#119
B+
Rank #83
Langfuse
An open-source LLM observability and evaluation platform for tracing prompts, costs, and quality metrics across LLM applications. Standout trait is its dual-database design combining Postgres for transactional data and ClickHouse for trace-scale analytics.
72.5
/ 100
⚙ Full Stack
Next.js/TypeScript backend with Postgres for auth/project metadata, ClickHouse for high-volume trace/observation storage, and Redis for queuing — a genuinely modern analytics-at-scale stack. Official docker-compose bundles all components together.
📈 Scaling Analysis
Genuinely designed for scale — ClickHouse handles high-cardinality trace ingestion far better than a single Postgres instance would, and the app tier is largely stateless behind that. The tradeoff is real operational complexity: four services to keep healthy versus a simpler single-DB app.
🚀 Running on Nexlayer
Four-pod deploy — app, Postgres (`postgres.pod:5432`), ClickHouse (`clickhouse.pod:8123`), and Redis (`redis.pod:6379`) — all wired via `.pod` DNS, with mirror.gcr.io images for the Postgres/Redis/ClickHouse bases. `NEXTAUTH_URL` must point at the Nexlayer `<% URL %>`, not a hardcoded host.