Skip to main content
Navigator can use project documents as retrieval context. Uploaded documents are chunked, embedded, stored, and searched when the assistant needs grounded answers from your files.

​
Retrieval flow

​
Setup outline

1

Enable Supabase extensions

Run the project migrations that enable pgvector, pgmq, pg_cron, and the RAG utility functions.
2

Deploy the embed function

Deploy the Supabase Edge Function used to generate embeddings asynchronously.
3

Configure secrets

Add Supabase credentials and any optional AI Gateway settings.
4

Upload and test

Upload a document, monitor embedding progress, then ask a question that should retrieve from that document.

​
Search endpoints

Use project-specific search when a thread should only draw from one archive or customer workspace. Use global search for broad personal knowledge lookup.