Models

BotConnector uses Hugging Face as its live catalog. There is no fixed, hand-maintained model list — discovery queries the Hub, and BotConnector adds GGUF file inspection, capability labels and hardware-fit guidance.

In this section

• Discover models — search and filter the live catalog.
• Hugging Face source — how Hub metadata is used and cached.
• Quantization — pick the right file size.
• Capabilities — what the badges mean and where they come from.
• Installed models — manage what is on disk.
• Gated models — license-gated repositories.
• Model storage — where files live.

The model lifecycle

1. Discover — search GGUF repositories on the Hub.
2. Inspect — view real file list: quantizations, sizes, sharded layout, projector files.
3. Download — fetch exactly the chosen file(s).
4. Load — run through the managed runtime.
5. Use — chat, API, embeddings, or integrations.
6. Remove — delete from the model store when no longer needed.

Capability labels

Every catalog entry carries badges such as Chat, Tools, Vision, Coding, Reasoning, Embeddings or Audio. These are evidence labels inferred from Hub tags and metadata, not guarantees — a Vision model still needs a projector file and runtime support. Full explanation: Capabilities.

Open the catalog

The public catalog is at Model catalog. In BCCLI, list models exposed by your configured providers with:

bccli models