backchannel
Windows x64
For Windows 10 or later on 64-bit Intel or AMD processors.
- Download
Backchannel-windows-x64-v0.5.0.zip. - Unzip it and run
Backchannel.exe. - If SmartScreen appears, choose More info, then Run anyway.
Release v0.5.0
Asks which provider you use, and stops losing things.
First-run setup no longer seeds a cloud model behind your back: every agent starts unselected, models are grouped by provider, and a recommendation is marked in whichever providers you actually have. Strategic signals persist live and after the call, the briefing names speakers instead of identifiers, and every build path now installs the exact dependency tree that was tested.
Download Backchannel v0.5.0
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.
macOS Apple Silicon
For Apple Silicon Macs. Intel macOS builds are not included in this release.
- Download
Backchannel-macos-arm64-v0.5.0.zip. - Unzip it and move
Backchannel.appwherever you keep apps. - Right-click
Backchannel.appand choose Open the first time.
Linux x64
A portable bundle for 64-bit Intel or AMD Linux systems.
- Download
Backchannel-linux-x64-v0.5.0.tar.gz. - Extract it with
tar -xzf Backchannel-linux-x64-v0.5.0.tar.gz. - Run
Backchannel/Backchannel.
What changed
v0.5.0 follows v0.4.0 with explicit provider-aware setup, a call record that keeps what it surfaces, and build paths that agree with each other.
First-run setup asks instead of assuming
A fresh install no longer seeds a cloud model behind your back. Every enabled agent starts unselected and says so, models are grouped by Google, OpenAI, and local, and one role-appropriate recommendation is marked in each provider you actually have available. Connecting only an OpenAI key no longer reports that Google is required, and a keyless local setup can find the paths that work with no cloud account at all.
Strategic signals persist
Signals raised during a call are kept and stay readable both live and after the call, instead of scrolling away. Briefing context built from them is deduplicated, so the same observation stops arriving several times wearing different words.
Documents feed the live conversation
Uploaded document summaries are stored once and reused, so asking about a document mid-call draws on the summary rather than re-reading the file. Under Privacy First the excerpt path keeps that working without sending the document anywhere.
The briefing names people, not identifiers
Owners and attribution resolve to speaker names at read time, including in exported HTML, so a finished call reads as a record of a conversation rather than a table of UUIDs. The raw owner chip is gone and the rationale sits directly under the summary it explains.
Insights say which model produced them
Enhance Insights shows the model you asked for alongside the one that actually ran, so a fallback is visible rather than silent. Insight type labels are legible words on badges and section headers, and the Excel export is a single enriched file instead of a choice between two partial ones.
Ask questions without leaving the call
A new in-call ask bar answers from the conversation as it happens, so you can check what was said earlier without stopping the recording or waiting for the post-call briefing.
The live insight surface was redesigned
Insights arriving during a call are laid out to be read at a glance while you are still talking, so the list can be scanned mid-conversation rather than studied afterwards.
Steadier in the places that used to slip
Activity chips separate blocked work from failed work and report a current failure count rather than a running total. Batch readiness stops calling a blank or unrecognized model id ready, oversized briefing context items are retained rather than dropped, and the live ask bar no longer starves the transcript it depends on.
Built from what was tested
Sixteen known vulnerabilities were cleared from the web-facing dependency stack. The four ways this project builds its frontend now agree, so continuous integration, the Windows, macOS, and Linux bundles, and the Docker image all install the exact dependency tree recorded in the repository. Windows executables now carry proper product and version metadata.
First-run notes
| Need | What 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 -> Connections. Calls will not start until the transcription model has a usable credential; local ONNX transcription needs none. |
| 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 | From this release the app updates itself in place: it verifies the new build's signature first and waits for recording and post-processing to finish before restarting. If you install a build manually instead, quit the old version from the tray first. 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 v0.5.0 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.