But I don’t code in python!¶
We’ve made every effort to make sure that you can still use Rasa Core even if you don’t want to use python. However, do consider that Rasa Core is a framework, and doesn’t fit into a REST API as easily as Rasa NLU does.
Rasa Core with minimal Python¶
You can build a chatbot with Rasa Core by:
- defining a domain (a yaml file)
- writing / collecting stories (markdown format)
- running python scripts to train and run your bot
The only part where you need to write python is when you want to define custom actions. There’s an excellent python library called requests, which makes HTTP programming painless. If Rasa just needs to interact with your other services over HTTP, your actions will all look something like this:
from rasa_core.actions import Action import requests class ApiAction(Action): def name(self): return "my_api_action" def run(self, dispatcher, tracker, domain): data = requests.get(url).json return [SlotSet("api_result", data)]
Rasa Core with Docker¶
We provide a Dockerfile which allows you to build an image of Rasa Core
with a simple command:
docker build -t rasa_core .
The default command of the resulting container starts the Rasa Core server
--nlu options. At this stage the container does not
yet contain any models, so you have to mount them from a local folder into
The full run command looks like this:
docker run \ --mount type=bind,source=<PATH_TO_DIALOGUE_MODEL_DIR>,target=/app/dialogue \ --mount type=bind,source=<PATH_TO_NLU_MODEL_DIR>,target=/app/nlu \ rasa_core
You also have the option to use the container to train a model with
docker run \ --mount type=bind,source=<PATH_TO_STORIES_FILE>/stories.md,target=/app/stories.md \ --mount type=bind,source=<PATH_TO_DOMAIN_FILE>/domain.yml,target=/app/domain.yml \ --mount type=bind,source=<OUT_PATH>,target=/app/out \ rasa_core train
You may in addition run any Rasa Core command inside the container with
docker run rasa_core run [COMMAND].