NetForge RL

Multi-agent cybersecurity environment for RL research. Red team vs Blue team on generated networks.

---

Features (v2.0)

• JAX Backend: Vectorized step implementation (jax.vmap, jax.jit) providing >1,000,000 steps-per-second at 4096 parallel environments on CPU.
• JaxMARL & DLPack Integration: Native integration with hardware-accelerated MARL frameworks and zero-copy tensor conversion between JAX and PyTorch.
• Decoupled Rendering: Matplotlib + NetworkX visualizer operating strictly off the hot path, with moviepy frame recording capabilities.
• Semantic Bridge: Foundation model interfaces providing SIEM telemetry via state_to_text, multimodal payloads via build_vla_prompt, and LoRA+PPO fine-tuning recipes for Llama-3-8B.
• Functional Core: Immutable EnvState PyTree with deterministic apply_state_delta execution.

Quick Start

pip install 'netforge_rl[jax,render,finetune] @ git+https://github.com/reforcemind/NetForge_RL'

See the Quick Start Guide for execution loops.

Architecture Structure

netforge_rl/
├── core/               Immutable EnvState PyTree and interpreters
├── backends/jax/       Vectorized kernels and batched execution
├── environment/        Legacy PettingZoo backend
├── bridges/            JaxMARL adapter and zero-copy dlpack
├── render/             Decoupled visualization pipelines
├── semantic/           Interfaces for language model agents
├── baselines/          Reference heuristic policies and JAX PPO
└── scenarios/          Environment goal configurations

Citation

@misc{netforge_rl_2026,
  title  = {NetForge RL},
  author = {ReforceMind},
  year   = {2026},
  url    = {https://github.com/reforcemind/NetForge_RL}
}