metadata

title: Omniscient
emoji: 👁️‍🗨️
colorFrom: indigo
colorTo: purple
sdk: streamlit
python_version: 3.11
sdk_version: 1.35.0
app_file: app.py
pinned: false

🧠 Omniscient

"The all-knowing AI that sees everything, knows everything"

A versatile AI bot for image analysis and dataset curation with support for multiple AI models

🎮 Try it Live on HuggingFace! (Actively WIP)

✨ Features

🗃️ Dataset Curation

Generate and curate high-quality image datasets with intelligent filtering and categorization.

🔍 Single Image Analysis

Benchmark different AI models on individual images with detailed performance metrics.

🤖 Agentic Analysis

Multi-step AI reasoning and analysis with advanced decision-making capabilities.

🌐 Multiple AI Providers

Seamless integration with OpenAI, Anthropic, and Google AI platforms.

🚀 Quick Start

📋 Step 1: Setup Environment

cd simple_G_ai_bot

Create a .env file in the project root:

# 🔐 .env
OPENAI_API_KEY=your_openai_api_key_here
ANTHROPIC_API_KEY=your_anthropic_api_key_here
GOOGLE_API_KEY=your_google_api_key_here

📦 Step 2: Install Dependencies

uv sync

🎯 Step 3: Usage Examples

🏗️ Dataset Curation

Generate 50 urban outdoor samples:

python main.py --mode data --samples 50 --urban --no-indoor

⚡ Single Image Analysis

Benchmark GPT-4o on 5 samples:

python main.py --mode benchmark --models gpt-4o --samples 5

🧠 Agentic Analysis

Run multi-step analysis with Gemini:

python main.py --mode agent --model gemini-2.5-pro --steps 10 --samples 5

⚙️ Configuration

🔑 Environment Variables

Variable	Description	Status
`OPENAI_API_KEY`	OpenAI API key for GPT models	🔶 Optional
`ANTHROPIC_API_KEY`	Anthropic API key for Claude models	🔶 Optional
`GOOGLE_API_KEY`	Google AI API key for Gemini models	🔶 Optional

🛠️ Command Line Options

🌟 Common Options

--mode → Operation mode (data, benchmark, agent)
--samples → Number of samples to process (default: 10)

🏙️ Data Mode Options

--urban → Focus on urban environments
--no-indoor → Exclude indoor scenes

📊 Benchmark Mode Options

--models → AI model to use (e.g., gpt-4o, claude-3, gemini-pro)

🤖 Agent Mode Options

--model → AI model for agentic analysis
--steps → Number of reasoning steps (default: 5)

🎯 Supported Models

Provider	Models	Status
🔵 OpenAI	GPT-4o, GPT-4, GPT-3.5-turbo	✅ Active
🟣 Anthropic	Claude-3-opus, Claude-3-sonnet, Claude-3-haiku	✅ Active
🔴 Google	Gemini-2.5-pro, Gemini-pro, Gemini-pro-vision	✅ Active

📋 Requirements

Prerequisites:

🐍 Python 3.8+

📦 UV package manager

🔑 Valid API keys for desired AI providers

🔧 Installation

1️⃣ Clone the repository

git clone <repository-url>

2️⃣ Navigate to project directory

cd simple_G_ai_bot

3️⃣ Create .env file with your API keys

touch .env
# Add your API keys

4️⃣ Install dependencies

uv sync

5️⃣ Run the bot with desired mode and options! 🎉

💡 Examples

🏗️ Basic Dataset Generation

python main.py --mode data --samples 20

🌆 Urban Scene Analysis

python main.py --mode data --samples 30 --urban --no-indoor

⚔️ Model Comparison

# GPT-4o Analysis
python main.py --mode benchmark --models gpt-4o --samples 10

# Claude-3 Analysis  
python main.py --mode benchmark --models claude-3-opus --samples 10

🧠 Advanced Agentic Workflow

python main.py --mode agent --model gemini-2.5-pro --steps 15 --samples 3

🔐 Security Note

⚠️ Important: Never commit your .env file to version control. Add .env to your .gitignore file to keep your API keys secure.

📜 License

MIT License - see LICENSE file for details.

⭐ Star this repo if you find it useful!