Declarative
Apply YAML. Get an HTTP endpoint. Everything else (Deployment, Service, ConfigMaps, Secret mounts, rollouts) is handled for you.
Genkit Operator
Turn YAML into production-ready Genkit HTTP endpoints — declaratively, on Kubernetes.
What it does
Section titled “What it does”You describe your AI app as a handful of Kubernetes resources — Prompt,
Model, PluginConfig, Flow (or FlowSet) — and the operator builds,
runs, and exposes each flow at POST /<flow-name>. Provider credentials
are resolved from Secrets, prompts are projected as files, and a
content-hash drives rolling updates whenever anything changes.
Multi-provider
First-party plugin support for Anthropic, OpenAI, Google AI, Vertex
AI, AWS Bedrock, and Ollama. Pluggable per PluginConfig.
Built for GitOps
Server-Side Apply, owner references, content hashes — fits Argo CD and Flux out of the box.
Bring your own runner
The reference runtime is one of many. Build a custom runner image that consumes the same on-disk contract.