Pipelines
Build custom AI workflows using drag-and-drop nodes.
Last updated
Build custom AI workflows using drag-and-drop nodes.
Last updated
Pipelines are chains of logic that can be used to automate workflows. They can contain inputs, outputs, and multiple processing steps in between, all of which can be connected using a drag-and-drop interface.
The Pipelines tab allows you to access the Pipelines you have created or imported/customized from the Marketplace. Additionally, you can create new pipelines.
The "+ New" >> "Create pipeline" button gives you access to the Pipeline builder
The "Run" button opens up a "form-like" interface for you to give inputs and receive outputs from the Pipeline
Share a pipeline by clicking on the three dots and then "Share Pipeline". You will be prompted for the username of the user that you want to share with.
The Pipeline builder can be accessed by clicking the "+ New" button on the Pipelines tab.
Every Pipeline is composed of nodes and edges. Nodes represent the individual units of functionality that a Pipeline executes, such as retrieving data from various Data Loaders, processing inputs using a large language model (LLM), or formatting Text. Edges are the connections between nodes, allowing you to connect the output of one node to the input of a different node.
Every node input/output has a given type, and edges can only be added between compatible types. You will not be able to create edges that result in incompatible types unless you disable the Enforce Type Safety option under the "edit information" bar on the right. Sometimes, creating an edge will give you a Warning, which indicates that the types at either end of the edge may be compatible sometimes but not in all cases. You can select Hide Warnings if you want to remove these messages.
You can run a Pipeline directly through the Pipeline builder by clicking the Run button in the top-right, filling out the relevant inputs and outputs, and clicking Run. You will be able to see the intermediate outputs from the nodes as the Pipeline runs.
You can view and edit the name, description, and shared users associated with the Pipeline on the Edit Information tab. You can also save the Pipeline from here.
