MedBot Installation Guide

MedBot can be installed from its source code or by docker images.

Install MedBot by docker

1. Install docker-compose

2. Create your token

   echo TOKEN=$(tr -dc A-Za-z0-9 </dev/urandom | head -c 20) > .env

3. Create docker-compose file:

   ```
   $ cat docker-compose.yml

   services:
   rasa:
   image: medbot/medbot_server:latest
   container_name: chatbot-server
   restart: always
   ports:
   - 5005:5005
   volumes:
   - ./logs:/rasa-server/rasa/logs
   networks:
   - rasa-netowrk
   env_file:
   - .env
   command: [$TOKEN]
   app:
   image: medbot/action_server:latest
   container_name: action-server
   restart: always
   volumes:
   - ./logs:/action-server/logs
   networks:
   - rasa-netowrk
   expose:
   - 5055

   networks:
   rasa-netowrk:
   driver: bridge
   ```

4. Enjoy your chatbot 😉

   docker-compose -f docker-compose.yml up -d

Install from source code

1. Clone Rasa branch

   git clone https://github.com/arezae/chatbot --branch rasa --single-branch --depth 1

2. Directory tree

   To run the Rasa chatbot, we need to run actions-server and chatbot-server separately. So we've separated actions, datasets, and the chatbot from each other.

   Action-Server will contain actions and datasets, and Rasa-Server will contain the chatbot model and its autocorrect component. In the rest of the document, we will install requirements and then run chatbot and action servers.

   production/ ├── action-server │   ├── actions │   ├── datasets │   └── docker └── rasa-server ├── autocorrect │   └── data ├── docker └── rasa ├── data ├── models └── tests

3. Install requirements

   Create python environment

   pip install --user --upgrade pip pip install --user virtualenv python -m venv rasa_env

4. Activate your python environment

   source virtualenv/bin/activate

5. Install rasa chatbot requirements

   pip install --no-cache-dir -r production/rasa-server/requirements.txt

6. Install rasa chatbot server and actions server requirements

   pip install --no-cache-dir -r production/rasa-server/requirements.txt -r production/action-server/requirements.txt

7. Download Spacy weights

   python -m spacy download en_core_web_md

8. Download autocorrect module dictionaries. This dictionary consists of English and medical words.

   cd production/rasa-server python -c "import autocorrect; autocorrect.Speller('en_med')"

9. Train rasa model ⌛

   cd production/rasa-server/rasa rasa train

10. Run rasa server
    We are almost done. Now we can run the rasa and actions server.

    • Run the Rasa server. If you would like to run the Rasa server with tokens to authenticate requests, You can add --auth-token YOUR_TOKEN at the end of the following command

      cd production/rasa-server/rasa mkdir logs rasa run --log-file logs/rasa-server.log --enable-api

    • Run the action server

      cd production/action-server/actions mkdir logs rasa run --log-file logs/action-server.log actions --actions actions

11. Enjoy chatting with MedBot 🤗

    curl --location --request POST 'http://localhost:5005/webhooks/rest/webhook' \ --header 'Content-Type: application/json' \ --data-raw '{ "message" : "Can you give me dosage information of Abilify?", "sender" : "default" }'

Build a Docker image

1. Clone Rasa branch

   git clone https://github.com/arezae/chatbot --branch rasa --single-branch --depth 1

2. Directory tree

   To run the Rasa chatbot, we need to run actions-server and chatbot-server separately. So we've separated actions, datasets, and chatbots from each other.

   Action-Server will contain actions and datasets, and Rasa-Server will contain the chatbot model and its autocorrect component.

   production/ ├── action-server │   ├── actions │   ├── datasets │   └── docker └── rasa-server ├── autocorrect ├── docker └── rasa ├── data └── tests

   In the rest of the document, we will build separated images for actions and rasa and then run them together.

3. Genrate token

   • Generate token for requests authentication. You can do this manually or use the urandom device file.

     echo TOKEN=$(tr -dc A-Za-z0-9 </dev/urandom | head -c 20) > production/.env

   • Check your token:

     $ cat production/.env TOKEN=SOME_RANDOM_STRING

4. Choose your Dockerfile

   Default Dockerfile is Slim-based. But you can choose other Dockerfiles in the docker directory. For example for the rasa-based images:

   cp production/rasa-server/docker/Dockerfile-Rasa production/rasa-server/Dockerfile cp production/action-server/docker/Dockerfile-Rasa production/action-server/Dockerfile

5. Review docker-compose file

   • In the case of using an old version of docker-compose or docker-engine, you might need to add the version key, on top of the docker-compose.yml file. For example in the case of the 1.25.0 version of docker-compose, you need to add version: "3.7" or version: "3" on top of your docker-compose file like this:
     version: "3.7" services: rasa: build: context: ./rasa-server dockerfile: Dockerfile args: - VERSION=v0.1.0 . . .

   • Also the version of the model weights you want to use, can be assigned by the VERSION argument in the docker-compose file. This will download the VERSION.tar.gz file from Dropbox while building a Docker images.

   Please note that, if you want to use the special version of our bot, you should clone that version alongside assigning that version as the VERSION argument in the docker-compose file. You can find out available versions here.

6. Build an image

   Build a rasa chatbot and action server images

   docker-compose -f production/docker-compose.yml build

7. Make containers and run images

   Run docker-compose to start and run the chatbot and its actions together in an isolated environment

   docker-compose -f production/docker-compose.yml up -d

8. Test your chatbot

   • Talk with your chatbot

     curl --location --request POST 'http://localhost:5005/webhooks/rest/webhook' \ --header 'Content-Type: application/json' \ --data-raw '{ "message" : "Can you give me dosage information of Abilify?", "sender" : "default" }'

   • Debug your model. (Use your token instead of SOME_RANDOM_STRING)

     curl --location --request POST 'http://localhost:5005/model/parse?token=SOME_RANDOM_STRING' \ --header 'Content-Type: application/json' \ --data-raw '{ "text" : "Can you give me dosage information of Abilify?" }'

   • Or you can use the following command. (Again, you should use your token instead of SOME_RANDOM_STRING)

     curl --location --request GET 'localhost:5005/conversations/default/tracker?token=SOME_RANDOM_STRING' \ --header 'Content-Type: application/json' \ --data-raw '{ "text" : "Can you give me dosage information of Abilify?" }'

9. Monitor the chatbot

   • Check for errors and warnings in logs

     docker-compose -f production/docker-compose.yml logs -f -t

   • Monitor a live stream of containers resource usage statistics

     docker stats

10. In removing the chatbot case

    Stop and remove chatbot containers

    docker-compose -f production/docker-compose.yml down