Question 1

How do I fix missing javascriptcoregtk-4.1 and libsoup-3.0 system libraries when building a Tauri desktop app on Ubuntu?

Accepted Answer

Install Tauri's Linux system dependencies before running `cargo check` or `bun tauri:dev:desktop`. On Ubuntu 24.04/Debian:

```bash
sudo apt update
sudo apt install -y libwebkit2gtk-4.1-dev libjavascriptcoregtk-4.1-dev \
  build-essential curl wget file libxdo-dev libssl-dev \
  libayatana-appindicator3-dev librsvg2-dev libsoup-3.0-dev
```

Then rerun:

```bash
cd src-tauri
cargo check
```

or from the project root:

```bash
bun tauri:dev:desktop
```

This resolves errors such as `The system library javascriptcoregtk-4.1 required by crate javascriptcore-rs-sys was not found` and `The system library libsoup-3.0 required by crate soup3-sys was not found`. On Fedora or other distributions, install the equivalent packages from Tauri's official Linux prerequisites documentation. Also ensure `sccache` is installed because `src-tauri/.cargo/config.toml` requires it.

Question 2

How do I fix “Could not create default EGL display: EGL_BAD_PARAMETER” when running a Tauri AppImage on Linux Wayland?

Accepted Answer

For Thunderbolt AppImage v0.1.101 on Arch Linux/Wayland, preload the system Wayland client library before starting the AppImage:

```bash
export LD_PRELOAD=/usr/lib/libwayland-client.so
./Thunderbolt_0.1.101_amd64.AppImage
```

If you extracted the AppImage, run:

```bash
export LD_PRELOAD=/usr/lib/libwayland-client.so
cd squashfs-root
./AppRun
```

If the library is in a different location, find it with:

```bash
ldconfig -p | grep libwayland-client
```

The error occurs because the AppImage bundles a `libwayland-client.so` that can be incompatible with the user’s Wayland compositor. The permanent fix is to build the AppImage without that bundled library so it loads the system copy; this is addressed by excluding `libwayland-client.so` from the Linux AppImage. `WEBKIT_DISABLE_DMABUF_RENDERER=1` and `WEBKIT_DISABLE_COMPOSITING_MODE=1` do not resolve this specific error.

Question 3

Why does signing in on a new device restart onboarding and reset the onboarding state on existing devices after enabling PowerSync?

Accepted Answer

This was caused by the new device seeding the default setting `user_has_completed_onboarding = false` before PowerSync's first sync downloaded the existing `true` value. That local default write was queued in `ps_crud` and uploaded when sync was enabled, overwriting the account state through last-write-wins. The issue is fixed by PR #1317 / `fix(THU-919): stop a new device overwriting or losing account state on first sync`. Update to a release containing that fix. As a temporary workaround on older versions, avoid enabling sync on a newly signed-in device until the patched version is installed; if the state was already overwritten, complete onboarding again after updating.

Question 4

How do I fix LM Studio model test/add failing with `error_type: "ssrf"` in Thunderbolt Docker?

Accepted Answer

This happens because Thunderbolt's backend proxy runs inside Docker, where `localhost` refers to the container—not your host machine running LM Studio—and the proxy rejects the loopback request as SSRF.

The fix is PR #1075, which makes Custom providers on loopback URLs connect directly from the browser. Update Thunderbolt to a release containing this fix, then:

1. In LM Studio, start the API server and enable CORS in the server settings.
2. In Thunderbolt, set the Custom provider URL to:

`http://localhost:1234/v1`

3. Reload the model list, then click **Test Model** and **Add model**.

Using `/api/v1/` or the OpenAI-compatible `/v1/` path alone does not resolve the SSRF failure on older versions; the proxy-bypass fix is required.

Question 5

How do I fix Google or Microsoft OAuth login redirecting to an external error page because the Client ID is missing in Thunderbolt?

Accepted Answer

Set the required OAuth environment variables in your Docker/Kubernetes backend configuration, then redeploy: - Google: `GOOGLE_CLIENT_ID=` - Microsoft: `MICROSOFT_CLIENT_ID=` For Docker, for example: ```bash docker run -e GOOGLE_CLIENT_ID=xxx -e MICROSOFT_CLIENT_ID=xxx ... ``` In Kubernetes, add them under `containers[].env` or use a Secret/ConfigMap. After the fix from PR #690, Thunderbolt checks the backend `configured` flag and `buildAuthUrl` throws a clear error naming the missing variable instead of redirecting to Google/Microsoft. If you are on an older release, upgrade to a version containing that fix or set the missing Client ID environment variable manually.

thunderbolt

Core Features

What It Can't Do

Use Cases

Tags

Getting Started

Similar Projects