Observability
Also: tracing, trace
Being able to see what an AI system is doing inside, so problems can be found and explained.
The fuller definition
Observability is the ability to understand a running system through outputs such as logs, metrics, and traces of component activity.
- Logs
Individual events
- Metrics
Counts and timing
- Traces
Linked execution steps
A simplified relationship, not a screenshot or a complete design.
A concrete example
A trace reveals that an answer failed because retrieval returned no relevant passages, rather than because generation misread them.
The distinction that matters
Monitoring that says a service is online does not establish the accuracy of its work products.
Related terms
Where it is taught
Test an AI workflow on work you can assessSource and further reading
Anthropic · Evaluating AI agentsThe definition and example are written for this library. The link gives background, official documentation or legal authority as labelled. Legal scope depends on the jurisdiction and current text; product features depend on release and settings.