Curate and maintain RAG knowledge base

Select and chunk documents for RAG knowledge base

Preparing heterogeneous documents for retrieval-augmented generation systems.

• · source documents
• · chunking strategy
• · relevance criteria

• · chunked passages
• · metadata
• · indexed embeddings

• · embedding models
• · vector databases
• · chunking library

Source documents available; embedding model selected; chunk size tuned

Chunks too large (loses specificity); too small (loses context); metadata loss

• · RAG-specific focus
• · chunking strategies explained