// the find
metrico/gigapipe
⭐️ The Open-Source Polyglot Observability Warehouse: Light, Fast, Cloud Native, Drop-in Grafana LGTMP alternative :rocket: Indie All-in-One Opentelemetry, Loki, Prometheus, Tempo, Pyroscope On-Prem Alternative :star:
Gigapipe is a single Go binary that speaks Loki, Prometheus, Tempo, and Pyroscope APIs simultaneously, storing everything in ClickHouse. It's aimed at teams running the Grafana LGTM stack who want to cut the operational cost of running four separate systems without changing their instrumentation or dashboards.
ClickHouse as the storage layer is the right call — its columnar compression and vectorized execution handle high-cardinality log and metric workloads better than the purpose-built stores it replaces. The query transpilers are real implementations: LogQL, PromQL, and TraceQL each get their own parser and planner tree that generates native ClickHouse SQL rather than shelling out to the original binaries. The plugin system in reader/plugins means you can swap storage backends (DuckDB, GigAPI/S3) without touching query logic. Actively maintained — last push was yesterday, CI is green, and the go-version-sync workflow keeps dependencies current.
AGPL-3.0 is a hard stop for any commercial product that can't open-source its infrastructure layer — this is a bigger adoption blocker than the README acknowledges. The PromQL path routes through a WASM build of the Prometheus engine rather than a native Go transpiler, which adds a non-obvious runtime dependency and makes performance debugging harder than the pure-Go LogQL/TraceQL paths. Documentation lives on an external hosted site (gigapipe.com/docs/oss) that you don't control and can go down — the in-repo docs cover Docker quickstart and the profiling API but nothing about schema internals, retention, or cluster mode. The ruler/alerting subsystem exists but has no documentation at all beyond the test files, so you're reading code to figure out what alert evaluation actually does.