f5; }") as demo:
    with gr.Column():
        gr.HTML("""
        <div style="padding: 10px; border-radius: 12px; background: #1f1f2e; color: #fff; font-size: 16px; margin-bottom: 12px;">
          <b>Why Evo?</b> 🚀 Evo is not just another AI. It evolves. It learns from you. It adapts its architecture live based on feedback. No retraining labs, no frozen weights. This is <u>live reasoning meets evolution</u>. <span style="color:#88ffcc">Built to outperform, built to survive.</span>
        </div>
        """)

    with gr.Row():
        with gr.Column():
            query = gr.Textbox(label="🧠 Your Question", placeholder="e.g. What should you do if there’s a fire?")
            option1 = gr.Textbox(label="❌ Option 1", placeholder="Enter the first option")
            option2 = gr.Textbox(label="❌ Option 2", placeholder="Enter the second option")
            feedback = gr.Radio(["Evo", "GPT"], label="🧠 Who was better?", info="Optional – fuels evolution", interactive=True)
            evo_btn = gr.Button("⚡ Ask Evo", elem_id="evo-btn")
            retrain_btn = gr.Button("🔁 Retrain Evo", elem_id="retrain-btn")
            clear_btn = gr.Button("🧹 Clear")
            export_btn = gr.Button("📤 Export Feedback CSV")

        with gr.Column():
            evo_stats = gr.Textbox(label="📊 Evo Stats", interactive=False)
            evo_box = gr.Textbox(label="🧠 Evo", interactive=False)
            gpt_box = gr.Textbox(label="🤖 GPT-3.5", interactive=False)
            status_box = gr.Textbox(label="🔵 Status", interactive=False)

        convo = gr.Dataframe(
            headers=["Question", "Option 1", "Option 2", "Answer", "Confidence", "Reasoning", "Context"],
            interactive=False, wrap=True, label="📜 Conversation History"
        )

    # 🔍 Ask Evo
    def ask_evo(q, opt1, opt2, hist, selected):
        result = evo_chat_predict(hist, q, [opt1, opt2])
        evo_text = f"Answer: {result['answer']} (Confidence: {result['confidence']})\n\nReasoning: {result['reasoning']}"
        gpt_text = get_gpt_response(q)
        stats = get_model_config()
        stats_text = f"Layers: {stats['num_layers']} | Heads: {stats['num_heads']} | FFN: {stats['ffn_dim']} | Memory: {stats['memory_enabled']} | Phase: {stats['phase']} | Accuracy: {stats['accuracy']}"

        # Update history
        new_row = [q, opt1, opt2, result["answer"], result["confidence"], result["reasoning"], result["context_used"]]
        new_row_df = pd.DataFrame([new_row], columns=hist.columns)
        updated_df = new_row_df if hist.empty else pd.concat([hist, new_row_df], ignore_index=True)

        # Log feedback
        if selected in ["Evo", "GPT"]:
            feedback_log.append(new_row)

        return evo_text, gpt_text, stats_text, updated_df

    # 🔁 Retrain Evo using retrain.py
    def retrain_evo():
        if not feedback_log:
            return "⚠️ No feedback data to retrain from."

        # Save log to feedback_log.csv in required format
        with open("feedback_log.csv", "w", newline="") as f:
            writer = csv.writer(f)
            writer.writerow(["prompt", "context", "label"])
            for row in feedback_log:
                prompt = row[0]
                context = row[6]  # context_used
                label = 0 if row[3] == row[1] else 1  # 0 if answer == option1 else 1
                writer.writerow([prompt, context, label])

        os.system("python retrain.py")
        return "✅ Evo retrained from feedback_log.csv."

    # 🧹 Clear UI
    def clear_fields():
        feedback_log.clear()
        return "", "", "", "", "", pd.DataFrame(columns=["Question", "Option 1", "Option 2", "Answer", "Confidence", "Reasoning", "Context"])

    # 📤 Export feedback
    def log_feedback_to_csv():
        if feedback_log:
            timestamp = datetime.now().strftime("%Y%m%d_%H%M%S")
            filepath = f"feedback_{timestamp}.csv"
            with open(filepath, "w", newline="") as f:
                writer = csv.writer(f)
                writer.writerow(["Question", "Option 1", "Option 2", "Answer", "Confidence", "Reasoning", "Context"])
                writer.writerows(feedback_log)
            return f"✅ Feedback exported to {filepath}"
        else:
            return "⚠️ No feedback to export."

    # 🔘 Event bindings
    evo_btn.click(fn=ask_evo, inputs=[query, option1, option2, convo, feedback], outputs=[evo_box, gpt_box, evo_stats, convo])
    retrain_btn.click(fn=retrain_evo, inputs=[], outputs=[status_box])
    clear_btn.click(fn=clear_fields, inputs=[], outputs=[query, option1, option2, evo_box, gpt_box, convo])
    export_btn.click(fn=log_feedback_to_csv, inputs=[], outputs=[status_box])

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