zeekay 2d0675063a extract: initial v0.1.0 ...

Document text extraction service. Drop-in replacement for Huly's
rekoni — listens on :4004 and serves /v1/extract, /v1/extract/url,
/v1/health.

- Pure Go, no CGo, no shelled binaries
- Format detection by magic bytes (filename ignored)
- PDF via ledongthuc/pdf; DOCX/XLSX/PPTX via native ZIP+XML walk;
  HTML via x/net/html with script/style strip; RTF via small native
  parser; plain text and markdown pass-through
- Bearer JWT auth against IAM JWKS (RS256/RS512), 10m cache
- 25 MB streaming upload cap, distroless static runtime
- Tests for every format (real fixtures in testdata/)

Image OCR is out of scope for v0.1 and returns HTTP 415.

2026-05-18 13:57:26 -07:00

1.0 KiB

Raw Permalink History

View Raw