halo-ai
Local AI anywhere, for everyone. No cloud. No subscriptions. Just you and your hardware.
The Stack
A complete AI platform running on bare metal. LLMs, voice cloning, image generation, autonomous agents — all local, all yours.
Local LLM
86 tok/s on Qwen3-30B with Vulkan + Flash Attention. No API keys. No rate limits.
Voice & TTS
Clone any voice with 30 seconds of audio. Whisper STT. Kokoro TTS. All on-device.
Image Generation
ComfyUI with SDXL and Flux. 123.5GB unified VRAM on AMD Strix Halo.
Autonomous Agents
17 specialized AI agents with personalities, backstories, and real jobs.
AMD Optimized
Built for Strix Halo. ROCm + Vulkan. 128GB unified memory. No NVIDIA tax.
Zero Cloud
Everything runs on your hardware. Your data never leaves your network.
Meet the Family
Every agent has a role, a personality, and a purpose.
Echo
Bounty
Meek
Amp
Mechanic
Muse
Sentinel
Forge
Qwen3-30B-A3B · Vulkan + Flash Attention · AMD Strix Halo · Bare metal
Get Started
One command. Under 5 minutes. Your AI, your rules.
git clone https://github.com/stampby/halo-ai.git && cd halo-ai && ./install.sh