Please do not open a public issue for a security problem. Use GitHub's private reporting: https://github.com/peteyagmin/localnotebook/security/advisories/new.

Include what you found, how to reproduce it, and the version (Settings, or GET /api/v1/health). You can expect an acknowledgement, and a fix or an explanation, as time allows: this is a single-maintainer hobby project.

Only the latest release receives fixes.

• The server listens on 127.0.0.1 only. Every request to /api needs a random token created at each launch (compared in constant time), and requests from other origins are rejected.
• Documents, web pages and videos are treated as untrusted: parsers run in a separate process with a timeout (and a memory limit on Linux), and instructions inside documents are quoted data, not commands.
• Web fetches accept only http and https, check every resolved address, pin the connection to the checked address, re-check redirects, and refuse localhost and private networks. Responses are limited to 20 MB.
• Imported notebook files are checked for unsafe paths, links, oversized or suspicious archives, and damaged or newer databases before anything is unpacked into place.
• There is no telemetry, analytics or update check. The only network traffic is to your configured model server and to actions you start (web import, Discover, model downloads); each can be switched off in Settings.

• Your data folder, exported .lnb files and the optional model-server API key are not encrypted. The API key is stored in plain text in app.db and is shown in Settings. Use your operating system's disk encryption and file permissions.
• If you point the app at a model server that is not on your computer, your questions and passages from your documents are sent there. The app shows a banner when this is the case.
• Anyone who can run programs as you on your computer can read your data folder.