Skip to content

Orchestrate Nodes

Node orchestration defines how data flows through a workflow. EZChat Workflow Studio uses nodes and edges to form a graph. Each node handles one step, and each edge determines execution order and data flow.

Add Nodes

You can add nodes in these ways:

  • Click Add Node from the left toolbar.
  • Open the add node menu from an empty area on the canvas.

Currently supported nodes include:

NodeUsage
StartWorkflow entry point. Provides query and files.
LLMCalls a language model to generate text or intermediate results.
Knowledge RetrievalRetrieves relevant chunks from datasets.
EndDefines the final output template.
AgentUses a model and MCP tools to complete tasks.
Question ClassifierClassifies input into configured classes.
If / ElseRoutes execution based on conditions.
HTTPCalls external APIs.
Note MemoAdds notes on the canvas and does not participate in execution.

Workflow Studio add node menu

Connect Nodes

After connecting nodes with edges, downstream nodes can use outputs from upstream nodes. The settings panel generates selectable variables based on upstream nodes, for example:

Upstream nodeAvailable variables
Startquery, files
LLM / Agenttext, usage, model
Question Classifierclass_id, class_name, keywords
Knowledge Retrievaldocuments
HTTPstatus_code, body, headers

These variables can be used in prompts, condition checks, knowledge retrieval queries, or final output templates.

Common Canvas Actions

  • Pointer mode: Select, move, and configure nodes.
  • Hand mode: Pan the canvas.
  • Organize: Automatically arrange the workflow layout.
  • Undo / Redo: Undo or redo canvas changes.
  • Zoom / Mini map: Zoom and navigate quickly.
  • Export: Export the current view or the full workflow as PNG, JPEG, or SVG.

Design Recommendations

  • Start from the Start node, confirm input fields, then add processing nodes step by step.
  • Use clear labels for every node so trace logs and maintenance are easier.
  • First run a simple Start -> LLM -> End flow, then add retrieval, branches, or APIs.
  • After adding branches, test success, failure, and fallback cases.