unified_interface

Introduction

In this folder, I provide a unified interface for querying different models. Currently, it only supports models from OpenAI (the GPT models), Google (the Gemini models), and Together.ai (various open-source models). But, it shouldn't be too difficult to extend it to other providers.

To use the unified interface, you need to install the dependencies first.

• openai
• google-generativeai
• together

You will also need to obtain the API keys for these providers.

Explanation of the code

In api_factory.py, I define a set of classes to support different providers and a function create_api_client to create a client for a given provider.

Now you can use the same interface to query different models from different providers.

# First, choose a provider and a model
import api_factory

provider = "openai"
model = "gpt-4o"

# provider = "google"
# model = "gemini-1.5-flash"

# provider = "together"
# model = "meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo"

# Then, create a client for the chosen provider
api_client = api_factory.create_api_client(provider, api_key)
result = api_client.query_model(model, system_prompt, user_instruction)

In query_llm.py, I demonstrate how to query a LLM using a unified interface.

Caveats

Different models within each provider might have different parameters, so you need to go to api_factory.py to change those parameters directly.

I haven't tested querying those APIs asynchronously, so you might need to figure it out yourself.