# Vaultwares-Realtime

Real-Time Speech-to-Text Engine
Part of the VaultWares Ecosystem • docs.vaultwares.com • vaultwares.com

High-performance, local-first VaultWares Realtime service (faster-whisper / Whisper-based) with low-latency streaming, multi-language support, and seamless integration into vaultwares-pipelines and vault-flows.

Overview

This component provides real-time transcription and translation capabilities for video, audio streams, live meetings, and media processing pipelines across the VaultWares ecosystem.

Features

• Real-time streaming transcription
• Multi-language support + auto-translation
• Local-first execution (GPU/CPU optimized)
• Fault-Tolerant Hardware Management (Automatic CPU fallback for models)
• WebSocket / HTTP streaming endpoints
• Integration with vaultwares-pipelines for end-to-end media workflows
• Low-latency mode for vault-player and vault-flows
• Multi-segment handling optimization (Near real-time algorithms, VAD slicing)
• Audio Volume Normalization (Dynamic peak AGC)
• Agent-aware monitoring hooks
• Persistent JSON configuration (config.json)
• Glass UI 9-theme system
• WASAPI loopback support out-of-the-box (via soundcard)

Quick Start

git clone https://github.com/p-potvin/Vaultwares-Realtime.git
cd Vaultwares-Realtime
git submodule update --init --recursive
pip install -r requirements.txt
python run_server.py

Architecture & Agent Integration Fully synchronized with the VaultWares Agent Knowledge Dissemination System: → https://raw.githubusercontent.com/p-potvin/vaultwares-docs/main/agents/knowledge-dissemination.mdx Can invoke the full coordinated agent team via invoke_vaultwares_team skill from vaultwares-adk. Privacy & Security

All processing happens locally by default No cloud transcription unless explicitly configured Encrypted WebSocket option Full threat model in central VaultWares docs

Contributing See CONTRIBUTING.md and the central Brand Guidelines. License GPL-3.0 (see LICENSE) Built with ❤️ for privacy