main.py

from flask import Flask, request, jsonify
from sentence_transformers import SentenceTransformer
from pinecone import Pinecone
import google.generativeai as genai
from langdetect import detect

# تهيئة الموديلات
embedding_model = SentenceTransformer("intfloat/multilingual-e5-large")

pc = Pinecone(api_key="pcsk_3ax4D8_PH7vWF1KWAMRpyjmEnXhwxswmHSjvqgwovna3xGGbfsgZsMRtRyFi9uCpPyi4B9")
index = pc.Index("newindex")

genai.configure(api_key="AIzaSyBXtRzMkpm9RNDO09A9N3XoG_vfjgUe5Vw")
model = genai.GenerativeModel("gemini-2.0-flash")

app = Flask(__name__)
chat_history = []

def detect_language(text):
    try:
        return detect(text)
    except:
        return "unknown"

def get_answer_from_pinecone(user_question, embedding_model, index, top_k=5, similarity_threshold=0.7):
    try:
        question_vector = embedding_model.encode(user_question).tolist()
    except Exception as e:
        return [f"❌ Error embedding question: {e}"]

    try:
        search_result = index.query(
            vector=question_vector,
            top_k=top_k,
            include_metadata=True
        )
    except Exception as e:
        return [f"❌ Error querying Pinecone: {e}"]

    matches = [m for m in search_result.matches if m.score >= similarity_threshold]
    sorted_matches = sorted(matches, key=lambda x: x.score, reverse=True)

    answers = []
    for m in sorted_matches:
        answer = m.metadata.get('answer', '').strip()
        source = m.metadata.get('source', 'unknown')
        score = round(m.score, 3)
        if answer:
            answers.append(f"• ({score}) from [{source}]:\n{answer}")
    return answers if answers else ["⚠️ No similar answers found."]

def ask_gemini_with_combined_answer(user_question, pinecone_answers=[], history=[]):
    context = "\n".join([f"👤 {q}\n🤖 {a}" for q, a in history])
    extracted_info = "\n".join([f"• {ans}" for ans in pinecone_answers]) if pinecone_answers else "None"
    lang = detect_language(user_question)

    if lang == "ar":
        greeting = "مرحبًا! 😊"
        instructions = """
❗ هام: استخدم فقط المعلومات من قاعدة البيانات.

📜 المحادثة السابقة:
{context}

👤 المستخدم يسأل: {user_question}
📚 معلومات من قاعدة البيانات:
{extracted_info}

📌 الرد:
"""
    else:
        greeting = "Hello! 😊"
        instructions = """
❗ Important: Use only database information.

📜 Previous conversation:
{context}

👤 User asks: {user_question}
📚 Retrieved info:
{extracted_info}

📌 Response:
"""

    prompt = instructions.format(
        context=context or ("لا يوجد" if lang == "ar" else "None"),
        user_question=user_question,
        extracted_info=extracted_info
    )
    response = model.generate_content(prompt)
    return f"{greeting}\n{response.text.strip()}"

@app.route("/ask", methods=["POST"])
def ask():
    data = request.json
    question = data.get("question")
    if not question:
        return jsonify({"error": "Missing question"}), 400

    pinecone_answer = get_answer_from_pinecone(question, embedding_model, index)
    final_answer = ask_gemini_with_combined_answer(question, pinecone_answer, chat_history)
    chat_history.append((question, final_answer))

    return jsonify({
        "question": question,
        "pinecone_matches": pinecone_answer,
        "answer": final_answer
    })

@app.route("/")
def home():
    return "🤖 API is running. Use POST /ask with {'question': '...'}"

if __name__ == "__main__":
    app.run(host="0.0.0.0", port=7860)