LLM observability
Tracing every model call with inputs, outputs, latency, tokens, cost and the user outcome.
Logs and traces show what the agent actually did, step by step, and what it cost. Joined with product analytics, they show whether the feature helps users or just runs. Privacy rules decide what content can be stored.
Related terms
- Evals: A repeatable test set that scores an AI feature, so changes can be measured instead of guessed.
- Tokens: The pieces of text models read and write, and the unit you pay for.
- AI gateway: A proxy between your app and model providers that adds logging, caching, limits, billing and fallbacks.
