ColPali

ColPali is a document retrieval library for developers and researchers building search or retrieval-augmented generation systems. It matches text queries to page images, using visual content alongside text. The colpali-engine package is deprecated, and maintainers recommend Sentence Transformers for new projects and production use.

• Search visual documents: Retrieve pages using their text, charts and layout without a separate OCR or layout recognition pipeline.
• Train document retrievers: The Python code supports training and inference for the original ColPali model and related ColVision and bi-encoder variants.
• Choose model families: ColPali builds on PaliGemma, while ColQwen2 uses Qwen2-VL and supports images at different resolutions.

The code runs with PyTorch and is open source under MIT. Apple Silicon users can run ColQwen models through MPS. Model licenses are separate from the code license: the listed ColPali checkpoints use the Gemma license, and the listed ColQwen2 checkpoints use Apache 2.0.

Sentence Transformers supports ColPali-style models through MultiVectorEncoder. The original repository and package remain available for existing projects, research and reproduction of results.