scripts/batch_ingest.py

#!/usr/bin/env python3
"""
scripts/batch_ingest.py
-----------------------
Batch ingest all PDFs from a source directory, generate metadata stubs,
and render first-page previews in one shot.

Usage:
    python scripts/batch_ingest.py --source /path/to/pdfs --category government_report
"""

from __future__ import annotations

import argparse
import json
import shutil
import uuid
from pathlib import Path

from rich.console import Console
from rich.progress import track

console = Console()

CATEGORIES = [
    "government_report",
    "government_form",
    "law",
    "gazette",
    "book",
    "research_paper",
    "manual",
    "annual_report",
    "financial_report",
    "certificate",
    "contract",
    "invoice",
    "receipt",
    "newspaper",
    "magazine",
    "presentation",
    "other",
]


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(description="Batch ingest PDFs into the corpus.")
    parser.add_argument("--source", required=True, type=Path, help="Source directory of PDFs.")
    parser.add_argument("--category", default="other", choices=CATEGORIES)
    parser.add_argument("--language", default="km", choices=["km", "en", "mixed"])
    parser.add_argument("--pdf-dir", default=Path("pdf"), type=Path)
    parser.add_argument("--meta-dir", default=Path("metadata"), type=Path)
    parser.add_argument("--preview-dir", default=Path("preview"), type=Path)
    parser.add_argument("--dpi", default=150, type=int)
    parser.add_argument("--no-preview", action="store_true")
    return parser.parse_args()


def main() -> None:
    args = parse_args()
    source: Path = args.source.resolve()
    pdf_dest: Path = (args.pdf_dir / args.category).resolve()
    meta_dest: Path = args.meta_dir.resolve()
    preview_dest: Path = args.preview_dir.resolve()

    pdf_dest.mkdir(parents=True, exist_ok=True)
    meta_dest.mkdir(parents=True, exist_ok=True)
    preview_dest.mkdir(parents=True, exist_ok=True)

    pdfs = list(source.rglob("*.pdf"))
    console.print(f"Found [cyan]{len(pdfs)}[/cyan] PDF(s) in [bold]{source}[/bold]")

    try:
        from pdf2image import convert_from_path  # type: ignore[import]

        has_pdf2image = True
    except ImportError:
        has_pdf2image = False
        if not args.no_preview:
            console.print("[yellow]pdf2image not found – skipping previews.[/yellow]")

    for pdf in track(pdfs, description="Processing…"):
        dest_pdf = pdf_dest / pdf.name
        if not dest_pdf.exists():
            shutil.copy2(pdf, dest_pdf)

        meta_file = meta_dest / (pdf.stem + ".json")
        if not meta_file.exists():
            stub = {
                "id": str(uuid.uuid4()),
                "language": args.language,
                "category": args.category,
                "pages": 0,
                "native_pdf": False,
                "scanned": False,
                "has_tables": False,
                "has_images": False,
                "has_header": False,
                "has_footer": False,
                "source": None,
                "license": None,
                "pdf_path": str(dest_pdf.relative_to(Path(".").resolve())),
                "preview_path": None,
            }
            meta_file.write_text(json.dumps(stub, indent=2, ensure_ascii=False), encoding="utf-8")

        if has_pdf2image and not args.no_preview:
            doc_preview = preview_dest / pdf.stem
            doc_preview.mkdir(parents=True, exist_ok=True)
            if not list(doc_preview.glob("*.png")):
                images = convert_from_path(dest_pdf, dpi=args.dpi, last_page=1)
                images[0].save(doc_preview / "page_0001.png", "PNG")

    console.print("[green]✓ Batch ingest complete.[/green]")


if __name__ == "__main__":
    main()