How can I Implement MongoDB's Native Vector Fields in Payload CMS for Efficient Similarity Search?

[proper oxide]

Does anyone know if it's possible to create a vector field in Payload CMS that utilizes MongoDB's native vector data type (introduced in MongoDB 7.0) for efficient vector similarity searches?

I'm aiming to implement AI-powered search within Payload without relying on external tools that might complicate the setup. Specifically, I want to:

• Define a vector field in my Payload schema that maps directly to MongoDB's actual vector type, ensuring optimal performance for similarity queries.
• Perform vector similarity searches in a single database query without needing to fetch documents twice or perform additional processing outside of MongoDB.
• Retrieve and resolve the resulting documents in Payload with the desired depth (including nested relationships), based on the similarity search results.

Has anyone successfully implemented this or can offer guidance on how to set up such a vector field in Payload, execute efficient similarity queries, and seamlessly integrate the results with Payload's document resolution features?

Any examples or insights would be greatly appreciated!

[harsh prism]

New Community-Help Thread Created!

Help is on the way! To mark it as solved, use the /solve command. In the meantime, here are some existing threads that may help you:

Documentation:

• MongoDB - Preserving externally managed data
• MongoDB - Options
• Database - Selecting a database - When to use MongoDB
• Search Plugin
• Vercel Content Link - How it works

[olive halo]

Hi Livog, did you ever make any progress on this? I would be open to callobarating on plugin or something that does this. I know you can access the monog database directly but that is not very payload like

This is one of Payload entripise features so I doubt we would get much help (understandable)