Evidence ledger
When an agent moves money or changes records, “we think it’s fixed” isn’t enough. Someone will eventually ask what happened, what did you do about it, and can you prove it? The evidence ledger is Lumni’s answer: an append-only trail attached to every investigation.
What the ledger records
Section titled “What the ledger records”- The failure and its diagnosis — detector, suspicion score, the offending trace values.
- The evidence it’s built on — trace artifacts, the tool response, the policy version in force, the code/runtime change that preceded it.
- The hypothesis loop — proposed root cause, and whether a human confirmed or rejected it.
- The replay verdict — did the candidate fix resolve it?
- The resolution — patched, rolled back, dependency pinned, escalated, accepted-as-risk.
Why append-only
Section titled “Why append-only”Entries are added, never rewritten. That’s what makes the ledger usable as an audit artifact: the timeline of who knew what and when is preserved, so an incident review or a compliance request can be answered from the record instead of from memory.
Who uses it
Section titled “Who uses it”| Team | What they get |
|---|---|
| Engineering | A shared, linkable timeline of an incident and its fix |
| Finance / RevOps | Evidence of exposure caught and controls that fired |
| Compliance / Audit | A defensible trail for regulated, money-touching agents |