- Zig 95.9%
- Objective-C 3.9%
- C 0.2%
| Filename | Latest commit message | Latest commit date |
|---|---|---|
a6257c714c
miniaudio, being a temporary implementation of OS audio integration, was never really polished much, which left in it a couple big bugs relative to the lifecycle management of callers. - callers would join/leave concurrently as the audio thread was running causing UB for complete lack of coordination between threads - resource cleanup at the end of a call failed to check a flag and would leak audio resources the first point has been fixed by introducing a ring buffer that collects essentially join/leave requests that then get processed by the audio thread at every loop. currently the audio thread will process all requests at once, while instead it should try to stay within a time budget, but that's an improvement that can be added later. More problematic is the fact that currently the audio thread frees caller resources, which is something that should be avoided when working on a handcrafted OS audio integration. With this change we should not experience anymore crashes when joining / leaving a call when using the miniaudio backend. |
||
| .forgejo | ||
| ci | ||
| deploy | ||
| design | ||
| src | ||
| .gitignore | ||
| build.zig | ||
| build.zig.zon | ||
| LICENSE | ||
| LICENSE-APACHE | ||
| LICENSE-MIT | ||
| README.md | ||
Awebo Chat
Light-weight, self-hostable group chat application with voice rooms and screensharing support.
From small groups of friends and work groups, up to worldwide communities, and without any enshittification.
Awebo server can run on extremely cheap hosting solutions (like Hetzner's 2.99EUR/mo VPS).
Awebo is written in Zig.
Development stage
Awebo is currently pre-alpha. We have working calls (all platforms) and screensharing (sending from macOS, receiving on all platforms).
As of March 2026 Awebo is complete enough to host Awebo development voice calls on Awebo.
A call with 3 screensharing sessions active at the same time (March 2026):
Does Awebo use WebRTC?
No, we are using state of the art codecs (Opus DRED, h264, HEVC, AV1, FFV1) and then write everything above from scratch, including UDP packet framing and jitter buffers. WebRTC is a complicated protocol with bloated C++ implementations that needlessly bump up system requirements and increase resource consumption.
We are currently focusing on native awebo clients but we plan to also create a web client, in which case we will most likely add to Awebo support for WebTransport.
Running the project
Create a new server instance
zig build server -- server init --server-name banana --owner-handle admin --owner-pass admin
This is only needed the first time and whenever the schema changes (make sure to delete awebo.db before re-running the command).
Run the server:
zig build server -- server run
Run the client:
zig build gui
The first time you will be asked to add a remote server, after that the info will be cached in the local config directory. See logs to learn how to reset your local cache if needed.
NixOS
If you get a runtime SDL error that there are no devices, try:
nix-shell -p sdl3
zig build gui -fsys=sdl3
You might also need to do the same with -fsys=pulseaudio.
Testing with multiple users
If you want to launch two or more instances of the client,
each with a different logged user, use the -Dlocal-cache build option like so:
# from inside the awebo repository
mkdir user1
cd user1
zig build gui -Dlocal-cache
Repeat multiple times as needed replacing 'user1' with a different name.
The -Dlocal-cache flag will create a build of the client that stores cache and
authentication data in .awebo-cache and .awebo-config respectively, and by
dedicating a directory to each user you can achieve isolation.
This can also be useful to be able to connect as the same user twice, as you will not be able to do so without this flag (the client takes an exclusive lock to the sqlite database).