Hi everyone,
I was thinking more about the AI chatbot project and had a technical question around how information retrieval could be handled effectively within Jenkins.
Given that Jenkins has multiple knowledge sources (official docs, plugin metadata, community discussions), would it make more sense to:
-
Use a single unified vector index for all sources
or -
Maintain separate indices per source (docs vs plugins vs discussions) and combine results at query time?
My intuition is that separate indices might help with better relevance and filtering (e.g., prioritizing official docs over forums), but I’m curious how others have approached this in similar systems.
Would love to hear thoughts or any existing approaches used in Jenkins-related tooling.