Workflow Introduction

A workflow is a series of interrelated steps used to implement business logic or complete specific tasks. It provides a structured framework for data flow and task processing of applications or agents.

The platform provides a visualized canvas, which allows you to quickly set up a workflow by dragging nodes. In addition, you can debug workflows in real time on the canvas. In the workflow canvas, you can clearly view the data transfer process and task execution sequence.

When creating a workflow, you need to set different parameters for each node, such as input and output parameters. You can orchestrate more nodes through drag-and-drop to implement complex service process orchestration and quickly build applications.

Workflow is mainly used in complex service scenarios where the target task contains multiple complex steps and the success rate and accuracy of the output result have strict requirements.

When orchestrating a workflow, you can use the following nodes to design functions. For details, see Workflow Node Configuration Reference.

• Start node: The workflow starts from the start node. The information entered by users is transferred from the start node.
• End node: The end node is the final node of a workflow and defines the output information of the workflow.
• Large model node: introduces large model capabilities to workflows.
• IntentDetection node: classifies intents based on user input and directs the intents to different subsequent processing flows.
• Questioner node: provides the capability of collecting more information from users during a dialog.
• Plugin node: introduces API plug-ins, executes user-defined plug-ins based on the input of the node, and uses the plug-in execution result as the output of the node.
• Branch node: functions as a branch switch node during application orchestration. You can specify the workflow branch to be executed based on the input judgment conditions.
• Code node: introduces a code executor to execute specified Python code based on the input of the node. The output of the node is the code execution result.
• Knowledge Repo node: retrieves matched information from the specified knowledge base based on input parameters.
• Message node: defines a piece of text content and sends the content to users during workflow execution.
• Loop node: repeats a series of tasks.
• Variable Assignment node: dynamically sets intermediate variables during loop execution.
• Aggregation node: integrates output variables of multiple branches into one, facilitating unified configuration on downstream nodes.
• Input node: collects user input during workflow running.
• Workflow node: implements workflow nesting.
• MCP Service node: is one of the core components for invoking third-party capabilities in workflows. As an important carrier for function extension, this node can invoke the MCP service to execute specific function tasks. Each MCP service is essentially a tool set. It can provide modular services to expand the capability boundary of workflows and complete more complex tasks.
• Agent node: provides the capability of using large models and calling large model tools. You can configure deployed models on the node. You can compile prompts and bind plug-ins to enable the model to process tasks.