THESIS_CAIRONE_FIORENTINO

Politecnico of Turin Thesis: "Implementation and Evaluation of an Educational Chatbot based on NLP Techniques"

GENERATE TOKEN SLACK

NB: REMEMBER TO REGENERATE SLACK TOKENS AND CHANNEL ID IF ARE DISABLE (SEE THESIS PDF)

• Go on your Slack App ( App Slack )-> select your app or create new -> Settings -> Install App -> (regenerate two tokens) -> Copy Tokens in actions/Constants.py and credentials.yml files.

RASA - LOCAL MODE

Components Installation

• Writing in Anaconda console this command:
  • conda create --name venv python==3.8.0
  • conda activate venv
  • conda install ujson tensorflow
  • pip install rasa
  • pip install rasa[spacy] for SPACY configuration
  • pip install rasa[trasformers] for BERT configuration
  • rasa init if you want to create new project

MORE INFO: See this video: link

Code Execution

• First, in TESI_POLI_POBOT directory do the training in anaconda console with venv activate:

  • rasa train --config configs/{file_cofiguration}.yml --fixed-model-name {model_name}

• After run server rasa:

• rasa run --connector slack --model {model_name} --debug

• In an other prompr run server rasa ACTIONS:

• rasa run actions --action actions --debug

• If you want,to test locally the intent classification, run in Anaconda console:

  • rasa train --config configs/{file_cofiguration}.yml --fixed-model-name {model_name}
  • rasa shell --model {model_name}

Rasa Bot - Slack Connection

• after training and running the server in Anaconda console (see ## Execution of code paragraph):
  • in BOT_RASA/ngrok directory run this command (in console prompt):
    • ngrok http 5005 to create a tunnel for localhost:5005 address:
    • select the forwarding address choose to ngrok ({address} = http://<number_code>.ngrok.io)
    • in Slack App link save the address (selecting in Features menu):
      • Interactivity & Shortcut > Request URL: ({address}/webhooks/slack/webhook) > Save Changes
      • OAuth & Permissions > Redirect URLs > Add : ({address}/webhooks/slack/webhook) > Save URLs
      • Event Subscriptions > Enable Events (Request URL) > Change: ({address}/webhooks/slack/webhook) > Save Changes
    • expired time of ngrok address: 2 hours. (when we will use a public address, ngrok will become USELESS)
• tokens bot save as environment variable in PC server (for security issue)
• Slack bot: link

To Connect also the ACTION Server Rasa:

• ngrok http 5055 to create a tunnel for localhost:5055
• In endpoints.yml modify url of action_endpoint: {address}/webhook

RASA X

Components Installation

• Writing in Anaconda console this command:
  • conda create --name venv python==3.9.0
  • conda activate venv
  • conda install ujson tensorflow
  • pip install rasa[spacy] for SPACY configuration
  • pip install rasa[trasformers] for BERT configuration
  • pip install --upgrade pip==20.0.1 pip3 install rasa-x --extra-index-url https://pypi.rasa.com/simple pip install --upgrade pip

Code Execution

• First, in TESI_POLI_POBOT directory do the training in anaconda console with venv activate:

  • set PYTHONUTF8=1
  • rasa x --connector slack --config configs/{file_cofiguration}.yml

• In an other prompt run server rasa ACTIONS:

• rasa run actions --action actions --debug

Rasa Bot - Slack Connection

• after training and running the server in Anaconda console (see ## Execution of code paragraph):
  • in BOT_RASA/ngrok directory run this command (in console prompt):
    • ngrok http 5005 to create a tunnel for localhost:5005 address:
    • select the forwarding address choose to ngrok ({address} = http://<number_code>.ngrok.io)
    • in Slack App link save the address (selecting in Features menu):
      • Interactivity & Shortcut > Request URL: ({address}/webhooks/slack/webhook) > Save Changes
      • OAuth & Permissions > Redirect URLs > Add : ({address}/webhooks/slack/webhook) > Save URLs
      • Event Subscriptions > Enable Events (Request URL) > Change: ({address}/webhooks/slack/webhook) > Save Changes
    • expired time of ngrok address: 2 hours. (when we will use a public address, ngrok will become USELESS)
• tokens bot save as environment variable in PC server (for security issue)
• Slack bot: link

To Connect also the ACTION Server Rasa:

• ngrok http 5055 to create a tunnel for localhost:5055
• In endpoints.yml modify url of action_endpoint: {address}/webhook

To Connect also the ACTION Server Rasa X:

• ngrok http 5055 to create a tunnel for localhost:5002
• In credentials.yml modify url of rasa: {address}/api

RASA - DOCKER MODE

Code Execution

If is the first time, run this commands:

docker volume create db-volume
docker-compose up -d
docker-compose down
docker-compose up -d
python rasa_x_commands.py create --update admin me <password_rasa_server> (i.e. password)
docker-compose down
docker-compose up -d

Otherwise, in TESI_POLI_POBOT directory in console run only this command:

• docker-compose up -d

Rasa Bot - Slack Connection

After all container is in running mode, in web browser go in `http:://localhost:80' and insert the password choose the first time. (i.e. password)

If rasa x server not contain NLU data (possible first time) you can upload the file directly using interface of rasa x server in Training -> NLU data / Stories / Rules / Configuration (copy and paste the content of config choose in TESI_POLI_POBOT directory)

If you want to connect to slack:

• after training and running the server in Anaconda console (see ## Execution of code paragraph):
  • in BOT_RASA/ngrok directory run this command (in console prompt):
    • ngrok http 80 to create a tunnel for localhost:80 address:
    • select the forwarding address choose to ngrok ({address} = http://<number_code>.ngrok.io)
    • in Slack App link save the address (selecting in Features menu):
      • Interactivity & Shortcut > Request URL: ({address}/core/webhooks/slack/webhook) > Save Changes
      • OAuth & Permissions > Redirect URLs > Add : ({address}/core/webhooks/slack/webhook) > Save URLs
      • Event Subscriptions > Enable Events (Request URL) > Change: ({address}/core/webhooks/slack/webhook) > Save Changes
    • expired time of ngrok address: 2 hours. (when we will use a public address, ngrok will become USELESS)
• tokens bot save as environment variable in PC server (for security issue)
• Slack bot: link

RASA - TEST DATA

Test stories in tests directory

• if you want tests stories with cross-validation:
  • rasa train --config configs/{file_cofiguration}.yml --fixed-model-name {model_name} (if model does not exist)
  • rasa test --config configs/{file_cofiguration}.yml --cross-validation --runs {time_runs} --folds {num_folds} --out {dir_out} --model {model_name}

Test NLU data

• if you want test only the NLU data splitted:

  • rasa train nlu --nlu train_test_split/training_data.yml --config configs/{file_cofiguration}.yml --fixed-model-name ./models_NLU/{model_name} (if model does not exist)
  • rasa test nlu --nlu train_test_split/test_data.yml --out {dir_out} --model ./models_NLU/{model_name}

• if you want test only the NLU data totally:

  • rasa train nlu --config configs/{file_cofiguration}.yml --fixed-model-name {model_name} (if model does not exist)
  • rasa test nlu --nlu train_test_split/test_data.yml --out {dir_out} --model {model_name}

Test CORE data

• if you want test only the CORE data:
  • rasa train --config configs/{file_cofiguration}.yml --fixed-model-name {model_name} (if model does not exist)
  • rasa test core --stories tests/test_stories.yml --out {dir_out} --model {model_name}