Warning: This document is for an old version of rasa NLU. The latest version is 0.15.1.

Configuration

You can provide options to rasa NLU through:

  • a json-formatted config file
  • environment variables
  • command line arguments

Environment variables override options in your config file, and command line args will override any options specified elsewhere. Environment variables are capitalised and prefixed with RASA_, so the option pipeline is specified with the RASA_PIPELINE env var.

Default

Here is the default configuration including all available parameters:

{
  "name": null,
  "pipeline": [],
  "language": "en",
  "num_threads": 1,
  "path": "models",
  "response_log": "logs",
  "config": "config.json",
  "log_level": "INFO",
  "port": 5000,
  "data": null,
  "emulate": null,
  "log_file": null,
  "mitie_file": "data/total_word_feature_extractor.dat",
  "spacy_model_name": null,
  "server_model_dirs": null,
  "token": null,
  "max_number_of_ngrams": 7,
  "duckling_dimensions": null,
  "luis_data_tokenizer": null,
  "entity_crf_BILOU_flag": true,
  "entity_crf_features": [
    ["low", "title", "upper", "pos", "pos2"],
    ["bias", "low", "word3", "word2", "upper", "title", "digit", "pos", "pos2"],
    ["low", "title", "upper", "pos", "pos2"]]
}

Options

A short explanation and examples for each configuration value.

name

Type:str
Examples:"my_model_name"
Description:Defines a models name used to store it and to refere to it when using the http server. The default is null which will lead to a generated model name, e.g. "model_20170426-230305".

pipeline

Type:str or [str]
Examples:"mitie" or ["nlp_spacy", "ner_spacy", "ner_synonyms"]
Description:The pipeline used for training. Can either be a template (passing a string) or a list of components (array). For all available templates, see Processing Pipeline.

language

Type:str
Examples:"en" or "de"
Description:Language the model is trained in. Underlying word vectors will be loaded by using this language

num_threads

Type:int
Examples:4
Description:Number of threads used during training (not supported by all components, though. Some of them might still be single threaded!).

path

Type:str
Examples:"models/"
Description:Directory where trained models will be saved to (training) and loaded from (http server).

response_log

Type:str or null
Examples:"logs/"
Description:Directory where logs will be saved (containing queries and responses). If set to null logging will be disabled.

config

Type:str
Examples:"config_spacy.json"
Description:Location of the configuration file (can only be set as env var or command line option).

log_level

Type:str
Examples:"DEBUG"
Description:Log level used to output messages from the framework internals.

port

Type:int
Examples:5000
Description:Port on which to run the http server.

data

Type:str
Examples:"data/example.json"
Description:Location of the training data.

emulate

Type:str
Examples:"wit", "luis" or "api"
Description:Format to be returned by the http server. If null (default) the rasa NLU internal format will be used. Otherwise, the output will be formatted according to the API specified.

mitie_file

Type:str
Examples:"data/total_word_feature_extractor.dat"
Description:File containing total_word_feature_extractor.dat (see Installation)

spacy_model_name

Type:str
Examples:"en_core_web_sm"
Description:If the spacy model to be used has a name that is different from the language tag ("en", "de", etc.), the model name can be specified using this configuration variable. The name will be passed to spacy.load(name).

server_model_dirs

Type:str
Examples:"models/"
Description:Directory containing the model to be used by server or an object describing multiple models. see HTTP server config

token

Type:str or null
Examples:"asd2aw3r"
Description:if set, all requests to server must have a ?token=<token> query param. see Authorization

max_number_of_ngrams

Type:int
Examples:10
Description:Maximum number of ngrams to use when augmenting feature vectors with character ngrams (intent_featurizer_ngrams component only)

duckling_dimensions

Type:list
Examples:["time", "number", "money", "distance"]
Description:Defines which dimensions, i.e. entity types, the duckling component will extract. A full list of available dimensions can be found in the duckling documentation.

luis_data_tokenizer

Type:str
Examples:"tokenizer_mitie"
Description:Name of the tokenization component used to process luis data (Luis data annotates entities using token offset instead of character offsets, to convert the token offsets to character positions a tokenizer is required.) see Migrating an existing app

If you want to persist your trained models to S3, there are additional configuration options, see Model Persistence

storage

Type:str
Examples:"aws" or "gcs"
Description:Storage type for persistor. See Model Persistence for more details.

bucket_name

Type:str
Examples:"my_models"
Description:Name of the bucket in the cloud to store the models. If the specified bucket name does not exist, rasa will create it. See Model Persistence for more details.

aws_region

Type:str
Examples:"us-east-1"
Description:Name of the aws region to use. This is used only when "storage" is selected as "aws". See Model Persistence for more details.

entity_crf_features

Type:[[str]]
Examples:[["low", "title"], ["bias", "word3"], ["upper", "pos", "pos2"]]
Description:The features are a [before, word, after] array with before, word, after holding keys about which features to use for each word, for example, "title" in array before will have the feature “is the preceding word in title case?”. Available features are: low, title, word3, word2, pos, pos2, bias, upper and digit

entitiy_crf_BILOU_flag

Type:bool
Examples:true
Description:The flag determines whether to use BILOU tagging or not. BILOU tagging is more rigorous however requires more examples per entity. Rule of thumb: use only if more than 100 examples per entity.