Nova is a rule-based conversational chatbot developed as part of the DecodeLabs Artificial Intelligence Industrial Training Program.
The project demonstrates how a chatbot can process user input, match it against predefined conversational rules, and return consistent responses through a structured decision-making system.
Unlike machine-learning or generative AI chatbots, Nova does not generate responses using a trained model or external API. Instead, it uses deterministic Python logic, predefined response mappings, input normalization, exit-command detection, and fallback handling.
The chatbot includes both:
- A web-based interface developed with Streamlit
- A command-line interface developed with Python
This modular structure separates the chatbot engine from the presentation layer, making the application easier to understand, test, maintain, and extend.
The deployed chatbot is available through Streamlit Community Cloud.
The live application allows users to:
- Send messages through an interactive chat interface
- View the complete conversation during the active session
- Test predefined questions and commands
- Receive controlled fallback responses
- End a conversation using exit commands
- Restart the conversation using the sidebar
- Review example questions supported by the chatbot
The main objectives of this project were to:
- Understand the fundamental architecture of rule-based chatbots
- Build a deterministic conversational response system
- Store predefined responses using Python dictionaries
- Normalize user input before rule matching
- Handle empty and unsupported messages safely
- Detect commands that end the conversation
- Develop a continuous command-line conversation loop
- Create a browser-based chatbot interface with Streamlit
- Separate application logic from interface logic
- Deploy the completed project as a publicly accessible web application
Nova uses predefined rules to generate responses. The same recognized input always produces the same output.
This makes the chatbot:
- Predictable
- Transparent
- Easy to test
- Easy to control
- Suitable for structured conversations
Before attempting to match a user message, the chatbot cleans the input by:
- Converting text to lowercase
- Removing leading spaces
- Removing trailing spaces
- Reducing unnecessary whitespace
For example:
" WHAT IS YOUR NAME "
is normalized to:
"what is your name"
User: hello
Nova: Hello! How can I help you today?
User: what is your name
Nova: My name is Nova. I am a rule-based chatbot.
User: something unknown
Nova: I don't understand that yet. Type 'help' to see what you can ask.
User: exit
Nova: Goodbye! Thanks for chatting with me.
- Responds only to predefined inputs
- Does not understand broader conversation context
- Does not learn from previous interactions
- Does not use machine learning or a large language model
- Does not permanently store conversation history
- May not recognize differently worded questions
- Add keyword-based intent recognition
- Support multiple phrases for each response
- Add fuzzy matching for spelling mistakes
- Expand the predefined response library
- Store conversation history in a database
- Add automated tests
- Introduce multilingual support
The application is complete and publicly available through Streamlit Community Cloud.
Artificial Intelligence Intern | DecodeLabs
This project was developed as part of the DecodeLabs Artificial Intelligence Training Program to demonstrate the design and deployment of a rule-based conversational application using Python and Streamlit.
