# Runtime (/docs/runtime)



Runtime is the foundational layer of BullSequana AI. It brings together the essential services needed to run data and AI workloads in production, with built-in support for networking, security, storage, inference, observability, and delivery automation.

It is the most basic data and AI level of the platform: the secure, reliable base that higher-level services build on to move faster and scale with confidence.

Who This Section Is For [#who-this-section-is-for]

This section is mainly useful for:

* platform engineers
* infrastructure and Kubernetes operators
* security and networking teams
* SRE and observability-oriented roles

Helpful background includes Kubernetes operations, networking, access control, storage, and production platform support.

What Runtime Is [#what-runtime-is]

Runtime is the shared platform layer responsible for:

* networking and service exposure
* identity, authentication, and authorization
* secret management and security controls
* model inference execution and supporting operators
* workflow orchestration and event handling
* operational data stores and artifact storage
* monitoring, logging, tracing, and debugging
* CI/CD and release automation

In practical terms, Runtime is the set of components that keeps the platform running.

How Runtime Fits in the Platform [#how-runtime-fits-in-the-platform]

Runtime provides the core operational capabilities used by the rest of the platform:

* **CoreAI** builds AI-facing services and developer capabilities on top of Runtime.
* **ProAI** adds enterprise data, analytics, and large-scale data operations.
* **Use cases** combine these layers into applications that deliver business value.

Runtime Capability Areas [#runtime-capability-areas]

The Runtime layer is organized around several operational capability domains.

| Capability area                     | Purpose                                              | Example components                                 |
| ----------------------------------- | ---------------------------------------------------- | -------------------------------------------------- |
| Network                             | Expose, route, and protect platform traffic          | APISIX, MetalLB, NGINX, External DNS, cert-manager |
| Access and security                 | Manage identities, permissions, and secrets          | Keycloak, OpenFGA, OpenBao                         |
| Inference                           | Run and support model-serving workloads              | KubeAI, vLLM, FasterWhisper                        |
| Workflows and automation            | Coordinate asynchronous and durable processes        | Argo Events, Temporal                              |
| Data and storage                    | Persist operational state and artifacts              | PostgreSQL, PgAdmin, MinIO                         |
| Monitoring and debugging            | Observe health, logs, metrics, and traces            | Grafana, Prometheus, Loki, Alloy, Tempo            |
| Continuous integration and delivery | Build, version, publish, and deploy platform changes | Harbor, GitLab, Argo CD                            |

Together, these components form the execution environment used by the rest of the BullSequana AI platform.

Why Runtime Matters [#why-runtime-matters]

Without Runtime, the platform would have no consistent way to:

* expose services to users and other systems
* secure access across users, services, and environments
* execute AI inference workloads in production
* orchestrate workflows and automations
* store artifacts, metadata, and operational state
* monitor reliability, performance, and incidents
* deliver updates safely and repeatedly

Runtime therefore acts as the base layer for platform operations. It is essential infrastructure, even when the user-visible value is delivered somewhere else in the stack.