Conversational LangGraph Patterns

Educational repository showcasing conversational patterns built with LangGraph. Each pattern demonstrates core concepts for building agentic conversational systems with practical, real-world examples.

Currently featuring three patterns:

• Booking Pattern: Guide users through booking appointments
• Modify Appointment Pattern: Help users modify or cancel existing appointments
• Detect Correction Pattern: Handle user corrections and intent changes in conversations

📚 Available Patterns

1. Booking Pattern

A conversational agent that guides users through the process of booking an appointment. The agent asks for missing information, validates availability, and confirms bookings.

• Video Tutorial: Watch on YouTube
• Detailed Article: Read on Medium
• Code: patterns/booking/

Key Concepts:

• Incremental information gathering
• Conditional routing based on state
• Availability validation
• Confirmation handling

2. Modify Appointment Pattern

A conversational agent that helps users modify or cancel existing appointments. The agent identifies the customer, looks up their appointments, processes modifications or cancellations, and validates availability for changes.

• Video Tutorial: Watch on YouTube
• Detailed Article: Read on Medium
• Code: patterns/modify_appointment/

Key Concepts:

• Customer identification and lookup
• Appointment selection from multiple results
• Modification request parsing (time, date, service)
• Availability validation for changes
• Confirmation handling for modifications
• Cancellation processing

3. Detect Correction Pattern

A conversational agent that detects and handles user corrections in real-time. The agent identifies when users are correcting previous statements (changing dishes, sizes, rejecting offers) and adjusts the conversation flow accordingly. Uses regex-based detection for fast, reliable correction identification.

• Video Tutorial: Watch on YouTube
• Code: patterns/detect_correction/
• Domain: Food ordering system

Key Concepts:

• Correction-first pattern detection
• Regex-based intent classification (no LLM required)
• Dynamic state management for multi-turn conversations
• Context preservation during corrections
• Early exit optimization for rejections
• Fallback handling for incomplete information

Pattern Flow:

1. Detect Correction - Identifies if user is correcting previous statement
2. Detect Intent - Determines user intention (order, modify, cancel, FAQ, finish)
3. Process Intent - Executes business logic based on intent and corrections
4. Respond User - Sends response and maintains conversation state

Example Conversations:

User: "I want a large pizza"
Bot: "Perfect, large pizza. Anything else?"
User: "Actually, make it small"
Bot: "Perfect, changed to small size. Your order: small pizza. Anything else?"

User: "How much does it cost?"
Bot: "Pizza: $15-25. You were ordering a pizza - what size?"

User: "I want tacos"
Bot: "What size would you like?"
User: "No thanks, never mind"
Bot: "No problem. Have a great day!"  ← Early exit (fast path)

🚀 Getting Started

Prerequisites

• Python 3.10+
• pip

Installation

1. Clone the repository:

git clone https://github.com/juanludataanalyst/langgraph-conversational-patterns.git
cd langgraph-conversational-patterns

2. Install dependencies:

pip install -r requirements.txt

3. Set up your environment (optional):

cp .env.example .env  # if you have an .env.example file

Running the Patterns

Option 1: LangGraph Studio (Recommended)

View and test all patterns in an interactive web interface:

uv run langgraph dev

Then open http://localhost:8000 in your browser. You'll see all three patterns available to test.

Option 2: Test Locally

Run a specific pattern directly:

# Booking pattern
python patterns/booking/graph.py

# Modify Appointment pattern
python patterns/modify_appointment/graph.py

# Detect Correction pattern
python patterns/detect_correction/graph.py

🤝 Contributing

We welcome contributions! Whether you want to:

• Add new conversational patterns
• Improve existing patterns
• Fix bugs
• Enhance documentation

Please read CONTRIBUTING.md for details on how to contribute.

📖 Learning Path

1. Video: Watch the tutorial to understand the pattern
2. Code: Explore the implementation in patterns/<pattern-name>/
3. Article: Read the detailed explanation on Medium
4. Experiment: Modify and extend the pattern

📁 Repository Structure

langgraph-conversational-patterns/
├── README.md                         # This file
├── CONTRIBUTING.md                   # How to contribute
├── LICENSE                           # MIT License
├── requirements.txt                  # Python dependencies
├── langgraph.json                    # LangGraph Studio configuration
├── img/                              # Visual assets
│   ├── BookingDiagram.PNG            # Booking pattern architecture
│   ├── Modify_appointment_diagram.PNG # Modify appointment pattern architecture
│   └── Correction_diagram.PNG         # Detect correction pattern architecture
└── patterns/
    ├── booking/                      # Booking pattern
    │   ├── __init__.py
    │   ├── state.py                  # State schema
    │   ├── nodes.py                  # Graph nodes
    │   └── graph.py                  # Graph definition
    ├── modify_appointment/            # Modify/Cancel appointments pattern
    │   ├── __init__.py
    │   ├── state.py                  # State schema
    │   ├── nodes.py                  # Graph nodes
    │   └── graph.py                  # Graph definition
    └── detect_correction/             # Detect correction pattern
        ├── __init__.py
        ├── state.py                  # State schema
        ├── nodes.py                  # Graph nodes
        └── graph.py                  # Graph definition

📝 License

This project is licensed under the MIT License - see the LICENSE file for details.

🙋 Questions?

If you have questions or suggestions, please open an issue on GitHub.

---

Happy learning! 🚀