Rigging is a lightweight Python framework designed to simplify and enhance the interaction with large language models (LLMs) in production code. It targets developers building LLM-powered applications, agents, and services, offering a flexible and efficient way to integrate various LLM providers and features.

How It Works

Rigging leverages a Pydantic-based approach for structured LLM interactions, allowing seamless use of Pydantic models alongside unstructured text. It defaults to LiteLLM for broad model compatibility and supports defining prompts as Python functions with type hints and docstrings. Key features include simplified tool use (even for models without native support), a connection string system for managing models and configurations, and integrated tracing via Logfire.

Quick Start & Requirements

• Install via pip: pip install rigging
• Supports models from LiteLLM, vLLM, and Hugging Face Transformers.
• API keys can be passed via connection strings (e.g., gpt-4-turbo,api_key=...) or environment variables (e.g., OPENAI_API_KEY).
• Documentation: docs.dreadnode.io

Highlighted Details

• Defines prompts as Python functions with type hints and docstrings.
• Supports structured output parsing with Pydantic models.
• Offers integrated tracing with Logfire for monitoring.
• Enables tool use for models that don't natively support it.
• Provides async batching and iteration for scalable generation.

Maintenance & Community

Rigging is actively developed and used daily by its creators at dreadnode.

Licensing & Compatibility

The README does not specify a license.

Limitations & Caveats

The license is not explicitly stated in the README, which may pose a concern for commercial or closed-source integration.