Most operating systems still rely heavily on filenames when searching images. Once your collection grows into hundreds or thousands of photos, finding exactly what you are looking for becomes surprisingly difficult.

This project explores a different approach. Instead of searching filenames, a vision model understands each image, generates structured metadata, stores everything in MongoDB, and layers increasingly capable search techniques on top of that data.

Helpful links

• Read README: README.md
• Live example: images.seemongo.com
• Tutorial video: Building an AI-Powered Image Search Application with MongoDB
• Source code: github.com/learnmongo/ai-image-viewer

Built with MongoDB Search, MongoDB Vector Search, Ollama, Voyage AI, and Next.js.

Starting with nothing more than a folder of images, you build a search application capable of understanding what is actually inside each image. The search experience grows one layer at a time.

One of the goals of this project was to keep the overall architecture simple. Everything revolves around a single MongoDB document per image.

Images
   │
   ▼
Vision Model (Ollama)
   │
   ▼
Structured Metadata
   │
   ▼
MongoDB Documents
   ├── MongoDB Search
   └── Vector Embeddings
            │
            ▼
       Hybrid Search
            │
            ▼
      Next.js Application

Rather than introducing multiple databases or external search systems, MongoDB becomes the central source of truth. Metadata, embeddings, and search indexes all live together. That simplicity is one of the best parts of this architecture.

As the project grew, a few design decisions ended up making a big difference.

• Two LLMs are better than one. The vision model focuses on understanding the image. A second instruction model transforms that understanding into structured JSON. Prompts are easier to iterate on and documents are more consistent.
• Embeddings are generated separately. They can be regenerated later without analyzing every image again, and experimenting with different embedding models is much easier.
• Everything lives in one MongoDB document. Titles, descriptions, tags, colors, feelings, prompt history, model information, embeddings, and location data. The document grows as the application grows.
• Hybrid search gives you the best of both worlds. Keyword search is precise. Vector search understands meaning. Hybrid search combines both, which is the experience many real-world applications need.

If you are exploring the code after watching the video, these are the best places to start.

Suggested reading order

• README.md
• ARCHITECTURE.md
• tools/process/
• MongoDB documents
• generate-embeddings.js
• lib/image/queries/
• SEARCH.md
• OLLAMA.md
• PROMPTS.md

That order mirrors the tutorial and builds the project one concept at a time.

These sections map directly to the repository. Timestamps point to where each topic appears in the tutorial.

The README gives you the high-level overview. These documents take a deeper look at individual parts of the project.

Image search is really just the beginning. The same architecture can be applied to many other types of data.

• Documents
• Knowledge bases
• Product catalogs
• Support tickets
• Internal tools
• Retrieval-Augmented Generation (RAG)
• Recommendation systems

Once your data has been enriched with AI and stored in MongoDB, you can continue building on top of it without introducing additional systems. That is what makes this pattern so powerful.

Learn more: learnmongo.com