Pipeline crawl du lieu Chotot, lam sach thanh lakehouse/SQLite, roi hoi du lieu bang Text2SQL agent chay Vertex AI.
pip install -r requirements.txtTao/cap nhat .env o root:
GOOGLE_CLOUD_PROJECT=your-gcp-project
GOOGLE_CLOUD_LOCATION=us-central1
TEXT2SQL_MODEL=gemini-2.5-flash
TEXT2SQL_DB_PATH=data/marketplace.dbDang nhap Google Cloud neu dung Vertex AI:
gcloud auth application-default login
gcloud config set project your-gcp-projectpython crawler/main.pyTest it trang:
python crawler/main.py --page-end 1Raw JSON:
data/raw/category_id=<id>/date=<yyyy-mm-dd>/listings.json
python preprocessor/main.pyOutput:
data/clean/category_id=<id>/date=<yyyy-mm-dd>/*.csv
data/marketplace.db
python -m text2sql.cli "Co bao nhieu tin bat dong san o Ha Noi?"Cau phuc tap nen tang so vong:
python -m text2sql.cli "O Ha Noi, moi hang xe co bao nhieu tin dang?" --max-rounds 12API server:
python -m uvicorn text2sql.server:app --reload --host 127.0.0.1 --port 8000DB chinh: data/marketplace.db
listings: bang trung tam cho tin dang.categories: danh muc.listing_attributes: thuoc tinh bo sung dang key-value.real_estate_details: bat dong san.vehicle_details: xe.electronics_details: dien tu.product_details: san pham khac.job_details: viec lam.service_details: dich vu.pet_details: thu cung.