# `LLMDB.Sources.Meta`
[🔗](https://github.com/agentjido/llmdb/blob/main/lib/llm_db/sources/meta.ex#L1)

Cached source for the Meta Model API registry.

`pull/1` requests `GET https://api.meta.ai/v1/models` with a bearer token.
Set `META_API_KEY` or pass `:api_key`. Tests can pass `:req_opts` and
`:cache_dir`. `:url` overrides the registry URL.

`load/1` reads the cache without a network request or an API key. The default
cache directory is `priv/llm_db/remote`; configure `:meta_cache_dir` to change it.

The registry is specific to a team. A pull keeps previously discovered models
when the response omits them. It preserves registry IDs, creation timestamps,
and owners. It does not infer prices, limits, capabilities, or release dates.
New registry entries stay catalog-only until a reviewed overlay supplies an
execution contract. Muse Code and Muse Wonder Full are excluded.

Missing credentials return `{:error, :no_api_key}`. Failed requests return a
safe error and preserve the cache. HTTP errors retain their status code;
transport errors and invalid responses do not include response data or keys.

# `transform`

Converts registry records to sparse catalog entries under provider `:meta`.

---

*Consult [api-reference.md](api-reference.md) for complete listing*
