https://github.com/smart-models/smart-embedder

A lightweight, self-hosted embedding server built for hybrid search pipelines. It runs entirely on your own hardware on an NVIDIA GPU for high throughput, or on CPU when no GPU is available with no cloud dependency and no data leaving your machine.

bge-m3 docker embedding fastapi hybrid-search qwen-embeddings qwen-reranker rag reranker self-hosted

Last synced: 28 Jun 2026