The eve platform is a project that is basically the combination of a smart assistant (chatbot) with an application developed in React Native, in order to monitor the prenatal of pregnant women.
Eve-mongo is a docker container, which contains all the application models. The tracker store of Rasa Core is also here.
In your root directory (eve-platform), execute the commands below:
$ cd containers/eve-mongo
$ docker-compose -f mongo-compose.yml up -d- Now you are connected to mongodb://eve_mongo:27017.
To run the MongoDb shell, you must enter the mongo docker container and run:
$ docker exec -it eve_mongo /bin/bashNow, inside the container, run the command below.
#root mongo- It will appear MongoDB shell version and an available terminal to operate.
Since we are using a docker container, there are some volumes inside. To do backups, you must execute, in the containers/eve-mongo directory:
docker run --rm --network evemongo_evenetwork \
--link eve_mongo:mongo -v $(pwd)/backups:/backups \
mongo bash -c "mongodump --out /backups \
--host mongo:27017"- This will generate two backup files and you can go to /container/eve-mongo/backups to get them. If you wish to backup your production database, you can change the host argument and run the command above.
To restore backups, you must execute, in the containers/eve-mongo directory:
docker run --rm --network evemongo_evenetwork \
--link eve_mongo:mongo -v $(pwd)/backups:/backups \
mongo bash -c "mongorestore /backups \
--host mongo:27017"- You can run the same command to production database. Just change host argument.
Eve-api is the main api of the platform. Through requests, we communicate with basically the entire application. The api that makes the main communication with the database.
In your root directory (eve-platform), execute the commands below:
$ cd containers/eve-api
$ docker-compose -f api-compose.yml up- Open up URL http://localhost:5001/ and it will appear a message Flask is running in port 5001.
If you look inside /eve-api/config, you will find a file named mongo_configuration.json. In this file, we configure all mongo settings.
{
"DATABASE_NAME": "evedb",
"DATABASE_URL": "mongodb://eve_mongo:27017/evedb"
}- If you need to change it, you have to change in this code. Our API is already pointing to the mongo inside docker container, by default.
Eve-Rasa contains all the files and models to run a Rasa Chatbot.
To start our containers, make sure you have your trained models. If you have not, Rasa Core container will not run. By default, our Dockerfile is configured to build with docker-endpoints.yml, which is our configuration to communicate with Mongo Container and Rasa Action Server container. Our production configuration is inside endpoints.yml.
Go to containers/eve-rasa and run:
docker-compose -f rasa-compose.yml up -d- Now, our Rasa containers are running and you can check their status using Portainer or running the command below
docker ps-
To make sure everything is working, you can open URL http://localhost:5005 and http://localhost:5055/health. It will appear hello from Rasa: 0.15.0a1 (Rasa Core) and status OK (Rasa Action Server).
-
Rasa containers communicate with mongo to track and store all bot's conversations.
To do all the steps before, you can run them by a single script. Our start-rasa-containers.sh script runs Rasa Action Server and Rasa containers.
- Go to containers/eve-rasa and run the commands below
chmod +x ./scripts/start-rasa-containers.sh
./scripts/start-rasa-containers.shGo to containers/eve-rasa directory and run these commands below, so you can run train.sh script.
chmod +x ./scripts/train.sh
./scripts/train.sh
If you check containers/eve-rasa/models directory, you will find our trained model.
If you wish to use Rasa X, you should start a local environment for Rasa Action Server. To do so, just run these commands below, in containers/eve-rasa/actions directory.
chmod +x ./scripts/start-action-server.sh
./scripts/start-action-server.sh
Go to containers/eve-rasa directory and run these commands below, so you can run start-rasa-x.sh script. With Rasa X you can talk to your bot or start interactive learning. Make sure you have started your local action server, in other terminal. You must know that all Rasa running containers will be stopped.
chmod +x ./scripts/start-rasa-x.sh
./scripts/start-rasa-x.sh
You can execute Rasa Interactive Learning via terminal, but not inside docker container. Run the script below, and it will run outside our containers. You must know that all Rasa running containers will be stopped.
chmod +x ./scripts/train-interactive.sh
./scripts/train-interactive.sh
Talking to your bot
You can talk with your bot by sending POST requests with some parameters to URL http://localhost:5005/webhooks/rest/webhook.
Open a tool called Postman or your can use curl to do it. Your JSON example data is below, which you can send a message, with the sender (user_id).
{
"message": "Hello",
"sender":"1"
}This will return a bot response, like shown below.
[
{
"recipient_id": "10",
"text": "Hello!"
}
]- This will be stored in Rasa database in our Mongo container.
Heroku is hosting our Eve API, Eve Rasa Action Server and Eve Rasa Core. Basically, to refresh your changes there, you must run the script below.
- Go to the Container's directory you wish to Deploy. Check if your Dockerfile is pointing to production configuration's and run the command below.
chmod +x ../../scripts/deploy.sh
../../scripts/deploy.sh
- This will deploy your container image to Heroku's application.
Portainer is a tool to build and manage your Docker environments.
Go to your home directory (not related to your project). And to run Portainer, we will create a docker container, by the command below.
docker run -d -p 9000:9000 --name portainer --restart always -v /var/run/docker.sock:/var/run/docker.sock
-v {USER_SYSTEM_FOLDER}/portainer/data:/data portainer/portainer- Acessing a URL http://localhost:9000 from your browser, it will appear a home screen of Portainer.
It's a tool to edit your training examples for Rasa NLU.
Go to {PROJECT_DIRECTORY}/containers/eve-rasa/data/json (NLU data directory). And to run Rasa NLU Trainer, we will create a docker container, by executing the code below.
docker run -t --name rasa-nlu-trainer -v $(pwd):/rasa-nlu-trainer/src/state -p 5006:8080 -i \
dominicbreuker/rasa-nlu-trainer- This will search for the first .json file in the folder. Now, acessing a URL http://localhost:5006 from your browser, it will appear your formatted NLU json to edit.