Releases: flujo-app/CommunityAI
Release list
CommunityAI alpha — community inference without model downloads
Community inference now runs entirely on contributing peers. Your app sends messages and receives answers without downloading community model weights or running its input/output stages locally.
- The small local Qwen fallback remains available when the community cannot answer. The explicit local-only choice is preserved.
- A 100% sharing limit now makes the full GPU memory capacity available for sharing. No memory is permanently reserved for an unused fallback.
- Model health uses the shared discovery status, removing the separate local tensor router that could show zero available blocks while the public mesh was complete.
- The desktop accepts the new zero-byte community download status.
Existing September 9 updater installations download this release automatically and offer Restart to update. Click the update status in the sidebar to check immediately. Earlier installations need one manual installer upgrade. Settings and model caches are preserved.
| Platform | Small download | Complete offline installer |
|---|---|---|
| Windows | Online setup | Offline setup |
| Ubuntu/Debian | Online installer | Offline .deb |
The online installers fetch and verify the complete runtime. Windows publisher signing remains deferred for alpha. Checksums.
Two real public-mesh requests completed with an empty consumer cache and all artifact downloads forbidden. The temporary CPU mesh took about 107 seconds for a short completion and 158 seconds for a short chat; this is functional evidence, not a latency target. Text-serving peers currently require operator setup through the source CLI (drift text-peer); automatic placement of that role on desktop contributors remains future work.
Fresh Windows and Linux packages use source d7f4333cc8ff74ce060076c1d6effb67fb4c6d2c and the normal packaging checks. Large hosted installers are checked through object size, SHA-256 metadata and public start/end samples; small public artifacts receive full hash checks. This release does not claim a new full installed desktop or GPU test campaign.
CommunityAI alpha — simpler desktop and automatic updates
CommunityAI now downloads application updates automatically and shows Restart to update when ready.
- Home: the model in use and why, your CPU/GPU, GPU memory in GB, computing limit, and Start/Pause sharing.
- Models: expandable sections for block health, downloads and contributors. Retired managed model entries are repaired on upgrade.
- Sharing: immediate starting/stopping feedback and clear errors.
- Updates: visible download progress, resumable downloads, signed update information and verified installer files. You choose when to restart.
Already using the September 8 release? Install this update once using a download below. That release did not contain an updater. Later releases can update through the app. Settings and downloaded model files are preserved; no uninstall is needed.
| Platform | Small online installer | Complete offline installer |
|---|---|---|
| Windows | Online setup | Offline setup |
| Ubuntu/Debian | Online installer | Offline .deb |
The small installer downloads the full runtime before installing. On Linux, run the online installer with python3, or install the downloaded package with sudo apt install './communityai_0.1.0~alpha.20260909.2_amd64.deb'. Linux updates ask for administrator authentication.
Installer checksums · Release metadata
This is an alpha for closed testing. Windows publisher signing remains deferred, so Windows may show an unknown-publisher prompt. Model files download separately; community availability depends on contributors.
Validation: existing source checks and the Windows update-handoff fixture passed; packages include the normal build checks and uploaded size/checksum metadata and public download samples. The updater verifies complete installer downloads before running them. This release does not claim a new full desktop, GPU, cloud or installed Linux updater test run.
CommunityAI 0.1.0 alpha — Windows and Linux
CommunityAI alpha
CommunityAI provides local Qwen inference and optional community inference through
the desktop and a local OpenAI-compatible API. Separate VRAM and processing
sliders default to 100%; sharing starts only when enabled. Processing limits pace
contributed work and allow brief bursts.
The block-health grid shows observed coverage, reservations and failures.
Your own model downloads and worker downloads have separate progress displays,
with available peer details.
| Platform | Small online installer | Complete offline installer |
|---|---|---|
| Windows x64 | Online setup · 2.11 MB | Offline setup · 2.46 GB |
| Debian/Ubuntu amd64 | Online installer · 13.7 KB | Debian package · 2.30 GB |
This alpha is unsigned. Windows may show an unknown publisher. Compare the
download against the published checksums
before running it. Signing, Store distribution, automatic application updates and
a hosted APT repository follow after alpha.
Online installers verify the complete offline package's size and SHA-256. They
download the same runtime. Allow approximately 4.26 GB installed on Windows
or 5.16 GB on Linux, plus installer, filesystem/upgrade overhead and model
cache. Linux online installation additionally needs 4.60 GB of temporary
package copies. Model weights download separately; RAM/VRAM needs depend on
the model. GPU use requires a compatible NVIDIA driver.
Run Windows setup as your ordinary user. On Debian 12+/Ubuntu 22.04+ amd64, run
python3 communityai-0.1.0-alpha.20260908.1-linux-online.py; it requests sudo
after verification and checks a protected copy before APT installation.
Windows's complete hosted download, installation, CPU diagnostic and removal
passed. Linux's hosted download/APT/removal passed on Ubuntu 22.04; its
scoped audit
preserves the raw observer failure and references prior CPU/CUDA proof for the
identical package.
Upgrade and uninstall preserve external settings, native credentials and model
caches. Disable “Start CommunityAI when I sign in” before uninstalling;
intentional retained-data removal remains a separate choice.
Both runtime bundles identify source 84205f93; the later Windows downloader
identifies b6c8aad9. Artifact provenance records distinguish them.
Inference is best effort: local use depends on downloaded models and available
resources; community use depends on online contributors. Current complete public
route capacity is not guaranteed. The combined Gate 16 public canary has not
run. Existing bounded conversation/recovery evidence remains separate from
these installer checks.