The decision engine is fully offline and deterministic. An optional local LLM CLI can draft an explanation from the finished decision record; it cannot alter the verdict.
Quick start
python3 agent.py evaluate --input fixtures
python3 agent.py evaluate --input my-release-evidence --output decision.json
python3 agent.py verify --input my-release-evidence --record decision.json
bash smoke_test.sh
Provide changes.json, tests.json, risks.json, slo.json, and owners.json. See CUSTOMIZE.md for their schemas and integration points.
Decision flow
local evidence --> deterministic checks --> blockers / accepted risks --> signed record
| | |
tests, owners, SLO named, dated waiver replay later
| Component | Role | Can change verdict? |
|---|---|---|
| Deterministic core | Validates tests, ownership, risk acceptance, SLO and rollback plan | Yes |
| Optional LLM CLI | Writes reader-friendly decision prose | No |
This pack reads files only. It performs no network access, deployment, payment, or approval action.
verify recalculates the complete record and rejects any changed, removed, or added output field.