This Swift demo app shows you how to build real-time native AI-powered apps for Apple devices using Unum's Swift libraries and quantized models. Under the hood, it uses UForm to understand and "embed" multimodal data, like multilingual texts and images, processing them on the fly from a camera feed. Once the vector embeddings are computed, it uses USearch to provide a real-time search over the semantic space. That same engine also enables geo-spatial search over the coordinates of the images and has been shown to scale even to 100M+ entries on an 🍏 iPhone easily.
The demo app is capable of text-to-image and image-to-image search and uses vmanot/Media libra to fetch the camera feed, embedding, and searching frames on the fly. To test the demo:
# Clone the repo git clone https://github.com/ashvardanian/SwiftSemanticSearch.git # Change directory & decompress the images dataset.zip, which brings: # - `images.names.txt` with newline-separated image names # - `images.uform3-image-text-english-small.fbin` - precomputed embeddings # - `images.uform3-image-text-english-small.usearch` - precomputed index # - `images` - directory with images cd SwiftSemanticSearch unzip dataset.zip
After that, fire up the Xcode project and run the app on your fruity device!
Links:
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4
