React Native Local First Semantic Search
summary
Builds a local‑first semantic search inside a React Native app using an on‑device ONNX model from Hugging Face (Snowflake Arctic quantized4), ONNX Runtime React Native, and Hugging Face tokenizers, producing 384‑dimension embeddings via token averaging, indexed with hierarchical navigable small world (HNSW) for approximate nearest neighbors. The demo is a restaurant finder, returns results in milliseconds on an emulator, uses cosine distance, and publishes code on GitHub.
AI-written summary · about
chapters (11)
- 0:00Local first semantic search
- 1:25Semantic vs keyword search
- 2:49Vector embeddings in AI
- 5:18Choosing AI embeddings model
- 7:25Local AI inference
- 9:57Tokenization and text conversion
- 12:44Visualizing AI outputs
- 14:30Fast search with HNSW
- 16:59Search in React Native
- 19:27Live demo: restaurant finder
- 21:17Closing and resources
related talks
17mThink Fast: OnDevice Vector Features
26mBuild your Machine Learning Model on Edge with React Native
19mReact For Social Change
24mReact Native Ignite: Building an AI app
23mChallenges in Building React Native Applications with On-Device AI
47mBoost Your Mobile Apps with Machine Learning and NativeScript
10mMachine Learning And React Native
30mGraphQL And React Native - The urql Edition