ItzCrazyKns/Neon-Image-Search — reverse-engineered prompt

Build me a simple image search web app where I can upload a photo and get back the most similar images from a small image collection. I want it to use Google Vertex AI to turn both the stored images and the uploaded image into embeddings, then save those vectors in NeonDB and run similarity search with pgvector. The sample dataset can just be the flower images in the public folder, and there should be a script that processes those images first and stores their embeddings before the app runs.

Make the UI clean and easy, with an upload area, a search button, and a results grid that shows the closest matches ranked by similarity. Please include the basic environment setup for the database and Google credentials, and wire it up so I can run the embedding generation step and then start the app locally. If anything is unclear, look up the current docs online and make it work end to end.