Using Ollama, Voyage AI, MongoDB Search, Vector Search, and Next.js

Recently I’ve been building a small project to explore what a modern AI-powered search application might look like.

The idea is simple.

Take a folder of images, analyze each one with an LLM, store the generated metadata in MongoDB, and then make those images searchable using both traditional keyword search and semantic vector search.

Along the way, the project ended up touching on quite a few interesting topics:

• Running vision models locally with Ollama
• Generating structured metadata from images
• Designing MongoDB documents for AI-generated data
• MongoDB Search for keyword queries
• Vector embeddings with Voyage AI
• Hybrid search using both keyword and semantic retrieval
• Building a simple Next.js application to explore the results

The goal wasn’t to build the perfect photo management application.

It was to explore practical patterns for combining AI with MongoDB in a way that’s approachable, understandable, and easy to extend.

I’ll post the video here soon!