Hi everyone! 👋
I’m excited to write my first post on this platform, and I’m very happy to be part of this amazing developer community. 🚀
🚀 FAISS HNSW-based Vector Database for Node.js is Here!
I’ve been working on a high-performance FAISS-based vector database for Node.js, and I’m excited to finally share it with you all! 🎉
💡 eada-cpu is optimized for FAISS HNSW (Hierarchical Navigable Small World) indexing, allowing efficient KNN searches directly in Node.js—without needing Python dependencies.
🔥 Benchmark Results
| Metric | Value |
|---|---|
| Vector Dimension | 128 |
| Number of Vectors | 7,000,000+ |
| KNN Search Time |
4.05 ms 🚀 |
| Performance vs FAISS-Python | 10% - 15% faster |
| Dataset Size | ~5GB |
| Indexing Time | 1 hour 36 minutes |
💡 Fully Optimized for CPU!
This benchmark was run entirely on CPU, making it ideal for standard servers without requiring GPU acceleration.
This enables LLM RAG applications, recommendation engines, and vector searches to run efficiently and cost-effectively in Node.js.
🎯 Key Features
✅ FAISS HNSW support → High-speed, accurate KNN search
✅ Pure Node.js → No Python dependencies required
✅ Compatible with Windows / Linux / macOS (Intel & ARM64)
✅ CMake-based build system for easy cross-platform support
✅ N-API & Prebuilt support → Install easily with:
bash
npm i eada-cpu