LLM Proxy to reduce cost and complexity of using multiple LLMs


Install
pip install proxyllm==0.2.3

Documentation

ProxyLLM - Proxy Classes for Language Models

The goal of ProxyLLM is providing python library that enables developers to effortlessly interface with various LLMs such as OpenAI's GPT-4, Google's BERT, and others.

This package encapsulates the complex interactions with these LLMs behind intuitive and user-friendly proxy classes, allowing developers to focus on leveraging the power of these advanced models, without being bogged down by intricate API details and interfacing code.

Features (WIP)

  1. Multiple LLM Support: The package supports multiple LLMs, including but not limited to GPT-4, BERT, and others. It is also designed with flexibility in mind to accommodate future models.

  2. Uniform API: It provides a uniform API for different LLMs, allowing developers to switch between models without the need to extensively change the codebase.

  3. Error Handling: It includes built-in error handling and retry logic, ensuring that your application remains resilient against minor network hiccups and transient errors.

  4. Optimized Performance: With caching and other optimizations, this package makes sure you get the best possible performance out of your chosen LLM.

  5. Asynchronous Support: For developers who require high performance and non-blocking code execution, this package offers asynchronous methods.

Installation

You can install the proxyllm package via pip:

pip install proxyllm

Quickstart

Here's an example of how to create a ChatGPT proxy and use it to generate text:

import proxyllm

session = proxyllm.Session()
while (prompt := input("you : ")) != "exit":
    message = session.send(prompt)
    print(f"\n{message['role']} : {message['content']}\n")

Documentation

For detailed information on using this package, please refer to our documentation.

Contributing

We welcome contributions! Please see our contributing guidelines for details.

License

This project is licensed under the terms of the MIT license. See LICENSE for more details.