Introducing Contextual Retrieval

This is pretty old -- but it certainly makes sense. Attaching metadata for either embeddings or for LLM prompting can help

could you please tell me what would be the most up to date sources?

my idea is to interpret open source llm with llamaindex instead of using claude, with milvus

Here's one using some dedicated extractor
https://docs.llamaindex.ai/en/stable/examples/metadata_extraction/DocumentContextExtractor/

Here's one more from scratch
https://docs.llamaindex.ai/en/stable/examples/cookbooks/contextual_retrieval/

The main thing that makes this feasible though is anthropics ability to cache. Without caching, it will be slower, and more expensive