AI Prompt for Building the Project

In this tutorial, the entire chatbot backend project is generated using an AI coding assistant. You can copy the prompt below and paste it into any AI coding tool to generate the complete project automatically.

Create a conversational chatbot backend using Python that works completely offline without any paid third-party APIs.

The chatbot should expose REST APIs using FastAPI and use NLP and machine learning to understand user intent and generate responses.

Technology Stack

* Backend Framework: FastAPI
* Programming Language: Python
* NLP Library: NLTK
* Machine Learning Library: scikit-learn
* Data Format: JSON knowledge base
* Model Storage: Pickle
* Server: Uvicorn

Functional Requirements

1. The chatbot must detect user intent using machine learning.

2. Convert user text into vectors using TF-IDF or CountVectorizer.

3. Train a classifier using scikit-learn to identify the intent of the user query.

4. Store chatbot knowledge in a JSON file called intents.json.

5. When a user sends a message to the API:

   * preprocess the text
   * vectorize the input
   * predict the intent
   * return a random response from the response list.

6. If the chatbot does not understand the question, return a fallback response.

Large Knowledge Base Requirement

Generate a knowledge base containing at least 10,000 conversational patterns covering common daily life conversations.

The dataset should include intents and conversation patterns from the following lifestyle categories:

Greeting and introductions
Small talk conversations
Daily routine questions
Time and date queries
Weather discussions
Food and cooking conversations
Health and wellness advice
Motivation and inspirational messages
Jokes and humor
General knowledge questions
Technology and programming questions
Travel discussions
Shopping conversations
Work and office discussions
Study and education topics
Friendship and relationships
Entertainment (movies, music, games)
Sports conversations
Internet and social media
Self improvement and productivity

Each intent must include:

* 50–200 different conversational patterns
* 10–20 possible responses

Example intent format

{
"tag": "greeting",
"patterns": [
"hello",
"hi",
"hey",
"good morning",
"good afternoon",
"good evening"
],
"responses": [
"Hello! How can I help you?",
"Hi there!",
"Hey! What can I do for you today?"
]
}

Dataset Generation Requirement

Automatically generate at least 10,000 conversation patterns distributed across all intents.

The dataset should simulate natural human conversations such as:

* greetings
* asking about someone's day
* casual chatting
* asking for help
* discussing food, travel, work, or study
* simple knowledge questions
* light humor and jokes

API Requirements

Create the following FastAPI endpoints:

POST /chat

Request
{
"message": "Hello"
}

Response
{
"response": "Hi there!"
}

GET /health

Response
{
"status": "Chatbot API is running"
}

POST /train

This endpoint retrains the chatbot model using the updated intents.json dataset.

Project Structure

chatbot-project

app/
main.py
chatbot.py
trainer.py
utils.py

data/
intents.json

model/
chatbot_model.pkl

requirements.txt
README.md

Implementation Requirements

* Use NLTK for text preprocessing (tokenization, stemming, stopword removal)
* Use TF-IDF vectorization
* Train a Logistic Regression or Naive Bayes model
* Save the trained model using pickle
* Load the model on FastAPI startup
* Return responses through the API

Include:

* complete Python code
* dataset generation script that can generate 10,000 conversation patterns
* instructions to run the FastAPI server using:

uvicorn app.main:app --reload