Local model downloads

BCCLI v0.4 does not bundle a separate model downloader. Downloading is the runtime's job: it knows which format, quantisation, and directory it needs, and BCCLI only needs to know what the runtime ended up serving.

Download with the runtime

Follow the runtime's own instructions — a models directory you configure once, then a pull or import per model. The file lands where that runtime looks for it, not where BCCLI looks, because BCCLI does not look at files at all.

Confirm BCCLI sees it

bccli models

This is the only check that matters. If the model is in that list, it is ready:

bccli -m provider/model

If it is not, the file is not the problem to investigate — the provider is. Check bccli provider list, and that the runtime is running on the address BCCLI expects.

Before you download

Check that the model fits in memory on this machine. A model that does not fit loads slowly or fails, and the failure reads like a broken install rather than a size problem. Local AI covers hardware fit; System requirements covers the runtime itself.

Next

• Local AI provider setup — registering the runtime.
• Use Local AI in BCCLI — running against a local model.