DataRobot Application Framework

af-component-agent

Homepage · Agent Documentation · Support

The agent template provides a set of utilities for constructing a single or multi-agent workflow using frameworks such as Nvidia NAT, CrewAI, LangGraph, LlamaIndex, and others. The template is designed to be flexible and extensible, allowing you to create a wide range of agent-based applications.

The Agent Framework is component from the DataRobot App Framework Studio

Getting Started

To use this template, it expects the base component https://github.com/datarobot/af-component-base has already been installed. To do that first, run:

uvx copier copy https://github.com/datarobot/af-component-base .
# uvx copier copy git@github.com:datarobot/af-component-base.git .

To add the agent component to your project, you can use the uvx copier command to copy the template from this repository:

uvx copier copy https://github.com/datarobot/af-component-agent .
# uvx copier copy git@github.com:datarobot/af-component-agent.git .

To update an existing agent template, you can use the uvx copier update command. This will update the template files

uvx copier update -a .datarobot/answers/agent-{ agent_app }.yml -A

Developer Guide

Please see the Development Documentation.

Full deployment E2E (Playground Codespaces)

This repo includes an optional end-to-end test that exercises a full lifecycle: render → build → deploy → test → destroy.

• Prereqs: set DATAROBOT_ENDPOINT and DATAROBOT_API_TOKEN either:
  • as environment variables in your Codespace, or
  • in a local af-component-agent/.env file.
• Run:

task test-e2e

Notes:

• By default the E2E test runs for all agent frameworks (base/crewai/langgraph/llamaindex/nat).
• To run a subset, set E2E_AGENT_FRAMEWORKS, e.g. E2E_AGENT_FRAMEWORKS=base,nat.
• The test uses a local Pulumi backend and a unique stack name per run, then cleans up.

Get help

If you encounter issues or have questions, try the following:

• Check the documentation for your chosen framework.
• Contact DataRobot for support.
• Open an issue on the GitHub repository.