run.py va yakun

import sys
import uvicorn

from app.config import settings
from app.database import SessionLocal
from sqlalchemy import text


def tekshir():
    """Ishga tushirishdan oldin baza tayyorligini tekshiradi."""
    try:
        db = SessionLocal()
        n = db.execute(text(
            "SELECT count(embedding_local) FROM rag.chunks")).scalar()
        db.close()
    except Exception as e:
        print(f"[XATO] Bazaga ulanib bo'lmadi: {type(e).__name__}")
        print("       Avval ishga tushiring:  python bootstrap.py")
        return False

    if n == 0:
        print("[OGOH] Bazada embedding YO'Q — qidiruv bo'sh natija beradi.")
        print("       Ishga tushiring:  python -m scraper.embedder")
    else:
        print(f"       {n} ta vektor tayyor")
    return True


if __name__ == "__main__":
    print("=" * 56)
    print("RAG DEMO")
    print("=" * 56)
    print(f"  Rejim: {settings.rag_mode}")
    print(f"  Baza : {settings.db_name} @ {settings.db_host}")

    if not tekshir():
        sys.exit(1)

    print("  Manzil: http://localhost:8000")
    uvicorn.run("app.main:app", host="0.0.0.0", port=8000, reload=True)

# 1. Baza
python bootstrap.py

# yoki haqiqiy ma'lumot bilan:
python bootstrap.py --qayta --pdf hujjat.pdf
python bootstrap.py --qayta --url https://sayt.uz/sahifa

# 2. Vektorlar (Ollama ishlab turishi kerak)
python -m scraper.embedder

# 3. Quvurni ko'rish (dars uchun)
python demo_pipeline.py "Nihol narxi qancha?"

# 4. Ilova
python run.py

PDF / Sayt
    |  pdf_parser, web_scraper
    v
  Xom matn
    |  normalizer
    v
 Toza matn
    |  chunker  (overlap bilan)
    v
  Bo'laklar
    |  embedder  (bge-m3)
    v
 Vektorlar -> PostgreSQL + pgvector (HNSW indeks)


Savol
  |  embedder
  v
Vektor
  |  search  (<=> cosine, threshold)
  v
10 ta chunk
  |  reranker  (cross-encoder)
  v
 5 ta chunk
  |  generator  (prompt + LLM)
  v
Javob + manba