Knowledge Support Chatbot

A flexible conversational AI assistant that provides step-by-step guidance for product usage through natural language interactions. The system uses a finite state machine (FSM) architecture to manage dialogue flow and can be adapted to support any product by modifying the knowledge base in the data folder.

Features

• Natural language understanding for user queries
• Dynamic task identification and goal clarification
• Step-by-step guidance with adaptive detail levels
• User proficiency tracking
• Contextual response generation
• Easily adaptable to different products/domains
• Multi-step task guidance with primitive operations

Architecture

The system uses a hierarchical FSM architecture with:

• Parent FSM for high-level conversation flow
• Sub-FSMs for specific task handling
• RAG (Retrieval-Augmented Generation) for natural responses
• Modular knowledge base structure

Core Components

1. Dialogue Engine: Manages conversation flow and state transitions
2. User Profile: Tracks user state, proficiency, and task progress
3. Response Generator: Generates contextual responses using RAG
4. NLU Unit: Processes natural language inputs

Installation

1. Clone the repository:

git clone https://github.com/Elliot0122/knowledge-support-chatbot.git
cd knowledge-support-chatbot

2. Install dependencies:

pip install -r requirements.txt

3. Start the Ollama server (required for LLM functionality):

ollama serve

4. Run the application:

python run.py

Customization

The chatbot can be adapted to support different products by modifying the data files:

Data Structure

data/
├── fsm/ # Conversation flow definitions
│ ├── states.json # Main conversation states
│ └── submodule/ # Task-specific state machines
│   ├── opening.json
│   ├── goal_clarification.json
│   └── task_execution.json
├── nlg/ # Natural Language Generation
│ └── examples.json # Response templates
└── tasks/ # Product-specific tasks
  └── task_list.json # Task definitions and primitives

Customization Steps

1. Define Tasks: Update task_list.json with your product's tasks and their primitive operations
2. Update Responses: Modify examples.json with product-specific response templates
3. Adjust States: Customize FSM states and transitions if needed

Usage

The API server will start on http://localhost:9000. You can interact with it using:

import requests

# Initialize conversation
response = requests.get("http://localhost:9000/init")
print(response.json()["response"])

# Send a message
response = requests.post(
"http://localhost:9000/chat",
json={"text": "I need help with this product"}
)
print(response.json()["response"])

Project Structure

knowledge-support-chatbot/
├── src/ # Source code
│ ├── core/ # Core dialogue and business logic
│ ├── models/ # FSM and data models
│ └── utils/ # Utility functions
└── data/ # Knowledge base and configuration

Key Features

1. Adaptive Guidance

   • Adjusts detail level based on user proficiency
   • Provides step-by-step instructions
   • Offers additional details when needed

2. Flexible Task Management

   • Dynamic task identification
   • Goal clarification dialogue
   • Multi-step task decomposition

3. Natural Interaction

   • Context-aware responses
   • Natural language understanding
   • Conversational state management

API Endpoints

• GET /init: Initialize a new conversation
• POST /chat: Send a message to the assistant
  • Request body: {"text": "your message here"}
  • Response: {"response": "assistant's response"}

Requirements

• Python 3.11+
• Ollama (for LLM functionality)
• FastAPI
• Uvicorn