# app.py

'''import gradio as gr
import pandas as pd
from inference import (
    evo_chat_predict,
    get_gpt_response,
    get_model_config,
    get_system_stats,
    retrain_from_feedback_csv,
    load_model,
)
import os
import csv

FEEDBACK_LOG = "feedback_log.csv"

# 🧠 Ask Evo
def ask_evo(question, option1, option2, history, user_vote):
    options = [option1.strip(), option2.strip()]
    result = evo_chat_predict(history, question.strip(), options)

    # Create feedback_log.csv with headers if it doesn't exist
    if not os.path.exists(FEEDBACK_LOG):
        with open(FEEDBACK_LOG, "w", encoding="utf-8", newline="") as f:
            writer = csv.writer(f)
            writer.writerow(["question", "option1", "option2", "evo_answer", "confidence", "reasoning", "context", "vote"])

    row = {
        "question": question.strip(),
        "option1": option1.strip(),
        "option2": option2.strip(),
        "evo_answer": result["answer"],
        "confidence": result["confidence"],
        "reasoning": result["reasoning"],
        "context": result["context_used"],
        "vote": user_vote.strip() if user_vote else ""

    }

    # Log feedback
    with open(FEEDBACK_LOG, "a", newline='', encoding="utf-8") as f:
        writer = csv.DictWriter(f, fieldnames=row.keys())
        writer.writerow(row)

    # Prepare outputs
    evo_output = f"Answer: {row['evo_answer']} (Confidence: {row['confidence']})\n\nReasoning: {row['reasoning']}\n\nContext used: {row['context']}"
    gpt_output = get_gpt_response(question)
    history.append(row)

    stats = get_model_config()
    sys_stats = get_system_stats()

    stats_text = f"Layers: {stats.get('num_layers', '?')} | Heads: {stats.get('num_heads', '?')} | FFN: {stats.get('ffn_dim', '?')} | Memory: {stats.get('memory_enabled', '?')} | Accuracy: {stats.get('accuracy', '?')}"
    sys_text = f"Device: {sys_stats['device']} | CPU: {sys_stats['cpu_usage_percent']}% | RAM: {sys_stats['memory_used_gb']}GB / {sys_stats['memory_total_gb']}GB | GPU: {sys_stats['gpu_name']} ({sys_stats['gpu_memory_used_gb']}GB / {sys_stats['gpu_memory_total_gb']}GB)"

    return evo_output, gpt_output, stats_text, sys_text, history

# 🔁 Manual retrain button
def retrain_evo():
    msg = retrain_from_feedback_csv()
    load_model(force_reload=True)
    return msg

# 📤 Export feedback
def export_feedback():
    if not os.path.exists(FEEDBACK_LOG):
        return pd.DataFrame()
    return pd.read_csv(FEEDBACK_LOG)

# 🧹 Clear
def clear_all():
    return "", "", "", "", [], None

# 🖼️ UI
with gr.Blocks(title="🧠 Evo – Reasoning AI") as demo:
    gr.Markdown("## Why Evo? 🚀 Evo is not just another AI. It evolves. It learns from you. It adapts its architecture live based on feedback.\n\nNo retraining labs, no frozen weights. This is live reasoning meets evolution. Built to outperform, built to survive.")

    with gr.Row():
        question = gr.Textbox(label="🧠 Your Question", placeholder="e.g. Why is the sky blue?")
    with gr.Row():
        option1 = gr.Textbox(label="❌ Option 1")
        option2 = gr.Textbox(label="❌ Option 2")

    with gr.Row():
        with gr.Column():
            evo_ans = gr.Textbox(label="🧠 Evo", lines=6)
        with gr.Column():
            gpt_ans = gr.Textbox(label="🤖 GPT-3.5", lines=6)

    with gr.Row():
        stats = gr.Textbox(label="📊 Evo Stats")
        system = gr.Textbox(label="🔵 Status")

    evo_radio = gr.Radio(["Evo", "GPT"], label="🧠 Who was better?", info="Optional – fuels evolution")

    history = gr.State([])

    with gr.Row():
        ask_btn = gr.Button("⚡ Ask Evo")
        retrain_btn = gr.Button("🔁 Retrain Evo")
        clear_btn = gr.Button("🧹 Clear")
        export_btn = gr.Button("📤 Export Feedback CSV")

    export_table = gr.Dataframe(label="📜 Conversation History")

    ask_btn.click(fn=ask_evo, inputs=[question, option1, option2, history, evo_radio], outputs=[evo_ans, gpt_ans, stats, system, history])
    retrain_btn.click(fn=retrain_evo, inputs=[], outputs=[stats])
    clear_btn.click(fn=clear_all, inputs=[], outputs=[question, option1, option2, evo_ans, gpt_ans, stats, system, history, evo_radio])
    export_btn.click(fn=export_feedback, inputs=[], outputs=[export_table])

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

# app.py
# app.py

import gradio as gr
import pandas as pd
import os
import csv

from inference import (
    evo_chat_predict,
    get_gpt_response,
    get_model_config,
    get_system_stats,
    retrain_from_feedback_csv,
    load_model,
)

GENOME_LOG = "genome_log.csv"
FEEDBACK_LOG = "feedback_log.csv"

# 🧠 Ask Evo
def ask_evo(question, option1, option2, history, user_vote):
    options = [option1.strip(), option2.strip()]
    result = evo_chat_predict(history, question.strip(), options)

    # Create feedback_log.csv if it doesn't exist
    if not os.path.exists(FEEDBACK_LOG):
        with open(FEEDBACK_LOG, "w", encoding="utf-8", newline="") as f:
            writer = csv.writer(f)
            writer.writerow(["question", "option1", "option2", "evo_answer", "confidence", "reasoning", "context", "vote"])

    row = {
        "question": question.strip(),
        "option1": option1.strip(),
        "option2": option2.strip(),
        "evo_answer": result["answer"],
        "confidence": result["confidence"],
        "reasoning": result["reasoning"],
        "context": result["context_used"],
        "vote": user_vote.strip() if user_vote else ""
    }

    with open(FEEDBACK_LOG, "a", newline='', encoding="utf-8") as f:
        writer = csv.DictWriter(f, fieldnames=row.keys())
        writer.writerow(row)

    evo_output = f"Answer: {row['evo_answer']} (Confidence: {row['confidence']})\n\nReasoning: {row['reasoning']}\n\nContext used: {row['context']}"
    gpt_output = get_gpt_response(question)
    history.append(row)

    stats = get_model_config()
    sys_stats = get_system_stats()

    stats_text = f"Layers: {stats.get('num_layers', '?')} | Heads: {stats.get('num_heads', '?')} | FFN: {stats.get('ffn_dim', '?')} | Memory: {stats.get('memory_enabled', '?')} | Accuracy: {stats.get('accuracy', '?')}"
    sys_text = f"Device: {sys_stats['device']} | CPU: {sys_stats['cpu_usage_percent']}% | RAM: {sys_stats['memory_used_gb']}GB / {sys_stats['memory_total_gb']}GB | GPU: {sys_stats['gpu_name']} ({sys_stats['gpu_memory_used_gb']}GB / {sys_stats['gpu_memory_total_gb']}GB)"

    genome_df = get_top_genomes()
    return evo_output, gpt_output, stats_text, sys_text, history, genome_df

# 📊 Top genome stats
def get_top_genomes(n=5):
    if not os.path.exists(GENOME_LOG):
        return pd.DataFrame()
    try:
        df = pd.read_csv(GENOME_LOG)
        if "score" in df.columns:
            df = df.sort_values(by="score", ascending=False)
        return df.tail(n)
    except Exception:
        return pd.DataFrame()

# 🔁 Manual retrain button
def retrain_evo():
    msg = retrain_from_feedback_csv()
    load_model(force_reload=True)
    return msg

# 📤 Export feedback
def export_feedback():
    if not os.path.exists(FEEDBACK_LOG):
        return pd.DataFrame()
    return pd.read_csv(FEEDBACK_LOG)

# 🧹 Clear UI
def clear_all():
    return "", "", "", "", "", "", pd.DataFrame(), {}, pd.DataFrame()

# 🖼️ UI Layout
with gr.Blocks(title="🧠 Evo – Reasoning AI") as demo:
    gr.Markdown("## 🚀 Evo is not just another AI. It evolves. It learns from you. It mutates based on feedback.\n\nNo retraining labs. No frozen weights. This is live reasoning meets evolution.")

    with gr.Row():
        question = gr.Textbox(label="🧠 Your Question", placeholder="e.g. Why is the sky blue?")
    with gr.Row():
        option1 = gr.Textbox(label="❌ Option 1")
        option2 = gr.Textbox(label="❌ Option 2")

    with gr.Row():
        with gr.Column():
            evo_ans = gr.Textbox(label="🧠 Evo", lines=6)
        with gr.Column():
            gpt_ans = gr.Textbox(label="🤖 GPT-3.5", lines=6)

    with gr.Row():
        stats = gr.Textbox(label="📊 Evo Stats")
        system = gr.Textbox(label="🔵 System Status")

    evo_radio = gr.Radio(["Evo", "GPT"], label="🧠 Who was better?", info="Optional – leave blank if both were wrong")

    history = gr.State([])

    with gr.Row():
        ask_btn = gr.Button("⚡ Ask Evo")
        retrain_btn = gr.Button("🔁 Retrain Evo")
        clear_btn = gr.Button("🧹 Clear")
        export_btn = gr.Button("📤 Export Feedback CSV")

    export_table = gr.Dataframe(label="📜 Conversation History")
    genome_table = gr.Dataframe(label="🧬 Top Genomes")

    ask_btn.click(
        fn=ask_evo,
        inputs=[question, option1, option2, history, evo_radio],
        outputs=[evo_ans, gpt_ans, stats, system, history, genome_table]
    )

    retrain_btn.click(fn=retrain_evo, inputs=[], outputs=[stats])

    clear_btn.click(
        fn=clear_all,
        inputs=[],
        outputs=[
            question, option1, option2, evo_ans, gpt_ans,
            stats, export_table, system, genome_table
        ]
    )

    export_btn.click(fn=export_feedback, inputs=[], outputs=[export_table])

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