google/langfun

OO for LLMs

What it solves

Langfun simplifies the interaction between natural language and computer programs. It addresses the challenge of managing complex prompts and ensuring that LLM outputs are structured and predictable, which is often difficult when using raw text strings.

How it works

Langfun uses a concept called "Object-Oriented Prompting." By treating language as functions and leveraging PyGlove, it allows users to define prompts using Python objects and types. This enables the seamless integration of structured data into prompts and ensures that the LLM's response is returned as a typed Python object rather than just a plain string.

Who it’s for

It is designed for developers building AI agents and applications that require structured outputs from LLMs, as well as those who whom need a unified API to interact with multiple models like Gemini, GPT, and Claude.

Highlights

  • Object-Oriented Prompting: Prompting using objects and types for better control.
  • Unified API: Compatible with popular LLMs including Gemini, GPT, Claude, and Llama3 without fine-tuning.
  • Modular Prompts: Supports a blend of text and different modalities.
  • Evaluation Framework: Includes a powerful framework for evaluating LLM outputs.
  • Developer-Friendly: Provides intellisense and easy debugging with minimal overhead.

Related

  • Project
    google/pyglove

    A symbolic programming library for Python that enables the manipulation and automatic search of Python programs, commonly used for AutoML and evolutionary computing.

  • Project
    dreadnode/rigging

    A lightweight LLM framework that simplifies production code integration by providing unified model access, structured Pydantic outputs, and type-safe prompt definitions.

  • Project
    langgptai/LangGPT

    A structured, reusable prompt design framework that treats prompt engineering as a programming language to create high-quality, scalable AI personas.

  • Project
    langfuse/langfuse

    An open-source LLM engineering platform for tracing, monitoring, and evaluating AI applications to streamline the development and development cycle.

  • Project
    microsoft/TypeChat

    A library that enables developers to build natural language interfaces by replacing prompt engineering with schema engineering, ensuring LLM responses are structured and validated against defined types.