When you need an audit trail beyond traces
Advanced governance and audit workflows for production AI ingest — optional beyond Agent Lab.
Most AITracer users only need Traces and Training data: prompt/response pairs from Agent Lab, chat, or coach.
Some teams also ingest production model calls and need policy results, integrity checks, and long-term audit exports. That is what the advanced audit and governance surfaces are for.
What belongs in a production audit record
When you ingest via SDK/API, a useful record typically includes:
- request and response text (or hashes, depending on retention policy),
- model and action name,
- token counts, latency, and estimated cost,
- optional policy / risk flags,
- optional integrity hash for later verification.
Agent Lab vs audit workflows
| Workflow | Typical user | What you use |
|---|---|---|
| Build a local agent | Developer | Run agent, coach, training export |
| Monitor production AI | Platform team | SDK ingest, traces, cost views |
| Compliance review | Security / legal | Governance, verification, audit vault |
If you are only training on Ollama, you can ignore audit vault entirely and still get full value from the product.
See Governance Engine and Audit Vault when you are ready for production controls.