uni-xervo

A Rust workspace for local and remote ML model inference, and document extraction built on top of it.

Crates

Crate	Path	Description
uni-xervo	crates/uni-xervo	Unified runtime for local and remote models across pluggable providers (Candle, ONNX Runtime, mistral.rs, whisper.cpp, OpenAI, Gemini, …): dense / learned-sparse (SPLADE, BGE-M3) / multi-vector (ColBERT) / single-pass hybrid embeddings, reranking, generation, NLP, OCR, document extraction, and transcription.
uni-xervo-pdf	crates/uni-xervo-pdf	Optional companion: tiered PDF document extraction that escalates per page across native text → OCR → doc-VLM, with provenance and cross-tier verification.

Documentation

• Guides and reference: https://rustic-ai.github.io/uni-xervo (sources under website/).
• API reference (rustdoc): https://rustic-ai.github.io/uni-xervo/api/uni_xervo/.
• Changelog: CHANGELOG.md.

Development

./scripts/test.sh      # fmt check + cargo check/test across the workspace

Licensed under Apache-2.0 (LICENSE).