How to Apply Filters in VDB for RAG Chatbot

Problem/Motivation

I’m integrating a Zilliz vector database with an AI-powered RAG chatbot. I have filterable attributes (e.g., case studies, approaches) stored as integers. When I query the VDB via the AI explorer, these attributes are returned, but I’m not sure how to pass filters into the query so that results are pre-filtered.

In short, I want to apply filters at query time so Zilliz returns only the filtered chunks.

Steps to reproduce

Use Zilliz for AI search and connect an AI assistant with knowledge grounded in results from the VDB.
Observe that filterable attributes are present in results but cannot be applied as query-time filters through the current interface.

Proposed resolution

Enable query-time filtering so the VDB returns only chunks matching selected attributes (e.g., case studies). Currently, the chatbot must post-filter results, which increases latency and requires fetching more chunks than necessary.

Remaining tasks

I am not sure if I am able to complete these steps from my end. If so I'd love to know where I should start making the changes:

• Expose a filter input in the AI bot UI so users can choose attribute filters before running a search; or
• Implement server-side filtering by passing filter expressions to the Zilliz VDB at query time.