// the find
mr-karan/logchef
Lightweight, single-binary log analytics interface for ClickHouse, focused on high-performance querying and visualization
Logchef is a self-hosted log analytics UI that sits in front of ClickHouse, giving you a query interface, alerting, and team access control without standing up an Elastic stack. It targets small-to-medium engineering teams who already have or are willing to run ClickHouse and want something lighter than Grafana Loki or a SaaS log product.
The schema-agnostic approach is the standout design choice — point it at any ClickHouse table and it works without forcing a specific ingestion schema, which matters a lot when you're retrofitting it onto existing data. The pluggable metadata store (SQLite default, Postgres for HA) is a pragmatic call that lets you start with zero ops overhead and scale when needed. The CLI is written in Rust and ships as a separate binary, so it's actually fast and doesn't pull the whole Go server as a dependency. OIDC + RBAC out of the box means you're not bolting on auth as an afterthought.
ClickHouse is a hard prerequisite with real operational weight — teams without ClickHouse experience will spend more time managing the database than using the log tool. The AI SQL generation is only as good as your schema documentation and the model's ClickHouse knowledge; LogchefQL exists alongside raw SQL, which means two query dialects to learn and two surfaces to maintain. At 758 stars and 48 forks, the contributor base is thin, so if mr-karan loses interest this stalls fast. No native log ingestion — you still need Vector, Fluent Bit, or something else to get logs into ClickHouse, and the integration story for that is just docs pointing at external tools.