ingest_docs.py

# ingest_docs.py

import os
import json
import faiss
from sentence_transformers import SentenceTransformer

RAW_DOCS_DIR = "data/docs/raw"
INDEX_PATH = "data/docs/docs.index"
META_PATH = "data/docs/docs_meta.json"

CHUNK_SIZE = 400

def chunk_text(text, size):
    chunks = []
    for i in range(0, len(text), size):
        chunk = text[i:i+size].strip()
        if chunk:
            chunks.append(chunk)
    return chunks

def main():
    model = SentenceTransformer("paraphrase-MiniLM-L3-v2", cache_folder="./model_cache")

    documents = []
    metadata = []

    for fname in os.listdir(RAW_DOCS_DIR):
        path = os.path.join(RAW_DOCS_DIR, fname)
        with open(path, "r", encoding="utf-8") as f:
            text = f.read()

        chunks = chunk_text(text, CHUNK_SIZE)
        for chunk in chunks:
            documents.append(chunk)
            metadata.append({
                "source_file": fname,
                "source": f"https://www.jenkins.io/doc/"
            })

    embeddings = model.encode(documents)
    index = faiss.IndexFlatL2(embeddings.shape[1])
    index.add(embeddings)

    os.makedirs("data/docs", exist_ok=True)
    faiss.write_index(index, INDEX_PATH)

    with open(META_PATH, "w", encoding="utf-8") as f:
        json.dump(metadata, f, indent=2)

    print(f"Ingested {len(documents)} document chunks.")

if __name__ == "__main__":
    main()