中文版

🚀 General-Evolve General Code Evolve Agent Automatically, efficiently, and stably perform optimization tasks such as algorithms, mathematical puzzles, and prompts.

🔥 ML-Evolve Machine Learning Agent Self-evolving ML Agent that autonomously understands data, builds models, and delivers an optimized solution.

⭐ LoongFlow Evolve Agent Framework A modular, highly extensible Agent framework for flexible customization and seamless integration.

LoongFlow: Inspired by Wang Yangming's "Enlightenment at Longchang," this concept signifies the deep integration of the model's "knowing" and the tools' "doing" — knowledge propels action, and action yields insight, ushering in the era of Agent cognitive autonomy. It transcends the role of a mere mechanical executor; through iterative refinement in the PES (Plan-Execute-Summarize) cycle, it shatters cognitive boundaries, achieving an evolutionary leap from a "passive tool" to an "autonomous intelligence."

A high-performance, stable, and scalable framework for evolutionary Agent development, featuring an innovative PES evolutionary paradigm to empower developers in building high-quality evolutionary Agents efficiently.

• High Efficiency: The innovative PES evolutionary paradigm, combined with multi-structural fused evolutionary memory, shifts from "random mutation" to "directed cognitive evolution." It significantly mitigates issues in traditional evolutionary methods, such as low generation quality, excessive ineffective evaluations, repetitive trial-and-error, and high randomness, thereby substantially enhancing evolutionary efficiency and convergence certainty. Compared to conventional methods, overall evolutionary efficiency is improved by approximately 60%.

• Stability: Upholding the principle of "engineering certainty," the system systematically encapsulates the inherent uncertainties of models through its architectural design, thereby reducing the burden of model reasoning and establishing a highly stable, reproducible intelligent evolution system. In practical evaluations, LoongFlow has demonstrated significant performance advantages.

• Ease of Use: LoongFlow provides comprehensive support, ranging from task-specific evolutionary Agents and a highly scalable evolutionary Agent development framework to modular atomic components. From applications to frameworks, it empowers developers to rapidly deploy evolutionary Agents for solving domain-specific problems, significantly reducing development and fine-tuning costs.

Validated on open mathematical problems proposed by Terence Tao and the AlphaEvolve team, the system outperformed all previously known best results on 11 of the problems.

Validated on 20 Kaggle machine learning competitions from the OpenAI MLE-Bench benchmark, the system achieved gold medals in 14 contests. Complete results will be announced after all competitions are concluded.

Additionally, validation was conducted on problems such as mathematical puzzles and MOE load balancing algorithms，Detailed examples can be found in Examples.

LoongFlow requires Python 3.12 or higher.

# Install uv/conda and clone repository
uv: https://docs.astral.sh/uv/getting-started/installation/
Miniforge: https://conda-forge.org/download/

# Install with uv
cd LoongFlow
uv venv .venv --python 3.12
source .venv/bin/activate
uv pip install -e .

# Install with conda
cd LoongFlow
conda create -n loongflow python=3.12
conda activate loongflow
pip install -e .

# Config LLM: Edit task_config.yaml, recommend to use gemini-3-pro-preview or deepseek-r1-250528
# Example: ./agents/general_evolve/examples/packing_circle_in_unit_square/task_config.yaml
# The model needs to configure providers as needed, default provider is openai. for example: openai/gemini-3-pro-preview
llm_config:
  url: "https://xxxxxx/v1"
  api_key: "******"
  model: "openai/gemini-3-pro-preview"

# Run your first evolve task, the evolution results are in the ./output directory
uv pip install -r ./agents/general_evolve/examples/packing_circle_in_unit_square/requirements.txt
./run_task.sh packing_circle_in_unit_square --background

# Check task log
tail -f ./agents/general_evolve/examples/packing_circle_in_unit_square/run.log

# Stop task
./run_task.sh stop packing_circle_in_unit_square

# Config LLM: Edit task_config.yaml, recommend to use gemini-3-pro-preview or deepseek-r1-250528
# Example: ./agents/ml_evolve/examples/ml_example/task_config.yaml
# The model needs to configure providers as needed, default provider is openai. for example: openai/gemini-3-pro-preview
llm_config:
  url: "https://xxxxxx/v1"
  api_key: "******"
  model: "openai/gemini-3-pro-preview"

# Init ml evolve
./run_ml.sh init

# Run your first evolve task, the evolution results are in the ./output directory
# ./run_ml.sh run <task_name> [--background] [other Python args]
./run_ml.sh run ml_example --background

# Check task log
tail -f ./agents/ml_evolve/examples/ml_example/agent.log

# Stop task
./run_ml.sh stop ml_example

from evolux.evolve import EvolveAgent

# Config evolve agent
agent = EvolveAgent(
    config=config,
    checkpoint_path=checkpoint_path,
)

# Register worker（Implement the Planner, Executor, and Summary interfaces）
agent.register_planner_worker("planner", PlanAgent)
agent.register_executor_worker("executor", ExecuteAgent)
agent.register_summary_worker("summary", SummaryAgent)

# Run agent
result = await agent()

For more details, please refer to EvolveAgent

from evolux.react import AgentContext, ReActAgent
from agentsdk.tools import TodoReadTool, TodoWriteTool, Toolkit

# Build agent context
toolkit = Toolkit()
toolkit.register_tool(TodoReadTool())
toolkit.register_tool(TodoWriteTool())

# Build default react agent
agent = ReActAgent.create_default(model=model, sys_prompt=sys_prompt, toolkit=toolkit)

# Run agent
result = await agent(message)

For more details, please refer to ReActAgent

Please read CONTRIBUTING.md for details on our code of conduct, and the process for submitting pull requests to us.

LoongFlow is licensed under the Apache License 2.