Aktiveeri venv

source ~/venvs/pdf-env/bin/activate cd ~/rag-demo/pdf-pipeline export $(grep -v '^#' .env | xargs -d '\n')

python -m src.db_schema # uuesti skeemiga

Vektor välja muudatus

./scripts/db_vector_välja_muutmine.sh

python -m src.extract_pdf # parandatud ekstraheerimisega python -m src.check_status # parema ülevaatega

python -m src.find_duplicates # duplikaatide kontroll python -m src.clean_and_normalize

python -m src.create_chunks python -m src.embed_chunks

Testimine

docker exec -it postgres_postgis psql -U osm -d pdf_research -h localhost -c "SELECT COUNT(*) FROM processed_documents;" docker exec -it postgres_postgis psql -U osm -d pdf_research -h localhost -c " SELECT id, raw_doc_id, page, LENGTH(content_text) AS len, has_table FROM processed_documents ORDER BY id LIMIT 20; "

Weaviate baasiga sünkroniseerimine

python -m src.sync_weaviate

python -m src.query_hybrid python -m src.generate_answer

API käivitamine

python -m src.rag_api

==============================================================

API päringute näidised

1. Interactive Docs (Swagger UI)

Ava brauseris:

http://localhost:8071/docs

2. cURL käsuga

# Lihtsustatud otsing
curl "http://localhost:8000/search-simple?q=young%20driver%20risk"
# POST requestiga (täielik kontroll)
curl -X POST "http://localhost:8071/search" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "young driver accident risk",
    "top_articles": 10,
    "top_chunks": 20,
    "temperature": 0.6,
    "max_tokens": 1500
  }' | jq '.'

3. Python requestiga

import requests
import json

# POST päring
response = requests.post(
    "http://localhost:8071/search",
    json={
        "query": "traffic flow prediction",
        "top_articles": 10,
        "top_chunks": 20,
        "temperature": 0.6,
        "max_tokens": 1500
    }
)

result = response.json()
print("Küsimus:", result["query"])
print("Vastus:", result["answer"])
print("Allikad:", result["articles"])

LOEMIND.md 2.0 KB Історія Запис