FeedDiscussion
How to Build Incremental RAG Indexing in .NET: A Real-World Walkthrough
Incremental RAG indexing is what separates a demo knowledge base from one you can still trust six months later. The first ingestion run is easy: read every document, chunk it, embed it, write it to a
codingdroplets.com13 min read
#net#csharp#aspnet-core#ai#artificial-intelligence#generative-ai#rag#vector-database#data-engineering#backend
"The trouble starts on day two" is exactly right, and deletions are the case most incremental indexers get wrong. An updated document at least produces a new embedding; a deleted one leaves its vectors in place answering questions forever, and nothing notices because nothing is looking for absence.
Testing the edit that shrinks the chunk count is a good sign you thought about it, since that is the same bug in miniature: chunks 5 through 8 need removing, not just 1 through 4 rewriting.
The other thing worth carrying into chunk metadata is a document version, so an answer can state how old its source is and a stale-content audit becomes a query rather than a project.