Apppublic

whoami-j/Ornith-MLX-Chat

sourceHugging Faceupdated 1mo agoView on Hugging Face

0likes

App README

🐦 Ornith-1.5-35B-A3B — MLX 4-bit chat

Gradio chat frontend for `ornith-ai/Ornith-1.5-35B-A3B-MLX-4bit`, served by a local MLX LM server.

How it works

—Run the model on any Apple Silicon Mac: mlx_lm.server --model ornith-ai/Ornith-1.5-35B-A3B-MLX-4bit (bind 0.0.0.0, port 8000).
—Expose it publicly with a tunnel (cloudflared tunnel --url localhost:8000, ngrok, etc.).
—Point this Space at the tunnel: hf spaces secrets set whoami-j/Ornith-MLX-Chat MLX_BASE_URL=https://<tunnel-host>/v1.
—Chat here. Zero GPU cost on Hugging Face — all inference happens on your Mac.

Config (env)

Var	Default	Purpose
`MLX_BASE_URL`	`http://localhost:8000/v1`	OpenAI-compatible endpoint (local or tunneled)
`MODEL_NAME`	`ornith-ai/Ornith-1.5-35B-A3B-MLX-4bit`	Model id passed to the server
`API_KEY`	`none`	Auth header sent to local server
`TEMPERATURE`	`0.6`	Sampling temperature