Roadmap
| Milestone | Goal | Status |
|---|---|---|
| M0 — Upstream Foundation | CI, contracts, immutable deploy path | ✅ Done |
| M1 — MVP Pilot | 7 skills verified, GenAI eval, qualification verdicts | ✅ Done |
| M2 — Production Hardening | MLflow AI Gateway integration, SSO, scorer profiles | In Progress |
| M3 — Platform Scale | Multi-tenant, cost tracking, alerting | Planned |
| M4 — Enterprise Governance | Compliance export, regulatory mapping | Planned |
| M5 — Ecosystem | Multi-cluster federation, marketplace | Future |
M1 Deliverables (Completed)
- 7 skills working end-to-end with Hermes + MLflow MCP
- 41 integration tests against real MLflow data
- OpenShell sandbox deployment on OpenShift
- Correct MLflow MCP 3.14 tool names across all configs
- Seed data tooling for local development
- Zero-code agent instrumentation (
usercustomize.py)
M2 Roadmap
- MLflow AI Gateway integration (automatic evaluation, cost tracking)
- CI/CD quality gate using
mlflow.genai.evaluate()in pipeline scripts - Agent registry backed by MLflow LoggedModel (when MCP tools available)
- SSO integration (OpenShift OAuth)
- Scorer profile YAML configuration