- Shipped
- September 27, 2026 at 9:25 PM UTC
- Author
- Kamo
- Commit
- 92034f8
docs_search, docs_read, translate_text and ocr_document (spec 4.20). A document is routed by DocsService's description before any byte is fetched: text is decoded here (a Markdown or YAML file stored as octet-stream by its extension), the PDFs and images Kamo OCR reads go to AIService F2, a document of more than five pages is refused before its download, and anything else is not readable without one. Document ids are canonical longs; language codes are checked as F1 checks them.