Local model download

BCCLI v0.4 does not bundle a separate model downloader. That is deliberate: the runtime that will serve the model is the thing that knows which file format, quantisation, and directory it needs, and BCCLI is not that runtime.

The short version

1. Download and place the model with the runtime or provider you already use.
2. Make sure that runtime is running and reachable.
3. Ask BCCLI what it can see:

bccli models

If the model is in that list, it is ready to use with bccli -m provider/model. If it is not, BCCLI cannot reach it, whatever the file on disk says.

Choosing a model that fits

A model that does not fit in memory will load slowly or not at all, and the failure looks like a broken install. Check the size before you download: Local AI covers hardware fit, and System requirements covers the runtime itself.

Where the file goes

The download directory belongs to the runtime, not to BCCLI. Put the model where that runtime looks for models — usually a models directory you configured when you installed it — and BCCLI will find it through the same provider listing as any other local model.

Next

• Local model downloads — the runtime's own download step.
• Use a model in BCCLI — pinning a local model to a run.