SuperCompress: Open-source LLM prompt compression for cost reduction

## SuperCompress: Open-Source LLM Prompt Compression [SuperCompress](https://github.com/arjunkshah/supercompress) - MIT licensed, cuts ~65% of LLM token costs before inference using a ~5K param CPU policy. Live demo: https://supercompress.dev CPU-only, model-agnostic, 100% oracle recall. pip install supercompress.