Skip to content
Bedrova bedrova
FeaturesCompatibilityDevelopersCompareDocs Coming soon

Docs

Discover & Hugging Face

Search Hugging Face, import by id or URL, read model cards, and download the right variant — safely.

Discover is where you find models to download. It searches Hugging Face directly and reads each repo’s metadata so you can pick the right build before committing disk to it.

Three ways to find a model

  • Search by keyword across Hugging Face.
  • Exact id — paste user/model to jump straight to a repo.
  • Full URL — paste a Hugging Face URL and Bedrova resolves it.

Curated starter and power-user picks are surfaced first so you don’t have to know what to search for on day one.

Reading a result

Each result shows the model card (README), the download variants it offers, size, architecture, license, and a compatibility read for your Mac from the fit-check. Two things to watch for:

  • MLX vs GGUF. Bedrova is MLX-first for text. It detects MLX and GGUF variants, recommends MLX when both can run, and labels GGUF as broad compatibility, usually slower. GGUF is text-only; vision, audio, image generation, perception, and search use their own provider paths.
  • Gated or remote-code repos. A repo that’s gated (requires accepting terms) or that needs trust_remote_code is flagged before you download. Remote code is off by default — see Model trust & verification.

Tokens and mirrors

Add your Hugging Face token (stored in the macOS Keychain, never echoed back) in Settings → Integrations for gated models and higher rate limits. You can also set a mirror base URL as a fallback download source; the canonical resolve URL is kept in job metadata either way.

Dry-run download plan

Before downloading, Bedrova can show a dry-run plan: the exact files, total size, and a free-space check, so there are no surprises. Confirm to start a tracked, resumable download — the mechanics are in Models & downloads.

See also