Flexible LLM library for code and agents

Rigging is a lightweight LLM framework to make using language models in production code as simple and effective as possible. Here are the highlights:

• Structured Pydantic models can be used interchangably with unstructured text output.
• LiteLLM as the default generator giving you instant access to a huge array of models.
• Define prompts as python functions with type hints and docstrings.
• Simple tool use, even for models which don't support them at the API.
• Store different models and configs as simple connection strings just like databases.
• Integrated tracing support with Logfire.
• Chat templating, forking, continuations, generation parameter overloads, stripping segments, etc.
• Async batching and fast iterations for large scale generation.
• Metadata, callbacks, and data format conversions.
• Modern python with type hints, async support, pydantic validation, serialization, etc.

import rigging as rg

@rg.prompt(generator_id="gpt-4")
async def get_authors(count: int = 3) -> list[str]:
    """Provide famous authors."""

print(await get_authors())

# ['William Shakespeare', 'J.K. Rowling', 'Jane Austen']

Rigging is built by dreadnode where we use it daily.

Installation

We publish every version to Pypi:

pip install rigging

If you want to build from source:

cd rigging/
poetry install

Supported LLMs

Rigging will run just about any language model:

• Any model from LiteLLM
• Any model from vLLM
• Any model from transformers

API Keys

Pass the api_key in an generator id or use standard environment variables.

rg.get_generator("gpt-4-turbo,api_key=...")

export OPENAI_API_KEY=...
export MISTRAL_API_KEY=...
export ANTHROPIC_API_KEY=...
...

Check out the docs for more.

Getting Started

Check out the guide in the docs

1. Get a generator using a connection string.
2. Build a chat or completion pipeline
3. Run the pipeline and get the output.

import rigging as rg
import asyncio

async def main():
    # 1 - Get a generator
    generator = rg.get_generator("claude-3-sonnet-20240229")

    # 2 - Build a chat pipeline
    pipeline = generator.chat(
        [
            {"role": "system", "content": "Talk like a pirate."},
            {"role": "user", "content": "Say hello!"},
        ]
    )

    # 3 - Run the pipeline
    chat = await pipeline.run()
    print(chat.conversation)

# Run the main function
asyncio.run(main())

# [system]: Talk like a pirate.
# [user]: Say hello!
# [assistant]: Ahoy, matey! Here be the salty sea dog ready to trade greetings wit' ye. Arrr!

Want more?

• Use structured pydantic parsing
• Check out raw completions
• Give the LLM access to tools
• Track behavior with tracing
• Play with generation params
• Use callbacks in the pipeline
• Scale up with iterating and batching
• Save your work with serialization

Examples

• Basic interactive chat: chat.py
• Jupyter code interpreter: jupyter.py
• OverTheWire Bandit Agent: bandit.py
• Damn Vulnerable Restaurant Agent: dvra.py
• RAG Pipeline: rag.py (from kyleavery)

Documentation

docs.dreadnode.io has everything you need.

Star History