// the find
tracewayapp/traceway
The only tool you need to know what is happening and how to fix it.
Traceway is a self-hosted observability platform that ingests OpenTelemetry data natively — no Collector required — and combines traces, logs, metrics, session replay, exception grouping, and AI observability in one system. It runs on Go with ClickHouse + Postgres in standalone mode, or SQLite for embedded use inside a Go app. Aimed at teams who want Datadog-level features without the per-event pricing or vendor lock-in.
Native OTLP/HTTP ingest means you can point any OTel SDK at it without a Collector or vendor shim in between. The JavaScript symbolicator is genuinely well-engineered: source maps compile once to a binary `.tw` format, are memory-mapped, and do sub-millisecond lookups on a cold cache — this is the kind of detail that separates a serious tool from a weekend project. The embedded mode (SQLite, zero external deps, runs inside your Go process) is a real differentiator for local dev and small deployments. Build tags to switch storage backends (SQLite vs ClickHouse + Postgres) without forking the codebase is clean architecture.
855 stars and only 25 forks on a project this ambitious is a thin contributor base — the bus factor is essentially one team, and the migration history (64 ClickHouse migrations, schema reshuffling still ongoing) suggests the data model is still settling. The AI observability feature currently only integrates with OpenRouter; if you're using Anthropic, OpenAI, or anything else directly, you get nothing out of the box. Session replay on mobile (Flutter, Android, React Native) sounds impressive but these are notoriously hard to maintain correctly — no indication of how much real-world coverage these have. ClickHouse adds significant operational complexity for self-hosters who just wanted 'one binary' — the embedded SQLite mode trades that simplicity for no horizontal scale path.