CAVRA Unified Enterprise Enhancement Roadmap
Last updated: 2026-07-03
This page converts the merged expert review into an implementation sequence with dependencies, status, verification expectations, and GitHub evidence. The source roadmap is maintained in the public repository at docs/product/cavra-unified-enterprise-product-enhancement-roadmap.md.
CAVRA is being planned as a unified AI governance control plane for two governed asset classes:
- Agent actions: file writes, shell commands, Git operations, MCP tool calls, CI/CD triggers, cloud operations, infrastructure changes, and production workflow actions.
- Models and artifacts: model registry entries, model metadata, deployment packages, AI supply-chain artifacts, assessment findings, drift signals, and compliance evidence.
The common control planes are Decision, Identity and Trust, Evidence, and Posture. This avoids building two unrelated products: one for agents and another for model risk.
| Phase | Focus | Primary Dependencies | Current Status | Exit Condition |
|---|---|---|---|---|
| 0 | Positioning and public roadmap | Review agreement, product scope decision | Completed | README, wiki, and product site describe unified agent-action plus model/artifact governance and link to this tracker. |
| 1 | Foundation trust | Phase 0 | In Progress | Security governance, API contract, signed release, SBOM, and buyer trust documentation are publishable. |
| 2 | Identity, data, and multi-tenancy | Phase 1 API contract and trust model | In Progress | Enterprise identity, RBAC/ABAC, tenant/workspace isolation, and production data architecture are implemented and tested. |
| 3 | Evidence, audit, and compliance | Phase 1 trust model, Phase 2 tenancy model | Planned | KMS-backed evidence, immutable audit log, and dynamic compliance mapping are production-ready. |
| 4 | Zero-trust scanning and connectors | Phase 2 tenancy, Phase 3 evidence/audit | Planned | Certified connector SDK, priority connectors, and model/artifact scanner agents work without raw model/data egress. |
| 5 | Policy lifecycle and event core | Phase 2 identity/data, Phase 4 connectors | Planned | Policy authoring, test, shadow, rollback, and event-driven continuous assessment paths are working. |
| 6 | Scale and ecosystem expansion | Phases 1-5 | Planned | Benchmarks, chaos tests, broader agent adapters, LLM guardrails, supply-chain checks, and red-team automation are validated. |
| ID | Phase | Problem(s) | Requirement | Dependency | Status | Tests and verification |
|---|---|---|---|---|---|---|
| R0.1 | 0 | P8, P16, P22 | Document CAVRA as one unified control plane for agent actions and models/artifacts. | None | Completed | README, wiki, and product-site validation. |
| R0.2 | 0 | P1-P22 | Publish this numbered product enhancement roadmap with dependencies and status. | R0.1 | Completed |
git diff --check; wiki Markdown render. |
| R0.3 | 0 | P1-P22 | Add a unified architecture-roadmap diagram for the public repo and wiki. | R0.1 | Completed | SVG readability and motion-safe fallback. |
| R0.4 | 0 | P22 | Make product website point buyers to the roadmap, trust posture, and implementation sequence. | R0.1, R0.2 | Completed | Product-site Playwright validation. |
| R1.1 | 1 | P6, P12 | Harden public security governance: responsible disclosure, supported versions, vulnerability handling, and release security criteria. | R0.2 | Completed | Release-security validator and Phase 1 trust-governance tests. |
| R1.2 | 1 | P10, P12 | Establish multi-maintainer governance with CODEOWNERS, maintainer onboarding, RFC process, and release cadence. | R1.1 | In Progress | CODEOWNERS and RFC docs exist; additional maintainer onboarding remains. |
| R1.3 | 1 | P6, P12 | Produce signed releases, SBOMs, provenance, and repeatable release attestations. | R1.2 | Completed | Community release workflow now runs the release trust gate, generates checksums, SPDX SBOM, SLSA/in-toto provenance, and release-trust evidence; Go runtime and Community attestation workflows retain GitHub keyless attestation controls. |
| R1.4 | 1 | P15 | Publish OpenAPI 3.x contract and API versioning discipline. | R0.1 | Completed | OpenAPI contract, export script, validator, workflow, and tests. |
| R1.5 | 1 | P22 | Publish CISO and buyer trust documentation. | R0.4 | Completed | Public trust pack and buyer review map. |
| R2.1 | 2 | P1, P13 | Implement OIDC/SAML, SCIM, RBAC, ABAC, break-glass, model-owner roles, and security-operator roles. | R1.4 | In Progress | Enterprise identity readiness contract, API endpoints, default policy, SAML bridge contract, SCIM lifecycle contract, RBAC/ABAC role model, break-glass controls, and runtime scoped approval enforcement are implemented publicly; live IdP and private SCIM worker validation remain. |
| R2.2 | 2 | P2, P13 | Implement production multi-tenant persistence with workspaces, Postgres, tenant isolation, and migration path from JSON/SQLite. | R2.1 | Planned | Tenant isolation, migration, and concurrency tests. |
| R2.3 | 2 | P11, P2, P13 | Define HA topology: stateless workers, queues, health checks, backup/DR, RTO/RPO, and data residency. | R2.2 | Planned | HA smoke, failover, restore, and residency validation. |
| R3.1 | 3 | P4 | Add KMS/HSM-backed evidence signing, key rotation, custody policy, and independent verifier support. | R2.2 | Planned | KMS integration, signature verification, and rotation tests. |
| R3.2 | 3 | P14 | Add immutable, append-only audit log separate from evidence bundles. | R2.2, R3.1 | Planned | Tamper-evidence, append-only, and SIEM export validation. |
| R3.3 | 3 | P5, P17 | Add clause-level compliance mapping packs for NIST AI RMF, ISO/IEC 42001, OWASP LLM/GenAI, NIST SSDF, and EU AI Act. | R3.2 | Planned | Control-pack schema and finding-to-clause tests. |
| R3.4 | 3 | P21, P5 | Build auditor, BI, executive, and board-ready reporting exports. | R3.3 | Planned | PDF/CSV/JSON export validation. |
| R4.1 | 4 | P3, P15 | Create public connector/plugin SDK with stable interfaces, certification rules, examples, and compatibility tests. | R1.4 | Planned | SDK, example connector, and compatibility tests. |
| R4.2 | 4 | P3 | Deliver priority certified connectors across SCM, CI/CD, SIEM, ITSM, and communications. | R4.1, R2.1 | Planned | Connector contract tests and live sandbox validation. |
| R4.3 | 4 | P3, P16, P20 | Add model registry connectors that work by reference: MLflow, SageMaker, Hugging Face, and Weights & Biases. | R4.1, R3.2 | Planned | No-raw-model-egress and metadata hash validation. |
| R4.4 | 4 | P16, P20 | Build zero-trust scanner agent that runs in customer VPC/on-prem and emits metadata, hashes, risk scores, and evidence only. | R4.3, R3.1 | Planned | Egress tests, scanner evidence tests, reference deployment validation. |
| R5.1 | 5 | P9 | Add OPA/Rego policy path alongside current policy engine with testable, Git-versioned policies. | R2.1, R4.1 | Planned | Policy unit tests and Rego parity tests. |
| R5.2 | 5 | P9 | Build policy lifecycle tooling: authoring UI, linting, versioning, shadow mode, dry run, rollback, and approval workflow builder. | R5.1 | Planned | Policy lifecycle integration tests and visual validation. |
| R5.3 | 5 | P18, P11 | Add event-driven continuous monitoring with event bus triggers for agent actions, model registration, drift, and promotions. | R2.3, R4.4 | Planned | Event replay, dedupe, latency, and stale-assessment tests. |
| R6.1 | 6 | P7, P11 | Publish latency, throughput, HA, and failure-mode benchmarks with SLO regression gates. | R2.3, R5.3 | Planned | Benchmark suite and CI regression thresholds. |
| R6.2 | 6 | P8 | Expand beyond coding agents through generic adapter SDK and action taxonomy. | R4.1, R5.1 | Planned | Adapter contract tests and sample non-coding agent scenario. |
| R6.3 | 6 | P19, P20, P21 | Add native LLM guardrail testing, AI supply-chain scanning, malicious model checks, and red-team automation. | R4.4, R5.3 | Planned | Prompt-injection tests, serialization scan tests, red-team report validation. |
| R6.4 | 6 | P16, P22, P3 | Publish zero-trust quickstart demo and reference deployments for Docker Compose, Helm, Terraform, Azure, and scanner operation. | R4.4, R6.1 | Planned | End-to-end reproducible demo and deployment smoke tests. |
Each requirement should move through the same evidence path:
- Requirement accepted in this tracker.
- Design or implementation committed.
- Tests, validators, screenshots, or deployment checks completed.
- GitHub evidence column updated in the repository source roadmap.
- Wiki mirror updated so readers can see current status.
The next clean engineering sequence is late Phase 1 into Phase 2:
- Complete the remaining R1.2 governance operating action by adding real additional maintainers/reviewers.
- Start R2.1 enterprise identity/RBAC/ABAC design and tests.
- Continue into R2.2 tenant isolation after the identity contract is stable.
- Start design records for R3.1 KMS/HSM signing, R4.1 plugin SDK, and R4.4 zero-trust scanner agent.