Skip to content

Troubleshooting

The message is “Backend Jade introuvable”. The installed backend is looked for in backend\Jade.exe next to the app. In development, run jade serve, then reopen the app.

The local engine does not answer after 60 seconds

Section titled “The local engine does not answer after 60 seconds”

Close and reopen the app. If it persists, check the antivirus (it may block Jade.exe) or reinstall. Read C:\ProgramData\Jade\logs\jade.log.

The app does not start after install, with PermissionError on .env

Section titled “The app does not start after install, with PermissionError on .env”

This was fixed in 1.10.1. Make sure you run 1.10.1 or later: the Users group needs modify rights on backend\.env.

The first launch is stuck downloading the models

Section titled “The first launch is stuck downloading the models”

Models are about 3.2 GB. The download resumes after a cut. The default source is the archive at MODELS_URL, with HuggingFace as a fallback, which can throttle anonymous downloads. Check the disk space, and set MODELS_URL to an archive you host. Host it on object storage that supports Range requests rather than behind a proxy with a short timeout: a download cut near 2 GB is the usual sign of that.

The system status panel in the settings shows the cause: inference server unreachable, models not ready, or less than 2 GB free on the data disk. Press refresh after fixing it.

Check that the folders are listed in the settings and online, that the index has finished, and that the user can read the files (NTFS filtering hides the others). Lower MIN_SCORE if too much is filtered out.

If a banner says the subscription expired or is suspended, contact whoever runs your jade server. See Licensing.

Use the feedback button: it builds a diagnostic zip with no secret and no conversation content.