How it works
Number Extractor
Your text is sent to a local WebGPU worker running Qwen3.5 2B. The model converts number words to digits and returns every number it finds, without sending your text to a server. It is prompted with English and Portuguese examples, including multiple numbers, negative values, decimals, and quantities such as half a dozen and meia dúzia. Identifiers and shorthand such as hello123world and 12.5k are parsed directly.
- Uses @huggingface/transformers with WebGPU acceleration
- Recognizes several numbers, number words, digits, currencies, and shorthand
- Downloads about 1.6 GB on first use, cached and shared with Local LLM Chat
- First run warms the model for faster follow-up requests
- Works best in English and Portuguese. Other languages can produce incorrect results; check the output.
Model: Qwen3.5 2B ONNX, based on Qwen3.5 2B under the Apache 2.0 license.
graph TD
A[Input text] --> B[Web Worker]
B --> C[Qwen3.5 2B model]
C --> D[Extracted numbers]
D --> E[History + copy]