kun-lab builds privacy protection system by triple design:
- Data storage mechanisms::
All user data (conversation logs, uploaded files, notes, etc.) are stored in encrypted form in a local SQLite database, with the default path to the .kunlab folder in the user directory, and are not synchronized with any cloud server. - network transmission control::
The basic conversation function runs completely offline, only when the user actively opens the "Internet search" will access the search engine through the HTTPS protocol (can be turned off at any time), and the query keywords are anonymized. - privilege management system::
In multi-user mode, the data of each account is isolated independently; the picture recognition function can be set to not save the original image and only retain the text recognition results.
Privacy Enhancements::
Users can permanently delete specific session records through "Settings"-"Data Management"; Support modifying encryption algorithm (AES-256 by default) when deploying the source code; All dependent libraries have been audited for security, and there is no data collection SDK.
Comparative Advantages: Compared with cloud services such as ChatGPT, kun-lab fundamentally avoids the risk of training data leakage and is particularly suitable for handling sensitive business documents.
This answer comes from the articleKunAvatar (kun-lab): a native lightweight AI dialog client based on OllamaThe
May not be reproduced without permission:AI productivity tools " How does kun-lab's localized data storage mechanism secure user privacy?
English 
简体中文 
日本語 
Deutsch 
Português do Brasil