Roadmap
This roadmap reflects the current direction of Kavach. It is a planning artifact, not a release commitment. Items should be read as direction rather than promise.
Vision
Evaluation alone is not enough for production AI systems. A high score on one run does not explain which governed configuration produced that score, whether quality regressed over time, or whether a candidate should be recommended for promotion.
AI systems change continuously — prompts evolve, models are upgraded, datasets change, evaluation criteria shift. Those changes require a governance control plane, not only an evaluator.
Kavach is aimed at a durable governance layer for AI systems, with emphasis on AI asset governance, experiment management, evaluation history, drift detection, and deployment recommendations. The long-term value is not a single evaluator or storage engine. It is a stable governance model that survives changes in providers, runtimes, and transport frameworks.
What Kavach Owns
- AI asset governance for prompts, models, and datasets
- Experiment management and candidate tracking
- Evaluation history and comparison
- Drift detection
- Ranking and leaderboard generation
- Framework-neutral governance and leaderboard APIs
What Kavach Does Not Own
- Prompt deployment
- Model deployment
- Infrastructure management
- CI/CD execution
- Application runtime orchestration
The platform should remain usable across those systems instead of becoming one of them. Kavach recommends and governs; downstream systems deploy.