How Langfuse is scaling LLM observability for the agentic era with ClickHouse Cloud

Langfuse leverages ClickHouse Cloud to enhance real-time observability and analytics for large-scale, complex language model (LLM) applications, transitioning from Postgres to ClickHouse for improved performance and scalability. This transition significantly reduced query latency and supported increased data ingestion volumes, leading to a shift towards immutable, wide-table modeling that optimizes memory usage and query speed. As LLM applications evolved from simple retrieval-augmented generation (RAG) workflows to complex agentic systems, Langfuse adapted its data model, ingestion patterns, and SDK design to meet the demands of these advanced systems. Observability challenges shifted focus from basic request success to understanding agent decision-making processes, prompting Langfuse to expand from trace trees to graph views and analytics tools, allowing teams to analyze latency, cost, and quality of millions of runs. Langfuse's move to ClickHouse, facilitated by an extensive pipeline redesign, addressed previous scalability issues and introduced a more adaptable, open-source data platform. The migration allowed for performance improvements while maintaining flexibility in user data ingestion and analysis. The company is now transitioning to an immutable data model, further enhancing performance and aligning with changing user expectations, while also backfilling legacy data to accommodate existing deployments. The recent acquisition of Langfuse by ClickHouse further integrates their capabilities, continuing the journey of scaling observability for agentic systems.

Use this post, company, and trend context to find content marketing opportunities, perform competitive analysis, or address product feature gaps via the Plushcap MCP server or the Plushcap API.