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

Public LLM API catalog source (https://api.llmapi.ai/v1/models).

`pull/1` explicitly fetches the public catalog without credentials. `load/1`
only reads its local cache. Application filters are applied by the engine,
never by this source.

The public catalog mixes Chat, Responses-only, and media models. Every exact
ID is retained, but Chat execution is declared only when all advertised
fallback routes support it. Capabilities are intersected across those routes.
Object generation uses ReqLLM's shared tool-based object adapter, not native
JSON Schema output. These contracts describe published metadata; live API
compatibility remains to be verified.

Top-level chat prices are USD per token, unlike upstream route prices. The
public Models page multiplies top-level rates by 1,000,000 for its display.
No output-token limits or strict/streamed-tool guarantees are published.

Sources:
- https://docs.llmapi.ai/api/v1/models
- https://docs.llmapi.ai/api/v1/chat/completions
- https://docs.llmapi.ai/models

Options: `:url`, `:cache_dir`, `:req_opts`, and an optional `:request` test hook.
The default cache directory is configured by `:llmapi_cache_dir` and falls
back to `priv/llm_db/remote`.

# `transform`

Transforms the public catalog into canonical data without filtering model IDs.

---

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