llm-reasoner 0.1.3

Transform any LLM into a methodical thinker that excels at systematic reasoning like OpenAI o1 and DeepSeek R1

🤔 LLM-Reasoner

Transform any LLM into a methodical thinker that excels at systematic reasoning like OpenAI o1 and DeepSeek R1

• Step-by-step reasoning with detailed explanations
• Dynamic confidence scoring and self-reflection
• Multiple reasoning strategies and approaches
• Robust error handling and recovery
• Support for mathematical notation using LaTeX
• Comprehensive thought process visualization
• Integration with multiple LLM providers

✨ What Makes It Special?

LLM-Reasoner enhances any Language Model with advanced reasoning capabilities:

• Think out loud and show its work (no more mysterious answers!)
• Double-check its own thinking
• Consider different angles before making up its mind
• Tell you how confident it is at each step
• Actually explain why it believes what it believes

🚀 Getting Started

Install LLM-Reasoner with pip:

pip install llm-reasoner

Configure your API keys:

# Using OpenAI? Pop this in:
export OPENAI_API_KEY="your-key"

# Team Google? Here you go:
export VERTEX_PROJECT="your-project"
export VERTEX_LOCATION="your-location"

# Claude fan? Got you covered:
export ANTHROPIC_API_KEY="your-key"

🎮 Quick Play

Try these commands to get started:

# Check out what models you can use
llm-reasoner models

# Ask it something cool
llm-reasoner reason "Why do planes stay up in the air?"

# Want a nice UI to play with?
llm-reasoner ui

🔧 Custom Models

You can register custom models with LLM-Reasoner in three ways:

1. Using the Web UI (Easiest):

1. Launch the UI with llm-reasoner ui
2. Click on "Register Custom Model" in the top section
3. Fill in your model details:
   • Model Name (e.g., "my-azure-gpt4")
   • Provider (e.g., "azure")
   • Context Window Size (optional)
4. Click "Register Model" to add it to your available models
5. Your new model will appear in the model selection dropdown

2. Using the CLI:

# Register a new model
llm-reasoner register-model my-custom-model azure --context-window 16384

# Set it as default
llm-reasoner set-model my-custom-model

# View all registered models
llm-reasoner models

3. Using Python:

from reasonchain import model_registry

# Register a custom model
model_registry.register_model(
    name="my-custom-model",
    provider="custom-provider",
    context_window=4096  # Optional
)

# Use your custom model
chain = ReasonChain(model="my-custom-model")

This allows you to use any LLM provider supported by LiteLLM. See LiteLLM's documentation for the full list of supported providers.

🎨 Interactive UI

Launch the visual interface with:

llm-reasoner ui

The UI provides:

• Model selection and registration via an easy-to-use form
• Parameter adjustment with intuitive sliders
• Real-time reasoning visualization
• Interactive exploration
• Custom model registration interface

🛠️ Using It In Your Code

Here's the simplest way to use LLM-Reasoner:

from reasonchain import ReasonChain
import asyncio

async def main():
    chain = ReasonChain()
    async for step in chain.generate("How does evolution work?"):
        print(f"🤔 Step {step.number}: {step.title}")
        print(step.content)

asyncio.run(main())

Want more control? Here's an advanced example:

chain = ReasonChain(
    model="gpt-4",                         # Pick your favorite model
    max_tokens=1000,                       # Let it think deeper
    temperature=0.3,                       # Control creativity
    prompt_template="Let's explore: {prompt}"  # Make it your own
)

# Get all the details about its thinking process
async for step in chain.generate_with_metadata("How do computers learn?"):
    print(f"💭 Step {step.number}: {step.title}")
    print(f"🎯 Confidence: {step.confidence}") 
    print(f"⏱️ Thinking time: {step.thinking_time}s")
    print(step.content)

🌟 Features in Detail

Each reasoning step includes:

• Step number (keeping things organized)
• Clear title (what it's pondering)
• Detailed thoughts (the good stuff)
• Confidence score (how sure it is)
• Thinking time (we track speed too!)
• Timestamp (when each thought happened)
• Next action (what it's planning)

Development

To contribute to LLM-Reasoner:

1. Clone the repository
2. Install development dependencies: pip install -e ".[dev]"
3. Run tests: pytest

🤝 Contributing

Found a bug or have ideas? We'd love to hear from you:

• Open an issue on GitHub
• Email us at [e-mail]
• Check out our examples

📜 License

MIT License - See LICENSE file for details.

---

Made with ❤️ for those who believe AI should show its work! ✍️