Skip to main content

Introduction to Rivet

Welcome to the Rivet User Guide! Rivet is a powerful Integrated Development Environment (IDE) and library designed for creating AI agents using a visual, graph-based interface. This guide will provide you with an overview of Rivet's capabilities and walk you through its various features and functionalities.

Key Components​

Rivet consists of two main components:

Rivet Application​

The Rivet Application is an editor/IDE for creating complex prompt chains and AI agents. It allows you to build Rivet project files that can be executed within your application. The Rivet Application comes with a suite of tools for designing and enhancing AI agents, such as a prompt designer, variations on nodes for A/B testing, and integrated testing to ensure your graphs work as expected for all inputs.

See this User Guide and the tutorial for more information on how to use the Rivet Application.

Rivet Core/Rivet Node​

These TypeScript libraries allow you to execute projects generated by the Rivet Application. They provide a simple API for integrating Rivet with your application. Once you've created a graph in the Rivet App, you can execute it within your application like a function call. This makes it easy to integrate Rivet's AI capabilities into your existing projects.

See the API Reference for more information on the APIs available and see integration getting started for more information on how to integrate Rivet into your application.

Node-Based Editor​

Rivet's node-based editor enables you to create, configure, and debug complex AI prompt chains and AI agent chains visually. This approach makes it easier to understand the flow of data and the state of your AI agent at any point in time. The editor allows you to view the input and output of every node, as well as AI responses in real-time, making it simple to identify and fix issues. Check out the overview of the interface and adding & connecting nodes for more information.

Library of Nodes​

Rivet features a library of node types to execute common functionality for nodes. Some essential node types include Text, Chat, Match, Loop Controller, Extract YAML, Extract JSON, Chunk, Trim Chat Messages, and External Call. These nodes can be connected together using wires, allowing data to flow between them.

Documentation for all nodes can be found in the Node Reference.

Live Debugging​

Rivet offers live debugging of AI chains as they run, allowing you to monitor the state of your AI agent in real-time and quickly identify any issues that may arise.

Remote Debugging​

Rivet also supports remote debugging, allowing you to debug AI chains running on a remote server. This is useful for debugging AI agents that are running in a production environment. See the remote debugging section for more information.

Get Started​

Now that you have an overview of Rivet and its capabilities, it's time to dive into the documentation and explore its features in more detail. The following sections will guide you through the process of installing Rivet, creating your first AI agent, and using the various tools and nodes available to build powerful AI-driven applications.