Question

You are building a RAG application for a SaaS platform that serves multiple client companies. Each client's internal documents are embedded and stored in a single Databricks Vector Search index. A strict requirement is that when an employee from Company A submits a query, the retrieval step must NEVER return chunks belonging to Company B, even if those chunks are semantically more similar. Which approach best satisfies this requirement during data preparation and retrieval?

Accepted Answer

Store a company_id column alongside each chunk in the Delta source table and apply a metadata filter on company_id in the Vector Search similarity query.

Answer

Create a single index but rely on a re-ranker after retrieval to demote chunks that do not belong to the querying company.

Answer

Prepend the company name to each chunk's text before embedding so the vector representation naturally separates companies.

Answer

Lower the number of returned results (k) so that only the most similar chunks are returned, reducing the chance of cross-company leakage.