How do you implement document versioning and freshness in RAG systems?

Document versioning in RAG ensures users get answers from the most current version of documents while maintaining access to historical versions when needed. Implementation requires timestamping each document chunk during indexing, maintaining version metadata linking chunks to their source document version, and implementing freshness-aware retrieval that prefers newer versions by default. When documents are updated, re-chunk and re-embed the new version while marking old chunks as superseded rather than deleting them. Implement a freshness decay function that slightly penalizes older document relevance scores. For regulatory and compliance contexts, maintain the ability to query against historical document states. Handle conflicting information across versions by including version dates in the LLM context and instructing it to prefer the most recent source. Automate document ingestion pipelines that detect changes in source systems and trigger re-indexing.