Workflows
Learn how to orchestrate automated pipelines, integrate external databases, call LLMs, and handle custom business logic using our visual canvas builder.
Workflows empower you to automate complex business processes, connect disparate data sources, and incorporate intelligent AI-driven decision-making into your applications. By leveraging our visual drag-and-drop workflow canvas, you can design, wire up, and execute multi-step pipelines without wrestling with boilerplate infrastructure code.
Core Concepts
Before building your first workflow, it helps to understand the foundational architecture:
- The Canvas: An interactive, node-based workspace where you visually assemble your automation logic. You drop nodes onto the grid and connect them using directional edges to define execution flow.
- Nodes: Self-contained execution units that perform a specific task (such as querying a database, making an HTTP call, or invoking an LLM). Each node accepts inputs, runs an operation, and outputs structured data.
- Edges & Data Flow: Connectors that link nodes together. Data flows dynamically down the graph, allowing downstream steps to reference outputs, parameters, and execution results from any previously executed upstream step.
Best Practices
- Keep Workflows Modular: Instead of building one massive monolithic workflow, break complex processes into smaller, reusable sub-workflows that can be triggered independently.
- Handle Failures Gracefully: Always configure fallback behaviors or error-handling branches for network-dependent steps like HTTP requests and external database connections.
- Leverage Structured Outputs: Take advantage of JSON parsing in database and API steps so downstream nodes can reliably query specific properties rather than parsing raw text strings.