One Database for LLM Observability: Traces, Metrics, and Conversations Build a unified LLM observability stack using OpenTelemetry GenAI Semantic Conventions and GreptimeDB. Store traces, metrics …

ITNEXT is a platform for IT professionals, developers, and technology enthusiasts, offering articles, tutorials, and insights on a wide range of topics, including software development, cloud computing, and machine learning. With a focus on practical solutions and real-world applications, ITNEXT provides actionable advice and best practices for navigating the complexities of modern technology landscapes. Developers can learn about  technologies, explore hands-on tutorials, and stay up-to-date on the latest industry trends through ITNEXT's engaging and informative content.

ITNEXT

LLM observability typically requires three separate systems: traces in Jaeger, metrics in Prometheus, and conversation logs in Elasticsearch. This post demonstrates a unified approach using OpenTelemetry GenAI Semantic Conventions and GreptimeDB as a single backend for all three signal types. The setup uses one line of Python instrumentation to emit traces, metrics, and conversation logs via OTLP directly into GreptimeDB. Key capabilities include SQL joins across traces and conversation logs using shared trace_id/span_id, GreptimeDB Flow for deriving per-model token and latency rollups from spans without double-writing metrics, full-text search on conversation content linked back to traces, and PromQL support for existing histogram metrics. Cost estimation, error rate by model, and p95/p99 latency queries are shown with concrete SQL examples. A Docker Compose demo is available supporting both OpenAI and Ollama.