Evidence Collection for Software Delivery

Evidence collection is easier when it happens during the work. Farna Control records the governed run instead of asking a team to reconstruct the delivery story from separate systems after the fact.

What a run records

A run keeps its trigger, pipeline snapshot, agent dispatches, node inputs and outputs, schema results, evaluation outcomes, human decisions, and final artifact. State-changing actions also append to the hash-chained audit log. The snapshot preserves the version of the pipeline, prompt, model backend, schema, and connector bindings that produced the result.

That gives a reviewer a single path from request to outcome. It does not turn a connector's self-reported result into an independent observation: the connected system remains the authority for what it reports, and Farna Control retains that provenance in the run.

Inspect and export the record

The run detail provides the evidence for one delivery. The audit log can verify its chain on demand, and the audit record can be exported for review. Team-tier audit viewing and export add organization-level access to the same evidence model; the Community tier still records run-level evidence.

Retention, backup, and access to the database remain your responsibility because Farna Control is self-hosted. Evidence is stored in your Postgres database and stays within the deployment boundary you operate.

Evidence depends on routing

Farna Control records work that enters a Farna Control pipeline. Work done entirely outside the pipeline keeps the record provided by its own tool. To collect a complete delivery trail, route the relevant steps through the pipeline and bind the connectors that can observe the external work.

Where to go next

• Audit Trail for AI-Generated Code: audit events and chain of custody
• Self-Hosted Agent Governance: data placement and custody
• Cost Visibility: run cost as part of the record