Module local_llm 

Location and accessors for the bundled local-model base URL.

Hub starts an MLX inference server on an ephemeral port and writes its OpenAI-compatible base URL here. Agents — started by launchd and therefore NOT inheriting Hub’s environment — read it back from this file.

Functions

base_url_path

Path to the file holding the local model base URL, under <mur_home>.

read_base_url

Read the base URL, trimming whitespace. None if absent/empty.

write_base_url

Atomically write the base URL (temp file + rename).