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

Using rasa NLU as a HTTP server


Before you can use the server, you need to train a model! See Training Your Model

The HTTP api exists to make it easy for non-python projects to use rasa NLU, and to make it trivial for projects currently using {wit,LUIS,api}.ai to try it out.

Running the server

You can run a simple http server that handles requests using your models with (single threaded)

$ python -m rasa_nlu.server -c config_spacy.json --server_model_dirs=./model_YYYYMMDD-HHMMSS

If your server needs to handle more than one request at a time, you can use any WSGI server to run the rasa NLU server. Using gunicorn this looks like this:

$ gunicorn -w 4 --threads 12 -k gevent -b rasa_nlu.wsgi

This will start a server with four processes and 12 threads. Since there is no standard way to pass command line arguments to the server, all your configuration options need to be placed in your configuration file (including the server_model_dirs!). You can set the location of the configuration file using environment variables, otherwise the default configuration from config.json will be loaded.


rasa NLU can ‘emulate’ any of these three services by making the /parse endpoint compatible with your existing code. To activate this, either add 'emulate' : 'luis' to your config file or run the server with -e luis. For example, if you would normally send your text to be parsed to LUIS, you would make a GET request to


in luis emulation mode you can call rasa by just sending this request to


any extra query params are ignored by rasa, so you can safely send them along.


POST /parse (no emulation)

You must POST data in this format '{"q":"<your text to parse>"}', you can do this with

$ curl -XPOST localhost:5000/parse -d '{"q":"hello there"}'

POST /train

You can post your training data to this endpoint to train a new model. this starts a separate process which you can monitor with the /status endpoint. If you want to name your model to be able to use it during parse requests later on, you should pass the name /train?name=my_model. Any parameter passed with the query string will be treated as a configuration parameter of the model, hence you can change all the configuration values listed in the configuration section by passing in their name and the adjusted value.

$ curl -XPOST localhost:5000/train -d @data/examples/rasa/demo-rasa.json

GET /status

This checks if there is currently a training process running (you can only run one at a time). also returns a list of available models the server can use to fulfill /parse requests.

$ curl localhost:5000/status | python -mjson.tool
  "training" : False
  "models" : []

GET /version

This will return the current version of the Rasa NLU instance.

$ curl localhost:5000/version | python -mjson.tool
  "version" : "0.8.2"

GET /config

This will return the currently running configuration of the Rasa NLU instance.

$ curl localhost:5000/config | python -mjson.tool
    "config": "/app/rasa_shared/config_mitie.json",
    "data": "/app/rasa_nlu/data/examples/rasa/demo-rasa.json",
    "duckling_dimensions": null,
    "emulate": null,


To protect your server, you can specify a token in your rasa NLU configuration, e.g. by adding "token" : "12345" to your config file, or by setting the RASA_TOKEN environment variable. If set, this token must be passed as a query parameter in all requests, e.g. :

$ curl localhost:5000/status?token=12345

Serving Multiple Apps

Depending on your choice of backend, rasa NLU can use quite a lot of memory. So if you are serving multiple models in production, you want to serve these from the same process & avoid duplicating the memory load.


Although this saves the backend from loading the same backend twice, it still needs to load one set of word vectors (which make up most of the memory consumption) per language and backend.

You can use the multi-tenancy mode by replacing the server_model_dirs config variable with a json object describing the different models.

For example, if you have a restaurant bot and a hotel bot, your configuration might look like this:

  "server_model_dirs": {
    "hotels" : "./model_XXXXXXX",
    "restaurants" : "./model_YYYYYYY"

You then pass an extra model parameter in your calls to /parse to specify which one to use:

$ curl 'localhost:5000/parse?q=hello&model=hotels'


$ curl -XPOST localhost:5000/parse -d '{"q":"I am looking for Chinese food", "model": "restaurants"}'

If one of the models is named default, it will be used to serve requests missing a model parameter. If no model is named default requests without a model parameter will be rejected.