This is unreleased documentation for Rasa Documentation Main/Unreleased version.
For the latest released documentation, see the latest version (3.x).

Version: Main/Unreleased

InvalidRecipeException Objects

class InvalidRecipeException(RasaException)

Exception in case the specified recipe is invalid.

Recipe Objects

class Recipe(abc.ABC)

Base class for Recipes which convert configs to graph schemas.


def recipe_for_name(name: Optional[Text]) -> Recipe

Returns Recipe based on an optional recipe identifier.


  • name - The identifier which is used to select a certain Recipe. If None the default recipe will be used.


A recipe which can be used to convert a given config to train and predict graph schemas.


def auto_configure(
config_file_path: Optional[Text],
config: Dict,
training_type: Optional[TrainingType] = TrainingType.BOTH
) -> Tuple[Dict[Text, Any], Set[str], Set[str]]

Adds missing options with defaults and dumps the configuration.

Override in child classes if this functionality is needed, each recipe will have different auto configuration values.


def graph_config_for_recipe(
config: Dict,
cli_parameters: Dict[Text, Any],
training_type: TrainingType = TrainingType.BOTH,
is_finetuning: bool = False) -> GraphModelConfiguration

Converts a config to a graph compatible model configuration.


  • config - The config which the Recipe is supposed to convert.
  • cli_parameters - Potential CLI params which should be interpolated into the components configs.
  • training_type - The current training type. Can be used to omit / add certain parts of the graphs.
  • is_finetuning - If True then the components should load themselves from trained version of themselves instead of using create to start from scratch.


The model configuration which enables to run the model as a graph for training and prediction.