app.py

# ===============================
# Derma Space: Dataset + Gradio Chatbot
# ===============================

import json
import random
import os
import gradio as gr
from datasets import load_dataset
from huggingface_hub import HfApi, login, upload_file

# ------------------------
# 1️⃣ ورود با Secret
# ------------------------
hf_token = os.environ.get("HF_TOKEN", None)
if hf_token is None:
    raise ValueError("HF_TOKEN not found in Secrets. Please add it in Space settings.")
login(token=hf_token)

# ------------------------
# 2️⃣ ساخت دیتاست ترکیبی امن
# ------------------------
def build_dataset():
    print("Creating a small general dataset...")
    general_examples = [
        {"domain":"general", "context":"Hello, how are you?", "response":"I'm good, thank you!"},
        {"domain":"general", "context":"What's your name?", "response":"I'm Derma ChatBot."}
    ]

    # ----- Dermatology QA (Mreeb)
    print("Loading Dermatology QA (Mreeb)...")
    derma = load_dataset("Mreeb/Dermatology-Question-Answer-Dataset-For-Fine-Tuning")['train']
    print("Columns in Mreeb dataset:", derma.column_names)

    derma_examples = []
    for item in derma:
        q = item.get('prompt')  # Mreeb uses 'prompt' instead of 'question'
        a = item.get('response')
        if q and a:
            derma_examples.append({"domain":"dermatology","context":q,"response":a})

    all_examples = general_examples + derma_examples
    random.shuffle(all_examples)

    # ذخیره به JSONL
    output_file = "derma_chat_mix.jsonl"
    with open(output_file, 'w', encoding='utf-8') as f:
        for ex in all_examples:
            f.write(json.dumps(ex, ensure_ascii=False) + "\n")
    print(f"✅ Dataset saved locally as {output_file} ({len(all_examples)} examples)")

    # آپلود به HF
    repo_id = "username/Derma"  # تغییر بده به نام کاربری خودت
    api = HfApi()
    api.create_repo(repo_id=repo_id, repo_type="dataset", exist_ok=True)
    upload_file(
        path_or_fileobj=output_file,
        path_in_repo=output_file,
        repo_id=repo_id,
        repo_type="dataset",
        commit_message="Initial upload of text-based chat dataset"
    )
    print(f"✅ Dataset uploaded: https://huggingface.co/datasets/{repo_id}")

# ------------------------
# 3️⃣ چت ساده با Gradio
# ------------------------
def simple_chat(user_input):
    with open("derma_chat_mix.jsonl", 'r', encoding='utf-8') as f:
        data = [json.loads(line) for line in f]

    best_match = None
    max_overlap = 0
    for item in data:
        overlap = len(set(user_input.lower().split()) & set(item['context'].lower().split()))
        if overlap > max_overlap:
            max_overlap = overlap
            best_match = item['response']

    if best_match:
        return best_match
    else:
        return "Sorry, I don't have a good answer for that. Try another question!"

# ------------------------
# 4️⃣ راه‌اندازی Gradio
# ------------------------
iface = gr.Interface(
    fn=simple_chat,
    inputs=gr.Textbox(lines=2, placeholder="Ask about dermatology or chat casually..."),
    outputs=gr.Textbox(label="Derma ChatBot"),
    title="Derma ChatBot",
    description="A simple English chatbot combining general conversation + dermatology QA."
)

# ------------------------
# 5️⃣ اجرای دیتاست + رابط
# ------------------------
if __name__ == "__main__":
    if not os.path.exists("derma_chat_mix.jsonl"):
        build_dataset()
    iface.launch()