Xtriever
github.com/mirth/xtriever- Category
- Developer Tools
- Rank
- No. 1904Tools index
Previous survey · No. 1870 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Data, Retrieval & Knowledge
- GitHub
- 5 stars
- Date
About
A hybrid retrieval engine for retrieval-augmented generation, written in Rust, built to run entirely on-device (iPhone, Android, or laptop) with no server or network connection required. It combines BM25 lexical search, a dense MiniLM embedding index, and a cross-encoder re-ranker through reciprocal rank fusion, and exposes bindings for Rust, Python, Swift, and Kotlin along with demo iOS and Android apps.
At a glance
Xtriever combines keyword and embedding search, then reranks candidates locally. It is worth studying if you need to search documents offline on a phone or laptop and can build an early-stage prototype.
Good fit
Developers evaluating local retrieval before adding an answer-generation step.
Consider something else when
Choose an established library if you need published packages or a supported production integration.
Start here
Index a small public document set on your target device. Compare exact-term and paraphrased queries against a keyword baseline and record latency and storage use.
Cost and requirements
The documented version builds from source. The Rust engine has Python, Swift and Kotlin interfaces and uses local model assets for embedding and reranking.
Limits to know
The repository describes an unpublished release and a planned learned ranker that has not been implemented. Test the full pipeline on your own hardware and documents.
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
Gives mobile and offline-first developers a genuine on-device retrieval engine instead of routing RAG queries through a server, useful for privacy-sensitive or connectivity-constrained agent applications.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.