LLMPromptNexus

A unified framework for interacting with Large Language Models (LLMs) through a standardized interface. LLMPromptNexus simplifies working with multiple LLM providers while providing powerful templating and batching capabilities.

🚀 Quick Start

pip install llmprompt-nexus

from llmprompt_nexus import NexusManager

# Initialize with your API keys
llm = NexusManager({
    "openai": "your-openai-key",
    "perplexity": "your-perplexity-key"
})

# Simple translation example
result = await llm.generate(
    input_data={
        "text": "Hello world",
        "source_language": "English",
        "target_language": "Spanish"
    },
    model_id="sonar-pro",
    template_name="translation"
)

🌟 Key Features

• Multiple LLM Providers: Seamlessly work with OpenAI, Perplexity, and more through a single interface
• Smart Template System: Pre-built and custom templates for common NLP tasks
• Efficient Batch Processing: Handle large-scale operations with automatic rate limiting
• Built-in Safety: Automatic retries, rate limiting, and error handling

📦 Installation

Using pip

pip install llmprompt-nexus

From source

git clone https://github.com/EEstevanell/llmprompt-nexus.git
cd llmprompt-nexus
pip install -e .

🔑 Configuration

1. Set your API keys as environment variables:

export OPENAI_API_KEY="your-key"
export PERPLEXITY_API_KEY="your-key"

2. Or provide them during initialization:

llm = NexusManager({
    "openai": "your-openai-key",
    "perplexity": "your-perplexity-key"
})

📘 Basic Usage

Using Built-in Templates

# Simple translation
result = await llm.generate(
    input_data={
        "text": "Hello world",
        "source_language": "English",
        "target_language": "Spanish"
    },
    model_id="sonar-pro",
    template_name="translation"
)

# Text classification
result = await llm.generate(
    input_data={
        "text": "I love this product!",
        "categories": ["positive", "negative", "neutral"]
    },
    model_id="sonar-pro",
    template_name="classification"
)

Using Custom Templates

Templates can be defined in two ways:

1. Using YAML files:

templates:
  technical_qa:
    template: |
      Context: {context}
      Question: {question}
      Provide a technical answer based on the context.
    description: "Technical Q&A template"
    system_message: "You are a technical expert."

2. Using Python dictionaries:

custom_template = {
    "template": """
    Analyze the following {language} code:
    
    {code}
    
    Provide:
    - Code quality score (0-10)
    - Best practices followed
    - Suggested improvements
    """,
    "description": "Template for code review",  # Optional
    "system_message": "You are an expert code reviewer.",  # Optional
}

# Use the custom template
result = await llm.generate(
    input_data={
        "language": "Python",
        "code": "def hello(): print('world')"
    },
    model_id="sonar-pro",
    template_config=custom_template  # Pass template directly
)

Batch Processing

texts = ["First text", "Second text", "Third text"]
batch_inputs = [
    {
        "text": text,
        "source_language": "English",
        "target_language": "Spanish"
    }
    for text in texts
]

results = await llm.generate_batch(
    inputs=batch_inputs,
    model_id="sonar-pro",
    template_name="translation"
)

🎯 Built-in Templates

LLMPromptNexus comes with several built-in templates:

• Translation: Convert text between languages
• Classification: Categorize text into predefined groups
• Intent Detection: Identify user intentions from text
• Question Answering: Generate answers based on context
• Summarization: Create concise text summaries

For detailed instructions and examples on using templates, see our Templates Guide.

⚙️ Template Schema

Templates follow this schema:

templates:
  template_name:
    template: |  # Required: The actual template text with {variables}
      Your template content here with {variable1} and {variable2}
    description: "Template description"  # Optional
    system_message: "System prompt for the LLM"  # Optional

Required fields:

• template: The template text with variables in {braces}

Optional fields:

• name: Template identifier (auto-generated if using dictionary)
• description: Brief description of the template's purpose
• system_message: System prompt for the LLM

📚 Documentation

For detailed documentation, visit our documentation site.

🤝 Contributing

Contributions are welcome! Please read our Contributing Guide for details on how to submit pull requests, report issues, and contribute to the project.

📄 License

This project is licensed under the Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0). This means you are free to:

• Share and redistribute the material in any medium or format
• Adapt, remix, and transform the material

Under these conditions:

• Attribution — You must give appropriate credit when using this work, especially in academic research
• NonCommercial — You may not use the material for commercial purposes

For the full license text, see the LICENSE file.