lilbee / model manager
Run and manage local AI models, in one program. lilbee browses Hugging Face, downloads a model, gives it a role, and runs it on your own GPU. It is a complete model manager, not a client of one.
You do not need a separate model runner. lilbee places large models across multiple GPUs sized to each card, and keeps models warm or swaps them on demand.
a complete model manager
Most local-AI tools make you stand up a model server and feed it files by hand. lilbee is the manager itself: browse Hugging Face from a built-in catalog, pull a model, assign it a role, and run it. One program covers the whole path from finding a model to answering with it.
what it does
browse and pull from Hugging Face
A built-in catalog searches Hugging Face Hub and downloads a model straight into your library, with no hunting for the right file.
give each model a role
Assign models to roles: chat, embedding, reranking, and vision. lilbee wires them into the search pipeline for you.
runs on your GPU
Metal, Vulkan, or CUDA. lilbee runs the model itself; you never point it at a server you set up first.
VRAM-aware multi-GPU fleet
Large models are split across your GPUs sized to each card, and kept warm or swapped on demand. See will it fit your GPU.
already have a runner? keep it
lilbee is a full model manager on its own, so you do not need anything else. If you already use Ollama or LM Studio, point lilbee at your setup and it will use those models instead. A cloud model is there when you want one, and never used unless you pick it.
questions
Is lilbee a model manager?
Yes, a complete one. lilbee browses Hugging Face, downloads models, gives each one a role, and runs them on your GPU. It also places large models across multiple GPUs. It is the model manager, not a client of one.
Do I need Ollama or LM Studio?
No. lilbee runs models itself, so you do not need a separate model runner. If you already use Ollama or LM Studio, you can point lilbee at them instead and keep them.
Where do the models come from?
From Hugging Face Hub. lilbee has a built-in catalog you browse, then pull a model straight into your library without hunting for files.
What hardware does it run on?
lilbee runs models on Metal, Vulkan, and CUDA, and spreads a large model across multiple GPUs sized to each device's VRAM.
Can it run several models at once?
Yes. You assign models to roles such as chat, embedding, reranking, and vision, and lilbee keeps them warm or swaps them on demand across your GPUs.