Files
llm/scripts/build_1c_rag_index.py

28 lines
928 B
Python

from __future__ import annotations
import argparse
from pathlib import Path
from common import build_lexical_index, read_jsonl, write_json
ROOT = Path(__file__).resolve().parents[1]
DEFAULT_CORPUS = ROOT / "plugins" / "1c" / "datasets" / "prepared" / "rag_corpus.jsonl"
DEFAULT_INDEX = ROOT / "plugins" / "1c" / "datasets" / "prepared" / "rag_index.json"
def main() -> int:
parser = argparse.ArgumentParser(description="Build a simple lexical RAG index for the 1C corpus.")
parser.add_argument("--corpus", type=Path, default=DEFAULT_CORPUS)
parser.add_argument("--output", type=Path, default=DEFAULT_INDEX)
args = parser.parse_args()
records = read_jsonl(args.corpus)
index = build_lexical_index(records)
write_json(args.output, index)
print(f"Wrote index with {index['doc_count']} document chunk(s) to {args.output}")
return 0
if __name__ == "__main__":
raise SystemExit(main())