Canonical ticket: docs/tickets/T-065-interpret-office-natively-without-pdf-render.md
Every docx/xlsx/pptx is ingested by rendering it to PDF with LibreOffice headless (office.py, T-054) and then running the PDF pipeline — there is no path that interprets the OOXML directly. Several downstream limitations (T-056 page explosion, borderless flatten, dropped comments/notes, the ~1 GB image variant) are consequences of how we ingest Office, not of the formats themselves. Exploratory, not a committed direction — "keep the render" (Option 4) is a fully legitimate outcome.
Canonical ticket: docs/tickets/T-065-interpret-office-natively-without-pdf-render.md
Every docx/xlsx/pptx is ingested by rendering it to PDF with LibreOffice headless (
office.py, T-054) and then running the PDF pipeline — there is no path that interprets the OOXML directly. Several downstream limitations (T-056 page explosion, borderless flatten, dropped comments/notes, the ~1 GB image variant) are consequences of how we ingest Office, not of the formats themselves. Exploratory, not a committed direction — "keep the render" (Option 4) is a fully legitimate outcome.