Skip to content

Bah 1.7.10 — Linux (EXPERIMENTAL · testers only)

Pre-release
Pre-release

Choose a tag to compare

View all tags
@github-actions github-actions released this 26 Jul 19:45
· 3 commits to main since this release
v1.7.10-linux-test
cafc3d6

EXPERIMENTAL — testers only

This is an untested Linux build. It compiles and passes CI (the app opens and does not crash headless), but nobody has run it on real hardware yet — I don't have a Linux machine.

Do not treat this as a stable release. The stable, tested build is Windows — see the Latest release.

How to run

chmod +x Bah-*.AppImage
./Bah-*.AppImage

If it refuses to start, try --no-sandbox. On some distros you may need libsecret and fuse2.

What should work

The whole core is cross-platform: the AI agent, ad/tracker blocking, YouTube ad skipping, local AI via Ollama, cloud API keys, document Q&A, background monitors, voice input.

What does NOT work on Linux

Anything that plays or downloads media — "play a song", "open a video about X", "download this music", "make a playlist". All of it resolves the video through yt-dlp, and the app fetches the Windows build of yt-dlp/ffmpeg, which cannot run here. It degrades with an error — it does not crash. Playing audio inside a page you opened yourself (YouTube in a tab) is normal Chromium and is unaffected.

Read-aloud (the 🔊 button and "read this page") — uses the OS's built-in text-to-speech voices. Windows always ships one; most Linux installs (and the AppImage sandbox) don't have speech-dispatcher set up, so there's no voice to speak with. As of this build the button just does nothing instead of pretending to read — if you install speech-dispatcher + a voice (e.g. espeak-ng) it should start working with no app update needed.

Also: auto-update is Windows-only. To update, download a newer AppImage.

Not a Linux problem, but worth knowing

There's no built-in free AI chat anymore. The old no-key fallback (Pollinations) never worked reliably enough to keep, so it was removed. To actually exercise the AI, use a cloud key (DeepSeek, Mistral, NVIDIA) or a local model with Ollama. Image generation still works for free, no key needed.

What I'd love to know

If you install it, the most useful thing you can tell me:

  1. Does it open at all? (that alone is useful)
  2. Does the AI agent actually click things? Open the assistant, ask it to do something on a page — e.g. "open youtube and search for X" — and tell me whether it really operated the page, or just talked about it. That's the part I most need confirmed on Linux.
  3. Anything that looks visually broken.

Open an issue with whatever you find, even if it's just "it doesn't launch". That's the point of this build.

Assets 3
Loading