backchannel
Release v0.4.0

Updates itself, and knows what your machine can run.

Backchannel now updates in place with a verified signature and a rollback path, measures whether your hardware can keep up before a call starts, and makes Privacy First a mode a whole workspace can actually run in -- agents, Analyze, Enhance Insights, and the briefing all keep working on a model you host yourself.

Download Backchannel v0.4.0

These unsigned open-source builds are delivered through the authenticated Backchannel portal. An approved Backchannel account is required; no GitHub identity or repository membership is required.

The Windows and Linux bundles include ffmpeg. On macOS, voice calibration and compressed audio imports still need ffmpeg on PATH.

Windows

Windows x64

For Windows 10 or later on 64-bit Intel or AMD processors.

  1. Download Backchannel-windows-x64-v0.4.0.zip.
  2. Unzip it and run Backchannel.exe.
  3. If SmartScreen appears, choose More info, then Run anyway.
Open v0.4.0 downloads
macOS

macOS Apple Silicon

For Apple Silicon Macs. Intel macOS builds are not included in this release.

  1. Download Backchannel-macos-arm64-v0.4.0.zip.
  2. Unzip it and move Backchannel.app wherever you keep apps.
  3. Right-click Backchannel.app and choose Open the first time.
Open v0.4.0 downloads
Linux

Linux x64

A portable bundle for 64-bit Intel or AMD Linux systems.

  1. Download Backchannel-linux-x64-v0.4.0.tar.gz.
  2. Extract it with tar -xzf Backchannel-linux-x64-v0.4.0.tar.gz.
  3. Run Backchannel/Backchannel.
Open v0.4.0 downloads

What changed

v0.4.0 follows v0.3.8 with in-place updates, capacity you can measure, and a Privacy First mode that carries the whole workspace.

Updates in place, verified

The desktop app checks for a new version, verifies its Ed25519 signature before touching anything on disk, and applies it with a rollback path if the swap fails. Shutdown is gated, so an active recording or a running post-processing pass finishes before the restart.

Privacy First carries the workspace

Admission is judged by where a model is served rather than what its vendor is called, so a model on your own machine or LAN keeps the analysis agents, post-import Analyze, Enhance Insights, and the call briefing running with the switch on. A refusal names the agent and the model to change instead of only saying no.

Know what the machine can run

The Local Model Fit Test times each self-hosted model against the real agent prompts and a realistic transcript, scoring it per role with per-model cycle budgets and an adjustable contention slider. Call-start capacity admission sums diarization, transcription, captions, and agents against measured headroom, and a benchmark from different hardware is marked out of date rather than shown as current.

Captions without the cloud

Point the audio bridge at the on-device captioner and interim captions run with no cloud call, including under Privacy First. It is CPU-heavy by nature, so the fit test projects whether the machine can sustain it alongside diarization and transcription.

Long calls survive slow models

A diarizer that falls behind sheds audio instead of exhausting memory, self-hosted requests get a budget and timeout sized for local hardware, agent replies are schema-validated so a local model's output is not silently dropped, and insight revalidation falls back to a self-hosted model when a cloud quota runs out mid-run.

Ending a call tells the truth

Stopping the browser share finalizes the system track at that moment rather than surfacing a stray speaker minutes later. End Call names any analysis stage that degraded, saves that summary with the call so a dropped connection cannot lose it, and a long post-processing run is no longer reported as a lost connection.

First-run notes

NeedWhat to do
Release access Use the email and password for your approved Backchannel recipient account.
Provider key Follow the guided setup on first launch, or add a key later in Admin -> API Keys. Calls will not start until the transcription model has a usable credential.
Voice calibration Works out of the box on Windows and Linux. On macOS, install ffmpeg and make sure it is on PATH before enrolling your voice.
Upgrading Quit the old version from the tray before starting the new one. Session data is stored per-user and shared by every version. A previously configured OpenAI-compatible endpoint is migrated automatically on first launch. The Admin About tab shows a what's-new notice after upgrading.
Desktop launch Backchannel opens a dedicated local app window in your default browser when it supports app windows, otherwise it uses a standard browser tab. It remains available from the system tray or menu bar.
Docker Check out the v0.4.0 source tag and rebuild Compose to run the same release through containers.

Start with the desktop app or self-host the stack

Both paths run Backchannel on your machine. Choose the desktop bundle for the fastest start, or Compose for maximum control.