Decision-maker overview
A concise non-implementation path through what CALIBER is, what it solves, and where to go next for technical proof.
This page is the shortest non-implementation path through what CALIBER is, what problems it solves, and where to go next for technical proof.
At a glance
| Topic | Current position |
|---|---|
| Product shape | MLflow-integrated control plane for governed agentic workflows |
| Primary problem | Building, evaluating, calibrating, governing, and observing LLM workflows from one platform |
| Deployment topology | Standalone CALIBER ASGI service; communicates with MLflow over HTTP via MLFLOW_TRACKING_URI. Embedded mode is unsupported. |
| Governance posture | Explicit evidence, review, approval, and apply/release control paths |
| Best technical proof | Layered architecture |
| Market context | Competitive analysis |
| Forward plan | Roadmap |
1. What CALIBER is
CALIBER is not only an evaluation tool and not only a workflow canvas. It is a governed platform for the full lifecycle around LLM-powered artifacts and workflows:
- author
- test
- calibrate
- review
- apply or release
- observe
2. Why teams adopt it
The platform is most useful when a team needs more than isolated prompt testing. Its value comes from keeping the lifecycle, evidence, and control surfaces in one system rather than scattering them across separate point tools.
3. Deployment model
CALIBER runs as a standalone ASGI service that communicates with a separately running vanilla MLflow server over HTTP via MLFLOW_TRACKING_URI. CALIBER owns its own metadata database (CALIBER_DATABASE_URL), separate from MLflow's backend store.
The embedded mlflow.app mode is not a supported deployment or developer path. See the architecture for the full topology support matrix.
4. Governance posture
CALIBER is designed to keep authority explicit.
In practical terms, that means:
- evidence generation is not the same as live release
- calibration is not hidden autopilot
- approvals and apply/release paths remain visible control points
For the full trust model, use:
5. Recommended next reads
| If you need... | Read this next |
|---|---|
| the technical proof | Layered architecture |
| the architecture entry path | Architecture reader guide |
| the product/user journey | Choose your CALIBER path |
| market position | Competitive analysis |
| roadmap and execution direction | Roadmap |