intel/enterprise-rag-late-chunking repository overview

⁠Intel® AI for Enterprise RAG Late Chunking Service

Part of the Intel® AI for Enterprise RAG ecosystem.

⁠Overview

The Intel® AI for Enterprise RAG Late Chunking Microservice enhances embedding quality by applying chunking after token-level embeddings are generated. This microservice takes documents as input, obtains token embeddings from an embedding service, then intelligently chunks the text and pools the token embeddings to create high-quality chunk representations. This approach preserves contextual information better than traditional "chunk-then-embed" methods.

⁠🔗 Related Components

This service works within the Intel® AI for Enterprise RAG ecosystem:

• Embedding service to obtain token-level embeddings with pooling layer
• Enhanced Dataprep service to store chunked embeddings in vector database

⁠License

Intel® AI for Enterprise RAG is licensed under the Apache License, Version 2.0.

Copyright © 2024–2026 Intel Corporation. All rights reserved.