Workflows
Build and Run
Design a workflow on the canvas, wire expressions, publish, and inspect a run.
Open Products → Workflows and create a workflow, or start from a template such as Agent Chat.
Canvas
The editor is a visual graph:
- Add a trigger (every workflow has one).
- Drop work nodes from the palette and connect them.
- End with an Output node.
- Save draft, then Publish when the graph is ready.
Connect nodes by their handles. The inspector on a selected node edits its display name and configuration.
Node types (console palette)
Use these public node types when you document or design a flow:
| Category | Types | Role |
|---|---|---|
| Triggers | Manual, API, Schedule, Webhook | How a run starts |
| Data | Set variable, Transform | Shape data between steps |
| Logic | Condition | Branch on an expression |
| Integration | HTTP request, social / chat-platform media | Call external APIs |
| AI | Chat completion, Agent invocation, WhatsApp agent | LLM via Gateway, or a Yaha Agent |
| Flow | Approval, Delay, Output | Human gate, wait, final result |
Expressions
Most fields accept {{ … }} expressions so you can pass data without code.
Common patterns:
{{ workflow.input.text }}
{{ workflow.input.media_id }}
{{ nodes.summarize.output }}
{{ variables.agent_id }}Use the inspector examples as the source of truth for the keys your graph actually exposes.
Publish and run
- Save draft while you edit.
- Publish to create a version. Schedule triggers become active on publish.
- Click Run in the editor and supply JSON input for a manual start, or wait for API, schedule, or webhook traffic.
Open the run drawer or Products → Workflows → Runs to see per-node status, logs, tokens, and cost. You can cancel a running workflow from the run detail.