Metadata filtering can significantly enhance retrieval accuracy in vector RAG applications

Recently, we introduced support for this feature in LangChain and LlamaIndex, utilizing node properties to refine results from Neo4j. Graph-based systems offer even more options for sophisticated metadata filtering, further expanding the capabilities for refining results.
Check it out:

Code: blogs/llm/graph_based_prefiltering.ipynb at master ยท tomasonjo/blogs ยท GitHub
Metadata filtering in LangChain: Neo4j Vector Index | ๐Ÿฆœ๏ธ๐Ÿ”— LangChain
Metadata filtering in LlamaIndex: llama_index/docs/docs/examples/vector_stores/neo4j_metadata_filter.ipynb at main ยท run-llama/llama_index ยท GitHub

Thanks to @bratanic_tomaz for this ongoing project!

1 Like