Actions
Last updated
Last updated
Actions are the building blocks of your workflow. Each action performs a specific task - from controlling flow and making decisions to processing data and generating AI content. By connecting actions together, you create workflows that can handle complex tasks automatically.
Every workflow starts with an Input action and ends with a Response action. Between these points, you can add any number of actions to process your data, make decisions, or generate content.
To connect new actions to your workflow, you can click or drag the ⊕ Plus button at the top-right corner of a previous action. This will present you with a dialog allowing you to choose from various types of actions
Control actions manage the flow of your workflow. They help you create branching conditions, handle repetitive tasks, and determine how data moves through your workflow. Use them to build logic like if-then decisions, loops for processing multiple items, or splits for handling different cases.
AI actions let you work with language models and other AI capabilities. LLM-based Prompt action is central here - it helps you craft prompts, choose models, and generate text. You'll also find tools for prompt management, model configuration, and AI-driven analysis.
Tool actions handle specific operations like making HTTP requests, searching vector stores, querying databases, or even executing your own custom code. They help you connect with external services, process data, and perform utility functions that support your workflow's main tasks.
Actions connect to each other through links, creating the path that data follows through your workflow. When you link actions:
Data flows from one action to the next
Each action can access results from previous actions
You can create parallel paths for simultaneous processing
You can merge paths to combine results
Some actions can work with streaming data, allowing for real-time processing in applications like live translation or transcription. Connecting new actions is as simple as using the ⊕ Plus button on one of the previous steps.
The best way to understand actions is to use them. Try starting with a template, or build a simple workflow that uses:
An Input action to receive data
A processing action (like an LLM action to generate text)
A Output action to return results
As you get comfortable, you can explore more complex patterns using Controls to add logic and Tools to expand functionality.
