llm-canvas 0.1.1

Visualize LLM conversations as a navigable canvas

LLM Canvas

Visualize and navigate Large Language Model conversations like never before.

LLM Canvas allows you to programmatically create interactive, branching conversation trees. Build complex conversation structures, explore different response paths, and visualize tool calls in a beautiful web interface — all while maintaining complete privacy with local deployment.

📰 News

🎉 August 2025: LLM Canvas v0.1.1 released with improved branching API and enhanced web UI

🌟 Key Features

• 🌳 Branching Conversations: Create and explore multiple conversation paths from any message
• 🔧 Tool Call Visualization: See how your LLM uses tools with clear input/output flows
• 📦 Zero Dependencies: Self-contained with built-in web UI

🚀 Quick Start

Installation

pip install llm-canvas
llm-canvas server --port 8000

Basic Usage

from llm_canvas import CanvasClient

# Create a client and canvas
client = CanvasClient()
canvas = client.create_canvas("My Conversation", "Exploring LLM interactions")

# Add messages
user_msg_id = client.add_message(canvas.canvas_id, "What is machine learning?", "user")
client.add_message(
    canvas.canvas_id,
    "Machine learning is a subset of AI that enables computers to learn from data...",
    "assistant",
    parent_node_id=user_msg_id
)

Start Local Server

# Start the local server
llm-canvas server --port 8000

# Server starts at http://localhost:8000
# Create and view your canvases in the web interface

💡 What Makes LLM Canvas Special?

Git-Like Conversation Management

Just like Git manages code versions, LLM Canvas manages conversation versions:

• Branches: Multiple conversation paths within a single canvas
• Commits: Each message is committed to a branch
• Checkout: Switch between different conversation branches
• Merge: Explore different paths and outcomes

Advanced Use Cases

1. Conversation Branching

# Create different response paths
main_branch = canvas.checkout("main", create_if_not_exists=True)
main_branch.commit_message({"role": "user", "content": "Explain quantum computing"})

# Create alternative explanations
simple_branch = canvas.checkout("simple-explanation", create_if_not_exists=True)
simple_branch.commit_message({"role": "assistant", "content": "Quantum computing uses quantum mechanics..."})

technical_branch = canvas.checkout("technical-explanation", create_if_not_exists=True)
technical_branch.commit_message({"role": "assistant", "content": "Quantum computing leverages superposition and entanglement..."})

2. Tool Usage Visualization

# Visualize how LLMs use tools
client.add_message(canvas_id, [
    {"type": "text", "text": "I'll check the weather for you."},
    {"type": "tool_use", "id": "weather_001", "name": "get_weather", "input": {"location": "San Francisco"}}
], "assistant")

client.add_message(canvas_id, [
    {"type": "tool_result", "tool_use_id": "weather_001", "content": '{"temperature": 72, "condition": "sunny"}'}
], "user")

🎯 Perfect For

• 🔬 AI Researchers: Analyze conversation patterns and model behavior
• 👨‍💻 LLM Developers: Debug and optimize chatbot interactions
• 📚 Educators: Teach AI concepts with visual examples
• 🚀 Startups: Prototype and demo conversational AI applications
• 🔒 Privacy-Conscious Users: Keep all data local and secure

📚 Examples & Documentation

Interactive Examples

Run the included examples to see LLM Canvas in action:

# Start the server first
llm-canvas server

# Then run examples (in a new terminal)
python -m examples.run_examples

# Or run specific examples:
python -m examples.hello_example          # Basic conversation
python -m examples.weather_tool_example   # Tool usage demo
python -m examples.vacation_planning_example  # Conversation branching
python -m examples.investment_decision_example  # Complex multi-branch scenario

Visit http://localhost:8000 to see all your canvases in the web interface!

Key Concepts

Concept	Description	Like Git
Canvas	A workspace for LLM conversations	Repository
Branch	Linear conversation threads	Branch
Message	Individual conversation entries	Commit
Checkout	Switch between conversation paths	git checkout
Commit	Add messages to a branch	git commit

📝 See detailed examples → examples/README.md

🛠️ Development

Local Development Setup

git clone https://github.com/LittleLittleCloud/llm_canvas.git
cd llm_canvas

# Install dependencies
pip install -e ".[server,dev]"

# Build frontend
cd web_ui && npm install && npm run build && cd ..

# Start development server
llm-canvas server --port 8000

Frontend Development

# Start backend server
llm-canvas server --port 8000

# Start frontend dev server (in another terminal)
cd web_ui
npm run dev  # Runs on http://localhost:5173

🔒 Privacy & Data

Local-First Philosophy: Your conversations stay on your machine.

• ✅ No external API calls for the visualization server
• ✅ No data collection or telemetry
• ✅ No account required for local usage
• ✅ Full control over your conversation data
• ⚠️ Session-based storage (data lost on server restart)

Note: The local server does not persist data between sessions. For permanent storage, consider our cloud plans or implement your own persistence layer.

🌐 Deployment Options

Local Server (Free & Open Source)

• ✅ Complete privacy control
• ✅ All visualization features
• ❌ No data persistence
• ❌ Session-only storage

Cloud Server (Coming Soon)

• ✅ Permanent data storage
• ✅ Cross-device access
• ✅ Team collaboration
• ✅ Backup & recovery

🤝 Contributing

We welcome contributions! LLM Canvas is open source and community-driven.

• 🐛 Bug Reports: GitHub Issues
• 💡 Feature Requests: GitHub Discussions
• � Pull Requests: See our Contributing Guide
• 📖 Documentation: Help improve our docs

📄 License

MIT License - see LICENSE for details.

---

Built with ❤️ for the AI community

Transform your LLM conversations from linear logs into navigable, visual experiences. Start exploring today!