LokalBot
Features Benchmarks Privacy Support
Download

Help and feedback

Support

LokalBot is an open-source community project. Start with the fixes below; the GitHub issue tracker is the public support desk.

Common problems

Only my side of the call was recorded

On a Mac, LokalBot records the other side through macOS system-audio capture. Without that permission the app warns that the recording is microphone-only. Open System Settings → Privacy & Security → Screen & System Audio Recording, allow LokalBot to record system audio, then make a short test recording in the meeting app you use. On Windows and Linux, LokalBot records the meeting app alone where it can, and otherwise everything your default output device plays, so in that case make sure the meeting plays through it; on Linux, system audio needs PulseAudio or PipeWire.

A meeting started but LokalBot did not ask to record

On a fresh install, LokalBot asks through a notification before it records a detected meeting, so allow its notifications (on a Mac, in System Settings → Notifications). On a Mac, browser meetings are detected through Accessibility, and calendar access helps match scheduled meetings. On Windows and Linux, LokalBot notices apps that start using the microphone; on GNOME, its tray icon needs the AppIndicator extension. You can always start a recording yourself or switch to automatic recording in Settings.

A model download failed or stalled

Models download from Hugging Face or the publisher's host the first time you use them. Check your connection and free disk space (speech models range from about 0.7 GB to 3.2 GB), then retry from Settings → Models.

The transcript is in the wrong language or misses words

Try another speech model in Settings → Models. On a Mac, Qwen3-ASR 1.7B is the default and covers 52 languages and dialects, and Whisper large-v3 turbo covers 99. On Windows and Linux, the default is Qwen3-ASR 0.6B, with Parakeet, SenseVoice, GigaAM for Russian, and Whisper turbo as alternatives. The choice matters most on noisy or multilingual recordings.

Where is my data, and how do I remove it?

By default the library lives in ~/Library/Application Support/me.dotenv.LokalBot/ on a Mac, %LOCALAPPDATA%\LokalBot on Windows, and ~/.local/share/LokalBot on Linux: recordings, transcripts, summaries, the search index, downloaded models, and logs (debug.log on a Mac; the logs folder on Windows and Linux). Delete individual meetings in the app, or quit LokalBot and remove that folder to erase everything.

Does LokalBot run on an Intel Mac or macOS 14?

No. On a Mac, LokalBot needs Apple Silicon (M1 or later) and macOS 15.0 or later. See the system requirements.

Does LokalBot run on Windows or Linux?

Yes, as an alpha: Windows 10 or 11 (x64) and Linux (x86_64, glibc 2.39 or later). The Windows build isn't code-signed yet, so SmartScreen asks before the first launch; choose More info, then Run anyway. Some features depend on your Linux desktop, and a few, such as Gmail in meeting briefs, are Mac only for now. See the system requirements.

Before opening an issue

Install the latest release, retry the smallest failing flow, and check the README.

Report a bug or request a feature

Open a GitHub issue

Include the LokalBot version, your operating system and its version (on Linux, also the distribution, desktop, and X11 or Wayland), your Mac model or PC's CPU and GPU, memory, the selected models, the exact action that failed, and reproducible steps. A redacted excerpt from the log can help.

Never post meeting audio, transcripts, screenshots, API keys, tokens, model-server credentials, or other private material.

Security reports

Do not disclose vulnerabilities in a public issue. Use GitHub's private vulnerability-reporting form. If it is unavailable, open a public issue with no exploit details and request a private contact channel.

Response expectations

Support is best-effort and response times are not guaranteed. Clear reproduction steps and a privacy-safe log excerpt make a response much more likely.

LokalBot

Your personal context engine.

Project

GitHub About Blog Benchmarks Privacy Terms Support

Community support. Keep sensitive data private.