Connecting Nodes
Last updated
Last updated
A Node functions as an abstract component within a workflow architecture. Nodes can serve various purposes, such as data preprocessing, managing LLM Agents, tools, and supporting retrieval-augmented generation. Nodes can be interconnected to form directed acyclic graph workflows, allowing development of applications of any complexity.
A Workflow begins with Input Node and finishes with Output Node. Both nodes will be automatically added to each new workflow.
Each node has its own set of inputs and outputs, which vary based on its functionality. An Input Node has two default parameters: input for text input, and files for file input.
Input: For text input
Files: For file input
Clicking on either the input or the output nodes will bring up their configuration options. Any parameter can be added to both nodes.
Nodes can be connected in a single workflow to create a sequence of operations. This is done by simply connecting the output of one node to the input parameter of another node.
By doing this, the necessary code to correctly pass the input data to the input transformer is generated automatically. For more information about the input and output transformers, please refer to the following link.
Alternatively, we can connect nodes directly without linking their parameters. This approach allows all parameters to be passed at once, offering flexibility to define custom logic within the input transformer, which is empty by default. This method is also used by the Choice Node, enabling conditional branching based on specific criteria. It also allows for the creation of a workflow structure, outlining the sequence of nodes without requiring data to be passed between them.
Without specifying the input transformer, this implementation will not function as expected. However, if we use the same instructions in the input transformer as in the previous example, both implementations will be equivalent.
