Sharing my MLX-LM CLI: "mlxlm" model search, sessions, aliases, and more

[CreamyCappuccino]

Hey everyone!
I’ve been working on a small CLI tool for MLX-LM and thought I’d share it here.
The goal is to provide a lightweight, script-friendly CLI inspired by Ollama’s workflow, but focused specifically on MLX-LM use cases rather than general model serving.
It also exposes JSON output and non-interactive modes, so AI agents or scripts can use it as a small local “tool backend” if needed.

🔧 Key features

• HuggingFace model search (with filters, sorting, pagination)
• JSON output mode (for automation / AI agents)
• Session management (resume previous chats, autosave, /new)
• Interactive alias system for long model names
• Prompt-toolkit UI (history, multiline, autocompletion)
• Multiple chat renderers (Harmony / HF / plain text)
• Offline mode, custom stop sequences, custom renderers, etc.

💡 Why a CLI?

Sometimes a terminal-first workflow is faster for:

• automation & scripting
• integrating into personal tools
• quick experiments without a full UI
• running on remote machines or lightweight environments

📎 Repository

https://github.com/CreamyCappuccino/mlxlm

Still evolving, but if anyone finds this useful or has ideas/feedback, I’d love to hear it!
I'll leave some screenshots down below.