LM Studio Weight Watcher

LM Studio Weight Watcher tells you when a model you already have in LM Studio has newer weights on Hugging Face.

It does not download or replace anything. It only checks and alerts.

Why It Exists

LM Studio downloads are snapshots. If the upstream model changes later, there is no built-in persistent update alert. This fills that gap.

Usage

Run once:

python .\lmstudio_weight_checker.py

Show all checked models:

python .\lmstudio_weight_checker.py --all

Output JSON:

python .\lmstudio_weight_checker.py --json

Use a custom models folder:

python .\lmstudio_weight_checker.py --models-root "D:\LM Studio models\.cache\lm-studio\models"

Run the watcher once and update alert state:

python .\lmstudio_weight_watcher.py --once

Start the tray app:

python .\lmstudio_weight_watcher.py

Use the launcher:

.\run_watcher.bat

Start at login:

powershell -ExecutionPolicy Bypass -File .\install_startup_task.ps1

Remove the startup task:

powershell -ExecutionPolicy Bypass -File .\remove_startup_task.ps1

Change the check interval or reminder timing:

python .\lmstudio_weight_watcher.py --check-interval-hours 4 --reminder-interval-minutes 30

Change the watcher timeout:

python .\lmstudio_weight_watcher.py --timeout-seconds 45

Notes

• By default, the script discovers the models folder from %APPDATA%\LM Studio\settings.json.
• It uses blob-hash comparisons with a relative-path hash cache to avoid re-hashing large .gguf files on every check.
• Persistent watcher state is stored by default in %APPDATA%\LM Studio Weight Watcher\state.json.
• Existing installs using the old LM Studio Weight Updater state path are still recognized automatically.
• The launcher first checks %LMSTUDIO_WATCHER_PYTHON%, then a local .venv, then python, then py -3.
• If the hidden startup launcher cannot start the watcher, it writes the reason to watcher-launch.log.

License

This project is released under The Unlicense.