tools/pdf.py

"""PDF generation from Markdown using markdown-pdf."""
import logging
import re
from io import BytesIO
from pathlib import Path
from typing import Optional

from markdown_pdf import MarkdownPdf, Section

_PROJECT_ROOT = Path(__file__).resolve().parent.parent

# markdown_pdf (via pymupdf) resolves every link target at build time. A link
# with an empty href or a dangling "#anchor" raises
# RuntimeError: No destination with id=...
# The LLM sometimes produces such placeholders (e.g. `[contact]()`,
# `[lien](#)`, `[foo](mailto:)`). We strip them before building the PDF.
_EMPTY_HREF_RE = re.compile(r"\[([^\]]*)\]\(\s*\)")
_HASH_HREF_RE = re.compile(r"\[([^\]]*)\]\(#[^)]*\)")
_EMPTY_SCHEME_HREF_RE = re.compile(
    r"\[([^\]]*)\]\((?:mailto|tel|https?|ftp):\s*\)",
    re.IGNORECASE,
)
_ANY_LINK_RE = re.compile(r"\[([^\]]*)\]\([^)]*\)")


def _sanitize_markdown_for_pdf(md: str) -> str:
    """Remove link patterns that confuse pymupdf's anchor resolver."""
    md = _EMPTY_HREF_RE.sub(r"\1", md)
    md = _HASH_HREF_RE.sub(r"\1", md)
    md = _EMPTY_SCHEME_HREF_RE.sub(r"\1", md)
    return md


def _strip_all_links(md: str) -> str:
    """Last-resort fallback: turn every `[text](url)` into plain `text`."""
    return _ANY_LINK_RE.sub(r"\1", md)


def _build_pdf(md_text: str, toc_level: int, logo_path: Optional[str]) -> BytesIO:
    pdf = MarkdownPdf(toc_level=toc_level, optimize=True)

    logo_file = _PROJECT_ROOT / logo_path if logo_path else None
    if logo_file and logo_file.is_file():
        logo_md = f"![Logo]({logo_path})"
        logo_css = "img { display: block; margin-left: auto; margin-right: auto; }"
        logo_section = Section(logo_md, toc=False)
        logo_section.root = str(_PROJECT_ROOT)
        pdf.add_section(logo_section, user_css=logo_css)

    content_css = (
        "body, p, li, ul, ol, td, th { font-size: 9pt; } "
        "h1 { font-size: 14pt; } h2 { font-size: 12pt; } h3 { font-size: 10pt; }"
    )
    content_section = Section(md_text)
    content_section.root = str(_PROJECT_ROOT)
    pdf.add_section(content_section, user_css=content_css)

    buffer = BytesIO()
    pdf.save_bytes(buffer)
    buffer.seek(0)
    return buffer


def markdown_to_pdf(
    md_text: str,
    toc_level: int = 2,
    logo_path: Optional[str] = None,
) -> BytesIO:
    """Convert Markdown to PDF using markdown-pdf (PyMuPDF + markdown-it-py).

    Supports UTF-8, tables, links, images, and TOC from headings.
    Optionally adds a dedicated logo section at the top (centered) if logo_path is set.
    Links with empty or dangling hrefs are stripped to avoid pymupdf errors; if
    the build still fails, every link is converted to plain text as a fallback.
    """
    if logo_path is None:
        try:
            from config.settings import settings
            logo_path = getattr(settings, "pdf_logo_path", None)
        except Exception:
            logo_path = None

    sanitized = _sanitize_markdown_for_pdf(md_text)
    try:
        return _build_pdf(sanitized, toc_level, logo_path)
    except RuntimeError as exc:
        logging.warning(
            "markdown_to_pdf: retrying without links after pymupdf error: %s", exc
        )
        return _build_pdf(_strip_all_links(sanitized), toc_level, logo_path)