Build a Multi-Channel Chatbot with Rasa

1. 1
   Install Rasa and Initialize Project

   Install the Rasa open-source library using pip. Then, run the `rasa init` command to create a new project with all the necessary boilerplate files for training data (`data/`), custom logic (`actions/`), and model configuration (`config.yml`).
2. 2
   Train Your First Model

   Navigate into your new project directory. Use the `rasa train` command to train the NLU (intent recognition, entity extraction) and dialogue management models based on the example data in your `data/` directory.
3. 3
   Test Your Bot in the Shell

   After training completes, start an interactive session with your bot using `rasa shell`. This lets you test its responses directly in the terminal and see intent classification details for each message you send. Type `/stop` to end the session.
4. 4
   Configure an LLM Fallback

   To handle open-domain questions, set up a fallback to call an LLM. First, define a rule in `data/rules.yml` to trigger a custom action when NLU confidence is low. Then, implement that action in `actions/actions.py` to call your preferred LLM API.
5. 5
   Connect to a Web Channel

   To deploy your bot on a website, enable the REST channel. Add the channel configuration to `credentials.yml`. Then, run the Rasa server with the API enabled. You can now interact with your bot by sending POST requests to its webhook URL.

#!/bin/bash
# This script installs Rasa, creates a new project, trains it, and starts a shell to chat.

# 1. Ensure you have Python 3.8 - 3.10 and pip installed

# 2. Install Rasa Open Source
# It is recommended to do this in a virtual environment
# python -m venv .venv
# source .venv/bin/activate
pip install rasa

# 3. Create a new project non-interactively
rasa init --no-prompt --path my-rasa-bot

# 4. Navigate into the project directory
cd my-rasa-bot

# 5. Train the initial model
rasa train

# 6. Run the bot and talk to it in your shell
echo "Rasa setup complete. Now running 'rasa shell' to chat with your bot."
echo "Try saying 'hello' or 'I am sad'. Type '/stop' to exit."
rasa shell