StarlightSearch/EmbedAnything

Health v2 · maintenance only

Adoption not part of the score

1305 stars · 143 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-30, confidence not recorded

EmbedAnything is a high-performance, memory-safe embedding pipeline written in Rust (with Python bindings) that generates embeddings from text, PDFs, images, audio, video, and webpages. It streams vectors efficiently to any vector database and supports dense, sparse, ONNX, model2vec, and late-interaction embedding models, including cloud providers.

Use cases

• generate embeddings for pdfs and documents
• build a rag pipeline with a vector database
• embed images and audio files locally
• index a directory of files into a vector store
• run embedding inference without pytorch
• chunk and embed webpages for search
• use colbert or colpali late-interaction embeddings

When to choose

• you need fast, low-memory, production-ready embedding ingestion in Rust or Python
• you want multimodal embedding (text, images, audio, video, pdfs) in one pipeline
• you want to avoid a PyTorch dependency for deployment
• you need streaming indexing into multiple vector databases with minimal code

When to avoid

• you need to train or fine-tune embedding models
• you only need a hosted embedding API with no local pipeline
• your stack is pure Python and you prefer native PyTorch tooling like sentence-transformers

Facets

library · maturity active

machine-learning llm-inference rag vector-database search-engine ocr audio-processing video-processing pdf etl machine-learning large-language-models developer-tools rust python cross-platform cloud self-hosted embeddings multimodal onnx candle vector-streaming chunking late-interaction colpali colbert reranker whisper information-retrieval retrieval-augmented-generation search natural-language-processing gpu

Member repositories

For agents

markdown · JSON · MCP: product_card(name="StarlightSearch/EmbedAnything")