Lumi-node/infinite-context

1---2title: Infinite Context3emoji: 🧠4colorFrom: blue5colorTo: purple6sdk: gradio7sdk_version: 4.0.08app_file: app.py9pinned: true10license: mit11tags:12  - retrieval13  - memory14  - llm15  - hat16---17 18# Infinite Context - Live Demo19 20**Give any LLM unlimited memory with sub-millisecond retrieval.**21 22## What This Demo Shows23 24This is a live demonstration of HAT (Hierarchical Attention Tree) - a retrieval system that:25 26- **100% accuracy** finding relevant conversations27- **< 1ms search time** across hundreds of thousands of tokens28- **1,400x context extension** for small models29 30## How to Use31 321. Click **Initialize** to create a simulated conversation history332. Ask natural questions like:34   - "What did we do to fix the React error?"35   - "How much did we speed up the Python script?"36   - "What was causing the Kubernetes pods to crash?"373. See HAT retrieve the exact relevant conversations in milliseconds38 39## Performance40 41| Metric | Value |42|--------|-------|43| Search Latency | < 1ms |44| Retrieval Accuracy | 100% |45| Context Extension | 1,400x |46 47## Links48 49- [GitHub](https://github.com/Lumi-node/infinite-context)50- [Docker Hub](https://hub.docker.com/r/andrewmang/infinite-context)51- [ArXiv Paper](https://arxiv.org/abs/2026.XXXXX) (coming soon)52 53## License54 55MIT56