WrappedLLM 0.1.2

A wrapper for various large language models including GPT, Claude, and Gemini

# LLM-Wrapper

LLM-Wrapper is a Python package that provides a unified interface for interacting with multiple Large Language Models (LLMs) including ChatGPT, Claude, and Gemini.

## Features

- Easy initialization of LLM clients
- Unified interface for generating outputs from different LLMs
- Support for multiple models within each LLM platform
- API key validation during initialization

## Installation

```bash
pip install WrappedLLM

Usage

Initializing LLMs

You can initialize LLMs individually or all at once:

from WrappedLLM import Initialize

# Initialize ChatGPT
Initialize.init_chatgpt("your_openai_api_key")

# Initialize Claude
Initialize.init_claude("your_anthropic_api_key")

# Initialize Gemini
Initialize.init_gemini("your_gemini_api_key")

# Initialize all LLMs at once
Initialize.init_all(
    chatgpt_api_key="your_openai_api_key",
    claude_api_key="your_anthropic_api_key",
    gemini_api_key="your_gemini_api_key"
)

Note: During initialization, a few tokens are used to verify that the provided API key is correct.

Generating Output

from WrappedLLM import Output

# Generate output using ChatGPT
gpt_response = Output.GPT("Tell me a joke about programming.")

# Generate output using Claude
claude_response = Output.Claude("Explain quantum computing in simple terms.")

# Generate output using Gemini
gemini_response = Output.Gemini("What are the benefits of renewable energy?")

Customizing Model Parameters

You can customize model parameters when generating output:

# Using a specific GPT model with custom temperature and max tokens
gpt_response = Output.GPT(
    "Summarize the history of artificial intelligence.",
    model="gpt-4o-mini-2024-07-18",
    temperature=0.7,
    max_tokens=2048
)

# Using a specific Claude model with custom temperature and max tokens
claude_response = Output.Claude(
    "Describe the process of photosynthesis.",
    model="claude-3-5-sonnet-20240620",
    temperature=0.5,
    max_tokens=1000
)

New Feature: GPT Function with Pydantic Class Outputs

You can now use the GPT function to return Pydantic class outputs. This provides a structured way to handle responses from the LLM.

Example:

from WrappedLLM import Initialize, Output
from pydantic import BaseModel

# Initialize the LLM
Initialize.init_chatgpt('your_api_key_here')

class CalendarEvent(BaseModel):
    name: str
    date: str
    participants: list[str]

# Use the Output class to interact with the LLM
response = Output.GPT(
    "Alice and Bob are going to a science fair on Friday.",
    system_prompt="Extract the event information.",
    response_format=CalendarEvent
)
print(response)

In this example, we define a CalendarEvent Pydantic class and use it as the response_format in the GPT function call. This allows the LLM to structure its response according to the defined class.

Available Models

You can get information about available LLM models using the get_llm_info() function:

from LLM.LLMModels import get_llm_info

llm_info = get_llm_info()

To get just the list of available models, you can use the LLM_MODELS dictionary:

from LLM.LLMModels import LLM_MODELS

available_models = LLM_MODELS

Note: While the get_llm_info() function includes information about image upload support, this functionality is not currently implemented in the LLM-Wrapper.

Error Handling

The package includes error handling for invalid API keys, unsupported models, and incorrect parameter values. Make sure to handle these exceptions in your code.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License.

This README provides an overview of the LLM-Wrapper package, including installation  instructions, usage examples for initializing LLMs and generating output, and information about customizing model parameters. It also mentions the API key validation during initialization and how to access the list of available models.