Introduction
What Tachyon is, what it isn't, and where to go next.
Tachyon is an open-source, typo-tolerant full-text search engine, distributed as a single binary. It adds BM25 relevance ranking, typo tolerance, filters, facets, sorting, and autocomplete to an application in under five minutes, without running a JVM cluster or a distributed coordination layer.
Tachyon is not a vector database and not a RAG engine. It does one thing: fast, self-hosted lexical search.
Before you deploy
Tachyon is under active development, and every feature described in these docs is tested end to end. Read Persistence for the exact durability guarantees before relying on it for data you can't afford to lose.
What it does
- Relevance — BM25, per-field boosts, typo tolerance
- Filters —
=,!=,<,<=,>,>=, ranges, set membership,&&,|| - Facets — counted over the whole result set, not just the current page
- Sorting — any numeric field, plus relevance (
_text_match), multi-clause - Autocomplete — prefix and typo-tolerant, ordered by term frequency
- Persistence — write-ahead log plus mmap'd on-disk segments, crash-safe
- Operations — Prometheus metrics, API-key auth, a REST API and nothing else
Where to go next
- Getting Started — requirements and what to read first
- Installation — run Tachyon with Docker or build from source
- Quickstart — create a collection, index documents, search
- Concepts — collections, documents, fields, segments
- API Reference — every endpoint, request, and response shape