llamastash
Share on:
Enter Your Mastodon Instance
Copy the link below to share to Mastodon
https://terminaltrove.com/llamastash/
A zero-overhead, terminal-native local-LLM manager.
Script Preview
Review this script before executing in your terminal.
llamastash is an all-in-one manager for launching and supervising local language models directly in your terminal.
It discovers local open weight model files in user directories and has support for existing Hugging Face, Ollama and LM Studio caches. Its TUI interface displays model information as well as the host info and starts selected models through llama.cpp or custom configured inference backends. A background daemon supervises running models without running the main interface.
llamastash comes with a built-in OpenAI-compatible proxy and presents local models through one endpoint. CLI commands return structured JSON for scripts and coding agents and you can configure an inference backend that is best for your host hardware.
This tool is very useful to machine-learning engineers, researchers, homelab enthusiasts and software developers who want to explore using open weight models to connect coding assistants and agents for local LLM inference for testing or benchmarking use-cases.
