Restructure solution layout by module
This commit is contained in:
22
src/AdvisoryAI/StellaOps.AdvisoryAI/AGENTS.md
Normal file
22
src/AdvisoryAI/StellaOps.AdvisoryAI/AGENTS.md
Normal file
@@ -0,0 +1,22 @@
|
||||
# Advisory AI Guild Charter (Epic 8)
|
||||
|
||||
## Mission
|
||||
Deliver the Advisory AI assistant service that synthesizes advisory/VEX evidence, policy context, and SBOM data into summaries, conflict explanations, and remediation hints—always with citations and guardrails.
|
||||
|
||||
## Scope
|
||||
- Service under `src/AdvisoryAI/StellaOps.AdvisoryAI` (retrievers, deterministics, orchestrator, guardrails, inference adapters, REST APIs).
|
||||
- Batch processing for CLI/automation, caching, observability, and integration with Console, CLI, and downstream systems.
|
||||
- Coordination across Conseiller, Excitator, VEX Lens, SBOM Service, Policy Engine, Findings Ledger, Web Gateway, Authority, DevOps, and Docs.
|
||||
|
||||
## Principles
|
||||
1. **Evidence preservation** – Raw advisory/VEX documents remain untouched; AI outputs reference them with citations.
|
||||
2. **Deterministic scaffolding** – Retrieval, mapping, and validators are deterministic; model variability is constrained.
|
||||
3. **Guardrails everywhere** – Redaction, injection defense, output validation, and audit logging are non-optional.
|
||||
4. **Policy-aware** – Outputs respect selected policy versions, thresholds, and trust configs.
|
||||
5. **Explainable** – Each result carries structured JSON, rationale, and provenance; humans can audit every statement.
|
||||
|
||||
## Definition of Done
|
||||
- Retrievers/deterministic tools implemented with tests and docs.
|
||||
- API endpoints documented (OpenAPI), RBAC enforced, guardrails active.
|
||||
- Console/CLI integrations operational; telemetry dashboards live.
|
||||
- Documentation suite published with compliance checklist.
|
||||
Reference in New Issue
Block a user