Interactive Text Adventure with LLM and Interrupts

A minimal “Choose Your Own Story” game built on LangGraph that lets a large language model generate the narrative, pause for user input, and then continue based on the choice.

📖 Overview

The graph starts by asking an LLM to write the opening of a story and propose several actions.
It then interrupts the flow, presenting those options in the console via questionary.
After the player selects an option, the graph resumes and asks the LLM to finish the scene with that choice.

This demonstrates:

State management with TypedDict
Custom interrupt nodes (langgraph.types.interrupt)
Human‑in‑the‑loop pattern
Checkpointing (in‑memory in this example)

⚙️ Installation

# Create a virtual environment (optional but recommended)
python -m venv .venv
source .venv/bin/activate   # Windows: .venv\Scripts\activate

# Install dependencies
pip install langgraph==0.1.9 \
          langchain==1.2.10 \
          langchain-openai==1.1.9 \
          questionary

OpenAI API key
The script uses ChatOpenAI from LangChain, which requires an OpenAI key. Export it as an environment variable:

export OPENAI_API_KEY="sk-..."
# Windows: set OPENAI_API_KEY=sk-...

📂 Project Structure

.
├── solution.py          # Main script – the graph definition and runner
└── README.md            # You’re reading it now

🚀 Running the Game

python solution.py

You’ll see something like:

LLM: "You find yourself in a dimly lit tavern..."
Options:
  1) Order a drink
  2) Ask about the town's rumors
  3) Leave immediately
Enter choice (1-3):

Type 1, 2 or 3 and press Enter. The LLM will then generate a short continuation based on your selection.

📋 Example Session

$ python solution.py
LLM: "You stand at the crossroads of destiny, with three paths before you..."
Options:
  1) Take the left path into the forest.
  2) Walk straight toward the looming castle.
  3) Turn back to the village.
Enter choice (1-3): 2

LLM: "You stride confidently toward the castle gates, feeling the weight of history upon your shoulders..."

Feel free to modify solution.py to add more nodes, change prompts, or persist checkpoints with a different saver.

🔧 Customization Tips

What	How
Different LLM	Replace `ChatOpenAI()` with another LangChain model (e.g., Anthropic).
Persisting state	Swap `InMemorySaver` for `SQLiteSaver`, `RedisSaver`, etc.
More complex choices	Add additional interrupt nodes or loop back to earlier parts of the story.

📜 License

This project is provided as a learning example and may be freely used, modified, and distributed.

2.8 KiB Raw Blame History Unescape Escape