| name | sam-deploy |
| description | Use when deploying Solace Agent Mesh for shared or managed use — Helm charts on minikube (local rehearsal) or EKS/GKE/AKS (cloud), choosing chart values (broker, persistence, images, LLM, OIDC/RBAC, ingress), promoting a local/desktop instance to a cluster, or running SAM in Docker for a team. Not for single-user local trial (sam-install-run), authoring agents/config content (sam-author-agent / sam-declarative-config), or diagnosing an already-broken deployment (sam-troubleshoot / sam-operate). |
| version | dev |
sam-deploy
This skill stands up SAM for shared use — others will depend on this instance. For "just let me try it on my machine" see sam-install-run (desktop first). For what runs on the deployment (agents, workflows, config) see sam-author-agent and sam-declarative-config. Deciding question: are others depending on this instance?
Where the deployment artifacts come from
The Helm charts ship as a downloadable zip from the Solace product portal: https://products.solace.com/prods/Agent_Mesh (login required), along with image-pull credentials. Install from the unzipped local chart directory. Do not point users at GitHub repositories, and do not produce helm repo add <url> commands from memory — older README/doc snippets referencing a public Helm quickstart repo are the Python-era flow, not this product's distribution channel. If a user lacks portal access, route them to their Solace account team or solace.com/support.
Paths, in teaching order
- Helm — the primary path, local and cloud alike. Same chart both ways: rehearse on minikube, promote to EKS/GKE/AKS by changing values, not architecture. See references/helm-minikube.md for the local flow and references/helm-cloud.md for what changes in the cloud.
- Docker — secondary, for "Kubernetes is overkill, one VM for the team." See references/docker.md. There is no official docker-compose file. If a user asks for one, never present guessed product facts (image names, ports, paths, env vars) as real — a skeleton with those marked as placeholders to fill from the shipped deployment docs is fine; a confident fabricated file is not. The supported shapes are the all-in-one container or per-binary containers.
Hard rules
- Always set
sam.platform: "go". The chart serves both runtimes and currently defaults to the Python one. This is the single most important value; it selects the Go GWE/AWE/STR component split.
- Values keys come from the chart, not memory. This skill names the load-bearing keys (references/helm-values.md); for anything beyond them, read the
values.yaml and docs/ inside the downloaded chart bundle — never guess. Python-era env names (LLM_SERVICE_*, SOLACE_BROKER_*) are not the Go chart's surface.
- The quickstart is genuinely self-contained. Default values bundle an embedded broker and persistence (Postgres + object storage) in-cluster — do not bolt on a separate broker install for a local rehearsal.
- Pre-flight is built in: a
sam-doctor job runs as a Helm pre-install/pre-upgrade hook by default and fails the install on broken prerequisites (broker, LLM endpoint, DB, storage, OIDC, TLS). If an install fails immediately, read the doctor job's logs first.
- Production isn't quickstart: shared deployments replace the embedded broker and bundled persistence with real services, and enable OIDC — a team instance without auth leaks prompts and artifacts.
References
| Topic | File |
|---|
| Minikube local rehearsal — prerequisites, install, access | references/helm-minikube.md |
| First-deploy values checklist (broker, persistence, images, LLM, auth, networking, sizing, sam-doctor) | references/helm-values.md |
| Cloud (EKS/GKE/AKS) — what changes vs local, workload identity, promotion from desktop | references/helm-cloud.md |
| Docker on one VM — supported shapes, what to configure | references/docker.md |
Hand-offs: configuring content on the new deployment (agents, gateways) → sam-declarative-config (sam config pull/plan/apply is also the desktop→cluster promotion mechanism); SSO/RBAC depth, secrets promotion, upgrades → sam-operate; "deploy failed / pods crashlooping" symptoms → sam-troubleshoot / sam-operate.
