Local model management

BCCLI v0.4 does not bundle a separate model-download command, and that is a boundary rather than a gap. Downloading is the runtime's job: it owns the directory, the file format, and the quantisation it can serve.

The workflow

1. Download or import the model with the runtime you use — Ollama, LM Studio, or any OpenAI-compatible server.
2. Make sure that runtime is running on this machine.
3. Register it once:

bccli provider list
bccli provider add <id>

1. Confirm the model arrived:

bccli models

If it is in that list, it is usable with bccli -m provider/model.

Managing providers

bccli provider remove <id>

Removes the provider from BCCLI. Removing it does not delete model files and does not touch other providers' credentials — it stops BCCLI from offering models it can no longer reach.

Why there is no bccli get

A download command in the CLI would have to guess which format and directory your runtime expects, and would be wrong for anyone using a different one. The runtime you already have is the correct place to make that decision.

Next

• Local AI provider setup — the same steps with more context.
• Use Local AI in BCCLI — running against the model you just added.