Overview

Next
AgentJet

AgentJet (AJet) is a cutting-edge, user-friendly agent tunning framework designed to optimize LLM models and agent workflows.

🛩️ Key Features

Get Started with Ease

AgentJet simplifies the process of tuning the models that power your agent workflows. It supports nearly all major agent frameworks (e.g. agentscope, langchain), as well as framwork-less agents built from HTTP requests.

Rich Tutorial Library

Rich examples as beginner's tutorial: math agent, werewolves rpg, appworld ... All with step-by-step guides. Covering various agentic frameworks.

Reliable and Reproducible

Checkout AgentJet's community-powered, robot-assisted open-benchmarking system. Share progress, compare training backbones, discover bugs and iterate faster than ever! Click here to see AgentJet performance across tasks/versions/backbones.

Multi-agent and Multi-turn

Built to support advanced multi-agent and multi-turn LLM workflows, AgentJet intergrates timeline-merging algorithms that automatically analyze and consolidate each agent's LLM timeline, accelerating training speed 1.5x ~ 10x.

High Resolution Logging

Log token-level rollout details, capturing token IDs, token loss masks, and token log probabilities with web UI display. This Support workflow development, agent diagnostics, and facilitate research on advanced LLM algorithm studies.

Any Training Engine

Support multiple training engines as backbone (VeRL and Trinity-RFT). Tinker backbone support will be released soon. Choose from vLLM and SGLang as you wish. Say goodbye to training engine gaps.

🛩️ Quick Start

Click Here for the Full Installation Document

We recommend using uv for dependency management. Click here for details and other training backbone (e.g. Trinity-RFT) options.

  • Clone the Repository: 

    git clone https://github.com/modelscope/AgentJet.git
cd AgentJet

  • Set up Environment: 

    uv venv --python=3.10.16 && source .venv/bin/activate
uv pip install -e .[verl]

# Note: flash-attn must be installed after other dependencies
uv pip install flash_attn==2.8.3 --no-build-isolation --no-cache-dir

  • Train the First Agent: 

    # You can start training your first agent with a single command using a pre-configured YAML file

ajet --conf tutorial/example_math_agent/math_agent.yaml

🛩️ Example Library

Explore our rich library of examples to kickstart your journey:

Math Agent

Training a math agent that can write Python code to solve mathematical problems.

AppWorld Agent

Creating an AppWorld agent using AgentScope and training it for real-world tasks.

Werewolves Game

Developing Werewolves RPG agents and training them for strategic gameplay.

Learning to Ask

Learning to ask questions like a doctor for medical consultation scenarios.

Countdown Game

Writing a countdown game using AgentScope and solving it with RL.

Frozen Lake

Solving a frozen lake walking puzzle using AgentJet's reinforcement learning.

🛩️ Core Concepts

AgentJet makes agent fine-tuning straightforward by separating the developer interface from the internal execution logic.

🛩️ The User-Centric Interface

To optimize an agent, you provide three core inputs:

Trainable Workflow

Define your agent logic by inheriting the Workflow class, supporting both simple and multi-agent setups.

Task Reader

Load training tasks from JSONL files, HuggingFace datasets, or auto-generate from documents.

Task Judger

Evaluates agent outputs and assigns rewards to guide the training process.
AgentJet Architecture

🛩️ Internal System Architecture

The internal system orchestrates several specialized modules to handle the complexities of RL training and agent interactions.

Module Description
Launcher Manages background service processes (Ray, vLLM) and routes the backbone
Task Rollout Bridges LLM engines and manages the Gym environment lifecycle
Task Runner Executes the AgentScope workflow and calculates rewards
Model Tuner Forwards inference requests from the workflow to the LLM engine
Context Tracker Monitors LLM calls and automatically merges shared-history timelines (1.5x-10x efficiency boost)

🛩️ Next Steps

Installation

Set up AgentJet environment and dependencies.

Quick Start

Run your first training in minutes.

First Agent

Build and train your own agent from scratch.

Examples

Explore detailed training examples.
Introduction