Category coverage
Every generated test card mapped to the four audit surfaces. LaunchAudit goes deep where launches actually break — admin/RBAC and middleware — not just the happy path.
Test results
Evidence-gated checks generated from repo truth, runtime crawl, and the risk model — grouped by category.
Context contract
Outperformance proof harness
Competitive categories tracked as proof gates, not marketing claims.
Flagship feature layer
Best-fit competitor ideas rebuilt as Fusion-native proof systems.
Model routing harness
Campaign tasks route to managed models, OpenRouter coding models, Docker-hosted local LLMs, or proprietary QA endpoints by category — each with a quality gate.
Persistence layer
Postgres tables and Blob artifact policies back projects and campaigns. Live mode is reported by the API, never assumed.
Postgres tables
Blob artifacts