ai-gradio 0.1.3

A Python package for creating Gradio applications with AI models

ai-gradio

A Python package that makes it easy for developers to create machine learning apps powered by OpenAI, Google's Gemini models, and CrewAI.

Installation

You can install ai-gradio with different providers:

# Install with OpenAI support
pip install 'ai-gradio[openai]'

# Install with Gemini support  
pip install 'ai-gradio[gemini]'

# Install with CrewAI support
pip install 'ai-gradio[crewai]'

# Install with all providers
pip install 'ai-gradio[all]'

Basic Usage

First, set your API key in the environment:

For OpenAI:

export OPENAI_API_KEY=<your token>

For Gemini:

export GEMINI_API_KEY=<your token>

Then in a Python file:

import gradio as gr
from ai_gradio import registry

# Create a Gradio interface
interface = gr.load(
    name='gpt-4-turbo',  # or 'gemini-pro' for Gemini
    src=registry,
    title='AI Chat',
    description='Chat with an AI model'
).launch()

Features

Text Chat

Basic text chat is supported for all models. The interface provides a chat-like experience where you can have conversations with the AI model.

Voice Chat (OpenAI only)

Voice chat is supported for OpenAI realtime models. You can enable it in two ways:

# Using a realtime model
interface = gr.load(
    name='gpt-4o-realtime-preview-2024-10-01',
    src=registry
).launch()

# Or explicitly enabling voice chat with any realtime model
interface = gr.load(
    name='gpt-4o-mini-realtime-preview-2024-12-17',
    src=registry,
    enable_voice=True
).launch()

Voice Chat Configuration

For voice chat functionality, you'll need:

1. OpenAI API key (required):

export OPENAI_API_KEY=<your OpenAI token>

2. Twilio credentials (recommended for better WebRTC performance):

export TWILIO_ACCOUNT_SID=<your Twilio account SID>
export TWILIO_AUTH_TOKEN=<your Twilio auth token>

You can get Twilio credentials by:

• Creating a free account at Twilio
• Finding your Account SID and Auth Token in the Twilio Console

Without Twilio credentials, voice chat will still work but might have connectivity issues in some network environments.

Video Chat (Gemini only)

Video chat is supported for Gemini models. You can enable it by setting enable_video=True:

interface = gr.load(
    name='gemini-pro',
    src=registry,
    enable_video=True
).launch()

AI Agent Teams with CrewAI

CrewAI support allows you to create teams of AI agents that work together to solve complex tasks. Enable it by using the CrewAI provider:

interface = gr.load(
    name='crewai:gpt-4-turbo',
    src=registry,
    title='AI Team Chat',
    description='Chat with a team of specialized AI agents'
).launch()

CrewAI Types

The CrewAI integration supports different specialized agent teams:

• support: A team of support agents that help answer questions, including:

  • Senior Support Representative
  • Support Quality Assurance Specialist

• article: A team of content creation agents, including:

  • Content Planner
  • Content Writer
  • Editor

You can specify the crew type when creating the interface:

interface = gr.load(
    name='crewai:gpt-4-turbo',
    src=registry,
    crew_type='article',  # or 'support'
    title='AI Writing Team',
    description='Create articles with a team of AI agents'
).launch()

When using the support crew type, you can provide a documentation URL that the agents will reference when answering questions. The interface will automatically show a URL input field.

Provider Selection

When loading a model, you can specify the provider explicitly using the format provider:model_name.

# Explicit provider
interface = gr.load(
    name='gemini:gemini-pro',
    src=registry
).launch()

Customization

You can customize the interface by adding examples, changing the title, or adding a description:

interface = gr.load(
    name='gpt-4-turbo',
    src=registry,
    title='Custom AI Chat',
    description='Chat with an AI assistant',
    examples=[
        "Explain quantum computing to a 5-year old",
        "What's the difference between machine learning and AI?"
    ]
).launch()

Composition

You can combine multiple models in a single interface using Gradio's Blocks:

import gradio as gr
from ai_gradio import registry

with gr.Blocks() as demo:
    with gr.Tab("GPT-4"):
        gr.load('gpt-4-turbo', src=registry)
    with gr.Tab("Gemini"):
        gr.load('gemini-pro', src=registry)
    with gr.Tab("CrewAI"):
        gr.load('crewai:gpt-4-turbo', src=registry)

demo.launch()

Supported Models

OpenAI Models

• gpt-4-turbo
• gpt-4
• gpt-3.5-turbo

Gemini Models

• gemini-pro
• gemini-pro-vision
• gemini-2.0-flash-exp

CrewAI Models

• crewai:gpt-4-turbo
• crewai:gpt-4
• crewai:gpt-3.5-turbo

Requirements

• Python 3.10 or higher
• gradio >= 5.9.1

Additional dependencies are installed based on your chosen provider:

• OpenAI: openai>=1.58.1
• Gemini: google-generativeai
• CrewAI: crewai>=0.1.0, langchain>=0.1.0, langchain-openai>=0.0.2, crewai-tools>=0.0.1

Troubleshooting

If you get a 401 authentication error, make sure your API key is properly set. You can set it manually in your Python session:

import os

# For OpenAI
os.environ["OPENAI_API_KEY"] = "your-api-key"

# For Gemini
os.environ["GEMINI_API_KEY"] = "your-api-key"

No Providers Error

If you see an error about no providers being installed, make sure you've installed the package with the desired provider:

# Install with OpenAI support
pip install 'ai-gradio[openai]'

# Install with Gemini support
pip install 'ai-gradio[gemini]'

# Install with CrewAI support
pip install 'ai-gradio[crewai]'

# Install all providers
pip install 'ai-gradio[all]'

Optional Dependencies

For voice chat functionality:

• gradio-webrtc
• numba==0.60.0
• pydub
• librosa
• websockets
• twilio
• gradio_webrtc[vad]
• numpy

For video chat functionality:

• opencv-python
• Pillow