adac.HotReload#

class adac.HotReload(fs: float, *, name: str = 'FDN', dsp_path: str | Path = '/tmp/adac-live.dsp', config_path: str | Path = '/tmp/faust-current-dsp.txt', every: int = 1, min_interval: float = 0.5, controls: dict[str, Any] | None = None)[source]#

re-emit a flamo model to faust and publish it to the running plugin.

Parameters:

  • fs (float) – sampling rate in hz, forwarded to flamo_to_faust().

  • name (str) – dsp name, appears in the generated faust header comment.

  • dsp_path (str | Path) – where the generated .dsp is written. the basename must stay stable across the run: the plugin matches reload events against the basename of the file it has loaded.

  • config_path (str | Path) – the plugin’s watched config file. its first line is kept pointing at dsp_path.

  • every (int) – publish at most every n-th call to update(). the first call always qualifies. default 1 (every call).

  • min_interval (float) – minimum seconds between published reloads. protects the plugin from reload storms when the training loop runs faster than the interpreter can swap factories. default 0.5.

  • controls (dict, optional) – macro controls (“rt60”, “dry_wet”, “pre_delay”) forwarded to flamo_to_faust(). slider addresses stay stable across publishes, and the plugin preserves parameter values by address, so knob positions survive every reload.

__init__(fs: float, *, name: str = 'FDN', dsp_path: str | Path = '/tmp/adac-live.dsp', config_path: str | Path = '/tmp/faust-current-dsp.txt', every: int = 1, min_interval: float = 0.5, controls: dict[str, Any] | None = None)[source]#

Methods

__init__(fs, *[, name, dsp_path, ...])

publish(code)

publish a faust source string to the plugin.

update(model, *[, force])

regenerate faust code from the model and publish it if due.
publish(code: str) → bool[source]#

publish a faust source string to the plugin.

lower-level entry point for callers who generate code themselves (for instance via json_to_faust after editing the config). skips the write when the code is identical to the last published version, so no spurious reload fires.

update(model: Any, *, force: bool = False) → bool[source]#

regenerate faust code from the model and publish it if due.

call this after each optimiser step. returns True when a new .dsp was actually written (and the plugin will reload), False when the call was throttled or the generated code is unchanged.

force=True bypasses the step and time throttles, for a final guaranteed publish after the training loop ends.