AMS Deployment Guides¶
Welcome. These guides explain how to deploy the Agent Management System (AMS) on Azure Kubernetes Service for a customer release environment.
You choose which application modules and adapters you need. We describe what infrastructure and shared services you must have in place. We do not ship Infrastructure-as-Code in this repository — you may provision with the Azure portal, Terraform, Bicep, or your own automation.
Use the language switcher in the header to move between English (default) and 日本語.
What you will deploy¶
| Layer | What it is | Required? |
|---|---|---|
| Azure infrastructure | AKS, Application Gateway (or equivalent ingress), DNS, TLS, registry access, databases/storage as needed | Yes — bring your own |
| Observability | Metrics, logs, traces (Grafana stack and OpenTelemetry path) | Yes — shared dependency |
| Keycloak | SSO / identity for AMS UIs | Yes — shared dependency |
| Infrastructure ingress | One shared AMS hostname; / landing owned by one module |
Yes — shared dependency |
| Application modules | Command Center, ARE, EOBS, FortifyOps | Choose one or all four |
| Adapters | Dify, EMA, LibreChat, Lyzer, MS Copilot workers | Optional per use case |
| Platforms | Self-hosted Dify, LibreChat, Langfuse | Optional; see platform notes |
Recommended reading order¶
- Getting started — pick modules and understand the landing rule
- Architecture — how pieces fit together
- Infrastructure — resource checklist
- Observability and Keycloak
- Infrastructure ingress — set who owns
/ - Deploy your application modules
- Add adapters and optional platforms if needed
- Follow the go-live checklist
How you run Helm¶
Access to the cluster API matters for every install step. Read:
- Private AKS when the API server is private
- Public endpoint or jump VM when you reach the cluster from a public API or a VM inside the VNet
Then apply the same Helm patterns to each chart you install.