Embedding Atlas

Use it when

• You have hundreds of thousands to millions of embedding vectors and cannot tell what structure, clusters, or outliers exist in them
• You need to cross-filter an embedding projection against metadata columns and inspect nearest neighbors of individual points interactively

What it solves

Not the fit when

• serving embeddings as a vector database
• computing or fine-tuning embeddings
• datasets far beyond a few million points where WebGPU rendering degrades
• vector similarity search serving
• embedding model training
• production RAG retrieval quality

Install

pip install embedding-atlas

Invoke

embedding-atlas <your-dataset> to launch the interactive viewer; or in a notebook: from embedding_atlas.widget import EmbeddingAtlasWidget; EmbeddingAtlasWidget(df)

Alternatives

No reviewed alternatives recorded yet.