robotmem: Persistent Experience Memory for Robot LLMs

An open-source episodic memory system that gives robots persistent memory across episodes. Instead of discarding
  experience after each episode, robots can learn, store, and recall physical experiences.

  Key results on MuJoCo FetchPush-v4:
  - +25% success rate with memory-augmented policies
  - +8% cross-environment transfer (FetchPush → FetchSlide)

  7 APIs: learn / recall / save_perception / forget / update / start_session / end_session

  Model-agnostic, SQLite-based, no external services needed.

  GitHub: https://github.com/robotmem/robotmem
  Website: https://robotmem.com
  PyPI: pip install robotmem