// the find
lnx-search/lnx
A flexible, performant and reliable search database without the AI bullshit.
lnx wraps the tantivy search engine behind a REST API, giving you a self-hosted Elasticsearch alternative without the JVM overhead. It targets developers who need fast full-text search with fuzzy matching and typo tolerance but don't want to run a distributed system. Single-node only, schema-full, configuration-heavy.
Built on tantivy and tokio, so the performance numbers are credible — millisecond indexing on large datasets is realistic, not marketing. Pre-computed fast-fuzzy queries are a genuine differentiator: spell correction gets baked in at index time rather than computed per query. Fine-grained thread pool tuning per index is unusual and useful when you're running multiple indexes with different access patterns. The query type coverage is solid — fuzzy, More-Like-This, regex, phrase prefix, nested boolean — you won't hit a wall on common search patterns.
The repo is effectively stalled: last push October 2025, latest stable is 0.9.0, and the README opens with a warning to not use the main branch. That's a bad sign for anything you'd put in production. No distributed mode means you're one machine; if that box falls over, search is down. No metrics endpoint makes it invisible to standard observability stacks. The schema-full requirement adds friction compared to Meilisearch — you can't just throw documents at it and iterate.