pdftoppm zero-pads the page number based on total page count:
- <10 pages: page-1.png
- <100 pages: page-01.png
- <1000 pages: page-001.png
The code hardcoded 'page-1.png' and 'page-N.png', which fails for any
multi-page document. Use filepath.Glob('page-*.png') to find the actual
output regardless of padding width.
Fixed in both ConvertToImage() (first-page preview) and the multi-page
OCR loop in ProcessDocument().
SearchDocuments excludes full_text for performance. The MD endpoint
needs the actual OCR content, not just the summary.
Added SearchDocumentsWithFullText() and SearchDocumentsWithFullTextFallback()
that select full_text explicitly. apiSearchMDHandler now uses these,
so format=md returns the complete OCR/markdown text for each document.
New endpoint returns all matching documents as concatenated plain-text
markdown, one section per document separated by ---.
Format:
# Document: {title}
ID: {id} | Category: {category} | Date: {date} | Vendor: {vendor}
{full_text or summary}
---
Parameters:
q - search query (required)
format - must be 'md' (required; distinguishes from HTML search)
Uses same FTS5 search as existing endpoints, limit raised to 200.
Falls back to LIKE search if FTS5 fails. Returns text/markdown content type.
POST /api/search (HTML partial) unchanged.
Replaces kimi-k2p5 for all vision tasks. K2.5 was outputting chain-of-thought
reasoning instead of JSON for non-English docs, requiring a fallback path.
qwen3-vl works first try, no retry needed, preserves original language correctly.
- Add 'DO NOT translate, preserve original language' to vision prompts
- Shorter/tighter JSON prompt to reduce K2.5 reasoning verbosity
- Fallback: when AnalyzeWithVision returns no JSON, do AnalyzePageOnly (plain text) then AnalyzeText (classify)
- Fallback to AnalyzePageOnly for single-page PDFs with empty/placeholder full_text
- Switch model back to kimi-k2p5 (only vision model on this Fireworks account)
- Build with CGO_ENABLED=1 -tags fts5 (required for SQLite FTS5)
- servePDF now supports ?download=1 query param
- Looks up document title and uses it as the Content-Disposition filename
- Download button on document page triggers actual download (not tab open)
- Added sanitizeFilename helper for safe Content-Disposition values
- Share table with random tokens and optional expiry (default 7 days)
- Public /s/{token} endpoint serves PDF directly
- Share/revoke UI on document page with copy-to-clipboard
- Caddy reverse proxy configured at docs.jongsma.me
James