Workflows
Workflows are the logical backbone of an agent, outlining how it functions. Each workflow is composed of four key elements:
-
A trigger that starts the workflow.
-
An AI model to process the workflow’s logic.
-
Plugin functions that the AI model can activate.
-
A set of instructions to guide the agent’s actions.
This structure makes workflows both powerful and adaptable. Instead of relying on a single, all-encompassing set of instructions, you can create specialized workflows tailored to specific scenarios.
Workflows can be optimized with the most suitable AI models and plugins for their tasks and triggered in various ways.
Benefits of Workflows
-
Clear and maintainable logic for agent operations.
-
Task-specific optimization using specialized models.
-
Reusable workflows for different scenarios.
-
Simplified testing and debugging of individual agent behaviors.
-
Complex agent behaviors built from simple, modular components.
Configuring Workflows
Workflow Triggers
Triggers define when a workflow starts. They can include:
-
Automatic events such as scheduled intervals (e.g., running every 10 minutes).
-
Plugin-based triggers such as receiving mentions on X/Twitter.
Custom workflows without automatic triggers can also exist and be directly activated by other workflows.
Plugin Access Control
By default, workflows can access all plugins. However, restricting plugin access can improve both performance and security. For instance:
-
A workflow triggered by Twitter mentions may not need access to wallet functions.
-
Limiting plugins to those required for the task ensures optimized context and greater accuracy.
AI Model Selection
Each workflow can utilize a different AI model based on the task. Choose from your installed models on the main builder page to optimize performance for tasks such as:
-
Summarizing content.
-
Analyzing images.
-
Executing plugin functions.
This modular design allows each workflow to focus on its strengths, ensuring efficient collaboration between components.
Last updated