VlocalhostAI

Install

Install Vlocalhostin about two minutes.

One installer, carrying its own Python — there is nothing to set up first. It records, transcribes and writes your notes entirely on this machine.

 Latest release checking…

one build · three machines
windows-x64sha256 ✓
macos-arm64sha256 ✓
linux-x64sha256 ✓
every release publishes the hash beside the file

Windows

Checking…

Windows 10 / 11 · 64-bit

windows
installing…✓ ready
  1. Download the installer
  2. Run it — Windows may warn, choose More info → Run anyway
  3. Vlocalhost AI appears in your Start menu
  4. Allow the microphone when asked

Unsigned, so SmartScreen warns. Verify against the published .sha256 with Get-FileHash if you want certainty. The ZIP is the same build without an installer: unpack it anywhere, run Vlocalhost.cmd, and delete the folder to remove it.

macOS

Checking…

macOS 12+ · Apple silicon & Intel

macos
installing…✓ ready
  1. Download the .dmg for your chip
  2. Drag Vlocalhost into Applications
  3. First launch is refused — System Settings → Privacy & Security → Open Anyway
  4. Allow the microphone when asked

Pick arm64 for M-series and x64 for Intel — Apple menu → About This Mac. The app is ad-hoc signed rather than notarized, which is why the first launch needs Open Anyway.

Linux

Coming soon

Debian / Ubuntu / Fedora · 64-bit

linux
installing…✓ ready
  1. Download the archive
  2. Unpack it anywhere in your home folder
  3. Run the install script inside it
  4. Allow the microphone when asked

Installs into your home folder; no root required. Needs libportaudio2 for the microphone — the script names the exact command for your distribution.

Every release publishes a .sha256 beside each download and a CycloneDX SBOM per platform. Builds carry signed provenance — verify one with gh attestation verify <file> --repo antigravitysoham-eng/vlocalhost-core.

After it opens

Three things, then it is yours.

1 · Pick what to listen to

microphone
speakers

Settings → What to listen to. Your microphone, the meeting audio from your speakers, or both — labelled separately, so the transcript shows who spoke. No bot joins the call.

2 · Install a summariser

terminal

$ ollama pull llama3.2

pulling manifest → 2.0 GB

✓ ready, on this machine

Transcripts work immediately. For written notes install Ollama and pull a model. That runs locally too.

3 · Press record

Ship Friday, not Wednesday
Mike · staging migration

Words appear while they are still being spoken and firm up when you stop. Stopping writes a Markdown summary and a timestamped transcript into your notes folder.

The long version.

Every screen, every prompt and every thing that can go wrong, written out step by step for each platform. The same document ships inside the download, and the app opens it from Settings → Installation guide.

Troubleshooting.

The five things people actually hit.

Windows says the publisher is unknown
Expected. The installer is not code-signed — a certificate costs money and that was a deliberate decision — so SmartScreen warns on every download. Choose More info → Run anyway. To verify first, every release publishes a .sha256 beside the installer: run Get-FileHash vlocalhost-…-setup.exe and compare the two strings.
macOS says the app cannot be opened
Also expected, for the same reason: the app is ad-hoc signed rather than notarized, because notarization requires a paid Apple developer account. Open System Settings → Privacy & Security, scroll to the message about Vlocalhost and press Open Anyway. You do this once.
Chrome says “Failed — Virus scan failed”
Not a detection. Chrome hands the finished download to Windows to be scanned and that call failed — usually third-party antivirus, a managed work laptop, or a Defender service that is not running. The file is never reported as malicious; the scan simply never answered. Open Downloads (Ctrl+J) and press Retry.
The microphone is not picked up
Open Settings → What to listen to, choose the microphone by name, and press Refresh if you plugged it in after the app opened. From a terminal, python vlocalhost.py --devices lists everything the app can see and what your current setting resolves to.
There are no summaries, only transcripts
Summarising uses Ollama, running locally. Install it and run ollama pull llama3.2. The transcript is always written regardless — a missing summariser never costs you the meeting.

Still stuck? Send a diagnostic report — the app writes one you can read before you send it.