Skip to main content

`GET /api/v1/llm/recommend?limit=...&n_ctx=...`

GET 
/api/v1/llm/recommend

Ranks the models this machine should run: plans against the embedded generation table, resolves the survivors on HuggingFace, reads each finalist's GGUF header over a range request, and orders what is left.

Answers 503 when the Hub cannot be reached, which is deliberately not the same as an empty 200. The catalogue is fetched live, so no network means no catalogue; reporting that as "no models fit" would be a different and false claim about the operator's machine.

Request​

Responses​

Models ranked for this machine, best first