Modular AI Library

A Python library that allows you to import and use various AI models easily. This library supports multiple AI model providers, making it straightforward to switch between different models without writing initialisation code each time.

Features

• Easy Integration: Quickly integrate with multiple AI providers.
• Flexibility: Easily switch between different AI models.
• Convenience: Avoid the hassle of writing repetitive initialisation code.

Installation

First, install the library using pip:

pip install modular-ai

Setup

Environment Variables

This library requires a .env file for storing API keys and URLs for the different AI providers (You don't need to have all of them). Create a .env file in the root of your project and add the respective API keys and URLs like so:

OPENAI_API_KEY=your_openai_api_key
OPENAI_API_URL=openai_api_url

GEMINI_API_KEY=your_gemini_api_key
GEMINI_API_URL=gemini_api_url

ANTHROPIC_API_KEY=your_anthropic_api_key
ANTHROPIC_API_URL=anthropic_api_url

COHERE_API_KEY=your_cohere_api_key
COHERE_API_URL=cohere_api_url

GOOGLE_API_KEY=your_google_api_key
GOOGLE_API_URL=google_api_url

Make sure to replace the placeholder values with your actual API keys and URLs.

Usage

Importing the Library

Start by importing the necessary classes from the library:

from ailib.models import Models

Fetching and Registering Models

Fetch and register models from each provider. This step initialises the models and makes them available for use.

# Fetch and register models from each provider
Models.fetch_and_register_models("openai", "openai")
Models.fetch_and_register_models("gemini", "gemini")
Models.fetch_and_register_models("anthropic", "anthropic")
Models.fetch_and_register_models("cohere", "cohere")
Models.fetch_and_register_models("google", "google")

Generating Text Using a Registered Model

Once the models are registered, you can use them to generate text. Here's an example using OpenAI's GPT-3 model:

# Get the OpenAI model
model = Models.get_model("openai", "gpt-4o")

# Generating text without token limit
response = model.generate_text("Hello, world!")
print(response)

# Generate text with token limit
response = model.generate_text("Hello, world!", max_tokens=128000)
print(response)

Listing Available Models

You can list all registered models for each provider:

# List all registered models
all_models = Models.list_models()
print(all_models)

Full Example

Here's a complete example demonstrating how to set up and use the library:

from ailib.models import Models

# Fetch and register models from each provider
Models.fetch_and_register_models("openai", "openai")
Models.fetch_and_register_models("gemini", "gemini")
Models.fetch_and_register_models("anthropic", "anthropic")
Models.fetch_and_register_models("cohere", "cohere")
Models.fetch_and_register_models("google", "google")

# Generate text using a registered model
model = Models.get_model("openai", "gpt-4o")
response = model.generate_text("Hello, world!")
print(response)

# List all registered models
all_models = Models.list_models()
print(all_models)

Running Tests

To ensure everything is working correctly, you can run the provided test suite. This will verify the initialisation, model listing, and text generation for different APIs. Make sure you have your .env file properly set up with actual API keys to run these tests.

python -m unittest discover -s tests