Profile
Back to NewsBack
GitHub Trending 2 min
Reader Mode
jantacy/mobilemodels: 手机品牌型号汇总

jantacy/mobilemodels: 手机品牌型号汇总

18 hours ago

Phone Brand & Model Summary

Phone manufacturers often expose only a raw, cryptic model code (e.g. MI-ONE PLUS) at the OS/analytics level instead of the human-readable name (e.g. "小米 1 联通版"). KHwang9883's MobileModels crowdsources the mapping between these codes and their real names, one markdown file per brand.

This repo pulls that project in as a git submodule, parses every brand's markdown file, and flattens the result into a single CSV — so you can just load one file instead of parsing 20+ markdown files yourself.

Output

brand_model.csv — one row per device, with columns:

| Column | Description | |---|---| | brand | Brand slug (e.g. xiaomi, huawei, apple) | | model | Raw model code as reported by the device (e.g. MI-ONE PLUS) | | area | cn or en, depending on which regional markdown file the entry came from | | brand_name | Brand name in Chinese (e.g. 小米) | | model_name | Human-readable model name (e.g. 小米 1 联通版) |

Auto-update

A GitHub Actions workflow (update.yml) runs daily:

  1. Pulls the MobileModels submodule to its latest commit
  2. Regenerates brand_model.csv by running phoneModel.py
  3. Commits and pushes if the data changed
No manual token setup is needed — it authenticates with the built-in GITHUB_TOKEN. If you fork this repo and the push step fails with a permissions error, check Settings → Actions → General → Workflow permissions and set it to "Read and write permissions".

Running locally

git clone --recurse-submodules https://github.com/jantacy/mobilemodels.git
cd mobilemodels
pip install -r requirements.txt
python phoneModel.py
Chat with me