We build tools for regulatory compliance.
Our open-source libraries power document processing and machine learning inference in Go — released from work on the BeneDoc platform.
Pure Go PDF library — zero CGO, no external dependencies. Covers the full document lifecycle:
- XFA forms — parse XFA into a typed form schema (fields, sections, rules, visibility logic); fill element-based and synthetic XFA datasets (including FDA eSTAR forms)
- AcroForms — create and fill text, checkbox, radio, listbox, combo, password, and multiline fields; generate appearance streams; flatten to static content
- Digital signatures — PKCS#7 (RSA + ECDSA), RFC 3161 TSA timestamps, LTV/DSS long-term validation, visible signature widgets
- Redaction — permanent content-stream redaction with opaque overlay; metadata scrubbing
- PDF/A compliance — validate and auto-repair PDF/A-1b/2b/3b (XMP metadata, ICC output intent, font embedding checks)
- Manipulation — encrypt/decrypt (AES-128), repair/linearize, text stamping, page numbering, merge, split, file attachments
- Extraction — tables (with merged cells and header detection), text, images, form field values
Web renderer for XFA PDF forms — Go backend + React/TypeScript frontend. Drop in a PDF and get a fully interactive form:
- Renders all XFA field types: text, numeric, date/time, password, checkbox, radio, listbox, multi-select, file upload, signature, separator
- Section-based navigation with per-section completion tracking
- Rules engine — three-valued Kleene logic evaluates XFA script rules for field visibility, value, and validation; handles compound
&&/||,rawValuepatterns, and.presencechecks - XML round-trip: export current values back into the XFA datasets stream; import previously saved XML
- File attachments embedded into the output PDF via pdfer
- Hot-reload dev setup: Air (Go) + Vite HMR (React) with live pdfer volume mount
Production-ready Go bindings for ONNX Runtime — no CGO required. Forked from shota3506/onnxruntime-purego and significantly extended with session pooling, IO binding, LoRA adapter support, profiling, prepacked weights sharing, global thread pools, and comprehensive ORT v23/v24 API coverage.