CrewAI Enterprise Chatbot

This project demonstrates a CrewAI flow for building chatbots with CrewAI Enterprise. It integrates a single crew/agent to handle conversational routing and a Streamlit-based chatbot interface for demonstration purposes.

Project Overview

The project includes:

• A CrewAI flow with a configured crew/agent for handling chat messages.
• A Streamlit app (demo/streamlit_app.py) that provides an interactive chatbot interface.
• Backend API endpoints that showcase CrewAI Enterprise functionalities.

Installation

1. Clone the repository:

   git clone <repo-url>
   cd demo-conversational-crew

2. Install dependencies:

   crewai install

   or

   uv sync

3. Configure all required environment variables in .env file.

Usage

Running the Terminal Chat

A script named terminal_chat.py is provided to run a terminal-based chat session. To run it, execute:

.venv/bin/python terminal_chat.py

This script demonstrates how to interact with the CrewAI flow through the terminal.

Running the Streamlit Chatbot

Your crew needs to be deployed to CrewAI Enterprise for this demo to work.

To launch the interactive chatbot demo using Streamlit, run:

For secure handling of your crew URLs and API keys, this project uses secrets.toml. Follow these steps to set up your environment:

1. Create a secrets.toml file in the .streamlit directory:

   mkdir -p demo/.streamlit && touch demo/.streamlit/secrets.toml

2. Add your environment variables to secrets.toml:

   # demo/.streamlit/secrets.toml
   base_url = "https://YOUR-CREWAI-DEPLOYMENT-URL.crewai.com"
   bearer_token = "YOUR-BEARER-TOKEN-HERE"
   # Add any other necessary environment variables

3. Ensure secrets.toml is listed in your .gitignore file to prevent accidental commits of sensitive information.

When deploying to production, use the appropriate method for your hosting platform to set these environment variables securely.

streamlit run demo/streamlit_app.py

Run some automated tests

You can run some automated tests with two consecutive messages by executing:

.venv/bin/python test_chat.py

API Calls

The project utilizes CrewAI Enterprise API calls to manage chat sessions and process messages. Key API interactions include:

• Initiating a Chat Session: A POST request is sent to the designated endpoint with a JSON payload specifying the chat context, starting the conversation.

• Processing Chat Messages: User inputs are routed through the configured crew/agent. API calls handle routing, processing, and returning responses in a structured format.

For detailed implementation, review the configuration files in src/conversational_routing/crews/assistant_crew/config/ and the Crew files.

The API is a standard CrewAI Enterprise API, so refer to the CrewAI Enterprise API documentation for more information.

The first message sends the following JSON payload to the /kickoff endpoint:

{
    "current_message": "Hello, I am a chatbot. How can I help you today?"
}

In response, the API will respond with a kickoff_id:

{
    "id": "UNIQUE-KICKOFF-ID"
}

In order to get the chatbot response, keep polling the /status/{kickoff_id} endpoint. The response will include the response and id fields.

{
    "state": "SUCCESS",
    "result": "{ \"response\": \"Assistant response here.\", \"id\": \"UNIQUE-CONVERSATION-ID\" }"
}

The Crew is using @persist to store the conversation history and uses UNIQUE-CONVERSATION-ID as a key.

Consecutive messages should be sent to the /kickoff endpoint with the following payload:

{
    "current_message": "I need help with something.",
    "id": "UNIQUE-CONVERSATION-ID"
}

Additional Information

• Verify that any required environment variables are set.
• Check the source code for additional comments and developer notes regarding API endpoints and operation.
• Contributions and issues are welcome. Feel free to open a pull request or issue on the repository.