No privacy policy or IP protection

[Eccenux]

The project seems to be missing a privacy policy and intellectual property protection information.

The docs mostly don't mention what data is sent and when the data is sent to the AI provider.

1. Are there any steps taken to avoid sending passwords and secret keys to remote servers?
2. Are there any steps taken to avoid sending personal information to remote servers (GDPR-related)?
3. What is sent when I just enable the plugin? Whole open projects? The POM files? All XML files (of which some might contain security keys)?
4. What is sent when I use the AI Assistant feature on the project?
5. What is sent when I use the AI Assistant feature on a folder?
6. What is sent when I use autocomplete? The previous line? The whole file I'm working on?
7. Is any other remote service involved other than the selected provider (e.g., a proxy)?
8. Is data sent to e.g. OpenAI sent as a temporary data (that is discarded) or is the data used to train model and do other undisclosed operations?

[stefanofornari]

Hi @Eccenux,
Thanks for your feedback; all are very valid points. At the moment any user shall use this plugin at his/her own risks.
I agree this should be clearly stated in the documentation and readme. @jGauravGupta maybe we can add something for the next release?

[jGauravGupta]

Thank you for raising this important concern, I will update the docs with these concerns.

Currently, Jeddict AI has two modes: ASK and AGENT.

In ASK mode, the plugin sends only the explicitly selected context to the configured AI provider (for example, the active editor selection, the prompt text entered by the user, or files/folders explicitly chosen for an AI operation — see the Context Scopes documentation: https://jeddict.github.io/page.html?l=tutorial/AIContext). The plugin does not automatically upload the entire project when it is enabled. Additionally, users can filter or restrict specific files and directories in the settings to ensure they are not included in ASK mode context.

In AGENT mode, only the paths of the “Session Context Files” are initially sent, not their contents. The agent then decides which files need to be accessed and sent as part of the interaction. In contrast, in ASK mode, the contents of all “Session Context Files” are sent along with the request.

In an upcoming release, we will introduce another mode called Interactive.
@stefano, could you please provide more details about this mode?

[stefanofornari]

From this standpoint, interactive mode works exactly like AGENT mode. The only difference is that for tools that may have a side effect (e.g. creating/editing files, running the build, etc.), Jeddict requires the user to confirm the action.