Use CALIBER
Workflows
The user and operator path for authoring, previewing, running, deploying, and debugging workflows in CALIBER.
Use this page for the practical workflow path: author, preview, run, deploy, inspect, and recover. Use the deep architecture page only when you need the underlying implementation model.
At a glance
| Task | Start here | Deep reference |
|---|---|---|
| design a workflow | start from the workflow mental model | Workflows architecture |
| look up available nodes | inspect the component catalog | Workflow components |
| run or deploy a workflow | use preview, run, and deployment surfaces | Workflows architecture |
| debug a failing step | inspect runtime evidence and queue state | Observability |
| recover a stuck or unsafe run | use the operator recovery docs | Operations runbook |
1. What workflows are for in CALIBER
Workflows are the orchestration surface for multi-step automation. They connect prompts, tools, skills, MCP access, files, and approvals into a governed run model.
2. Common tasks
| You want to... | Read this next |
|---|---|
| understand the workflow model | Workflows architecture |
| choose the right node | Workflow components |
| publish a workflow service | REST API overview |
| inspect run failures | Health and readiness, Operator troubleshooting |
3. Common failure modes
| Symptom | First thing to check |
|---|---|
| The workflow compiles but a live run fails | runtime inputs, approvals, or downstream dependencies |
| Runs are queued but not moving | event backend, worker health, or queue state |
| A deployment exists but consumers cannot call it | service path, auth, or project scoping |