llm-observability

What problem does it solve?

LLM observability and LLMOps capability pack provides the rules, tooling, and patterns needed to monitor, measure, and govern AI agents in production, including tracing, cost attribution, drift detection, and prompt/versioning governance.

Core Features & Use Cases

• OpenTelemetry GenAI semantic conventions adoption and telemetry conformance checks.
• End-to-end production observability: tracing, latency profiling (TTFT/ITL), and per-call cost attribution with four-layer token accounting.
• Drift detection, grounding checks, and online evaluation for high-stakes LLM deployments.
• Centralized prompt registry and versioning discipline to enforce stable operational templates.

Quick Start

Apply this pack to your LLM-powered system to enable production-grade observability and governance.