: Keep content logic in Jinja, layout in CSS (using @media print ), and generation pure Python. 2. Pattern: Zero-Copy PDF Merging with pypdf (formerly PyPDF2) The Impact : Merge hundreds of PDFs without memory explosion.
Combine asyncio.to_thread for CPU-bound PDF generation:
# pyproject.toml (workspace root) [project] name = "pdf-power-hub" [tool.uv.workspace] members = ["extractors/ ", "generators/ ", "signers/*"] : Keep content logic in Jinja, layout in
:
– Use pikepdf + xmltodict :
@app.post("/report") async def create_report(data: dict, background_tasks: BackgroundTasks): # offload to thread pool pdf_bytes = await asyncio.to_thread(_generate_report_sync, data) background_tasks.add_task(log_pdf_generation, data["id"]) return Response(pdf_bytes, media_type="application/pdf")
Two standards exist: (simple) and XFA (XML-based, dynamic). Modern Python handles both. : Keep content logic in Jinja
import pikepdf with pikepdf.open("document.pdf") as pdf: pdf.convert_to_pdfa( version="2b", output_intent=srgb_intent, attach_output_intent=True ) pdf.save("archival.pdf", compress_streams=True)