Expand description
Client side of the newline-delimited JSON protocol spoken by the Python
ML workers (chatterbox_turbo_worker.py, whisper_worker.py,
mmaudio_worker.py). The workers stay Python, in their own venvs; Rust
only starts them, writes one JSON request per line and reads one JSON
response per line.
Startup differs by caller, so both reads are exposed: stem-verify
parses the very first line as the ready message, while the SFX and TTS
clients skip model-loading noise until a {"ready": true} line appears.
Structs§
- Worker
- A running worker process.
Functions§
- is_
truthy - Python truthiness for the
readyfield. - worker_
script <package_dir>/<name>— where a worker script lives next to the Python package that ships it.