# VoiceBM

"Voice BM is not a replacement for anything. It is an enhancement to fill a hole that exists within the local smart AI consumer communities. I built it and shared it so you didn't have to."

VoiceBM is a voice biometrics engine for Home Assistant. It identifies who is speaking, publishes identity signals via MQTT, and supports enrollment + blocklisting workflows.

> Scope note: VoiceBM is an engine. It exposes identity and controls; your UI/LLM/frontend decides how to use them.

## What it does
- Creates speaker embeddings and matches against an enrolled gallery
- Falls back to a virtual **user** identity for unknown speakers
- Publishes identity + status topics over MQTT (Home Assistant discovery-friendly)
- Per-identity blocklist switches (including **user**) to fail STT silently when blocked
- Room-aware collection, roomless identity (identity is global)

## What it does NOT do
- It’s not a full conversational UX
- It does not decide “what to say” or how to build assistants
- It doesn’t assume a specific frontend (Home Assistant, OpenWebUI, local LLM, etc.)

## Quick start (high level)
1. Install dependencies (see docs)
2. Configure MQTT + paths
3. Start services
4. Enroll voices
5. Verify identities in Home Assistant

## Configuration
This repo ships with example configs/templates. You must provide your own:
- MQTT broker host/user/password
- Local IPs / URLs
- Any secrets

**Never commit secrets** to GitHub.

## Docs
- See `/docs` for setup and Home Assistant entity details (if present)
- See `/systemd` for service examples (if present)

## Project site
https://cybericebyte.github.io/VoiceBM/

## Companion AI artifacts
- ChatGPT GPT: https://chatgpt.com/g/g-68e55ecea6888191bb871536408fa73b-home-assistant-voicebm-voice-biometrics
- Claude Artifact (embedded): https://cybericebyte.github.io/VoiceBM/claude.html
- Claude Artifact (source): https://claude.ai/public/artifacts/918fa86a-7b8f-4ac5-a4a7-28c581d29cdd