Memex(RL)

Open-source reinforcement-learning framework and code release for long-horizon LLM agents that uses indexed experience memory: compact structured working-context summaries linked to full-fidelity external stored interactions.