EragAPI - Unified AI Service Interface

EragAPI provides a unified FastAPI-based interface for multiple AI providers, offering seamless integration with Groq, Gemini, Cohere, DeepSeek, and Ollama. Features streaming responses, CLI tools, and a system tray icon for server management.

Features

• Multi-Provider Support: Connect to Groq, Gemini, Cohere, DeepSeek, and Ollama.
• Streaming Responses: Server-Sent Events (SSE) for real-time outputs.
• Simple CLI: Start the server/list models with terminal commands.
• System Tray Integration: Background server management with tray icon.
• Ollama Compatibility: Works with locally hosted Ollama models.

Installation

1. Requirements:

• Python 3.9+
• Ollama (for Ollama integration)

2. Install dependencies:

pip install fastapi uvicorn pydantic python-dotenv google-generativeai cohere requests pystray pillow groq openai

Configuration

Create a .env file with your API keys:

GROQ_API_KEY="your_groq_key"
GEMINI_API_KEY="your_gemini_key"
CO_API_KEY="your_cohere_key"
DEEPSEEK_API_KEY="your_deepseek_key"

Usage

Start the API Server

python eragAPI.py serve

Add --tray to enable system tray icon.

List Available Models

# List all providers' models
python eragAPI.py model list

# Filter by provider
python eragAPI.py model list --api groq

API Endpoints

Chat Completion

POST /api/chat

{
  "model": "groq-mixtral-8x7b-32768",
  "messages": [{"role": "user", "content": "Hello!"}],
  "stream": false,
  "temperature": 0.7
}

Text Generation

POST /api/generate

{
  "model": "gemini-gemini-pro",
  "prompt": "Write a poem about AI",
  "stream": true,
  "temperature": 0.5
}

Example Requests

Basic Chat (cURL)

curl -X POST http://localhost:11436/api/chat \
-H "Content-Type: application/json" \
-d '{
  "model": "groq-mixtral-8x7b-32768",
  "messages": [{"role": "user", "content": "Explain quantum computing"}]
}'

Streaming Response (Python)

import requests

response = requests.post(
    "http://localhost:11436/api/generate",
    json={
        "model": "ollama-llama2",
        "prompt": "Tell me a joke",
        "stream": True
    },
    stream=True
)

for line in response.iter_lines():
    if line:
        print(line.decode('utf-8').replace('data: ', ''))

Model Naming Convention

Format: {provider}-{model_name}

Examples:

• groq-mixtral-8x7b-32768
• ollama-llama2-uncensored
• deepseek-deepseek-chat

Notes

• Ollama models must be pulled first (e.g., ollama pull llama2).
• Default models are used if no specific model is provided.
• Temperature range: 0.0 (deterministic) to 1.0 (creative).
• Stream responses for long-generation tasks to avoid timeouts.

License

MIT - Add appropriate license file.