UniAI

A Universal AI Interface Layer for Python

Seamlessly connect to and switch between LLM providers with unified context management and streaming support.

✨ Features

📦 Installation

pip install uniai

From source:

git clone https://github.com/Archie818/uniai.git
cd uniai
pip install -e .

🚀 Quick Start

from uniai import UniAI

# Initialize with your preferred provider
bot = UniAI(provider="openai", api_key="sk-...", model="gpt-4o-mini")

# Simple chat
response = bot.chat("Hello, who are you?")
print(response)

# Multi-turn conversation (context is automatically managed)
response = bot.chat("What can you help me with?")
response = bot.chat("Tell me more about the first thing you mentioned")

🌊 Streaming

for chunk in bot.stream("Tell me a story about a brave knight"):
    print(chunk, end="", flush=True)

🔀 Switch Providers

# Start with OpenAI
bot = UniAI(provider="openai", api_key="sk-openai-key", model="gpt-4o-mini")
bot.chat("Hello!")

# Switch to DeepSeek (preserves history)
bot.switch_provider(provider="deepseek", api_key="sk-deepseek-key", model="deepseek-chat")
bot.chat("Continue our conversation")

# Switch and clear history
bot.switch_provider(provider="openai", api_key="sk-...", model="gpt-4o", keep_history=False)

📋 Supported Providers

⚙️ Configuration

bot = UniAI(
    provider="openai",
    api_key="sk-...",
    model="gpt-4o-mini",
    system_prompt="You are a helpful assistant.",  # System instructions
    temperature=0.7,         # Randomness (0.0-2.0)
    max_tokens=2048,         # Max response length
    max_history=20,          # Conversation history limit
    timeout=60.0,            # Request timeout (seconds)
    max_retries=3,           # Retry attempts
)

📖 Advanced Usage

Full Response Object

response = bot.chat_with_response("What is Python?")

print(f"Content: {response.content}")
print(f"Model: {response.model}")
print(f"Tokens: {response.usage.total_tokens}")
print(f"Finish reason: {response.finish_reason}")

History Management

history = bot.get_history()      # Get conversation history
bot.clear_history()              # Clear history
print(len(bot.memory))           # Messages count

Custom Providers

from uniai.core.base import BaseProvider
from uniai.providers import register_provider

class MyProvider(BaseProvider):
    name = "custom"

    def _init_client(self):
        pass

    def chat(self, messages):
        # Your implementation
        pass

    def stream_chat(self, messages):
        # Your implementation
        pass

register_provider("custom", MyProvider)
bot = UniAI(provider="custom", api_key="...", model="...")

🛡️ Error Handling

from uniai.exceptions import AuthenticationError, RateLimitError, APIError

try:
    response = bot.chat("Hello!")
except AuthenticationError:
    print("Invalid API key")
except RateLimitError:
    print("Rate limit exceeded")
except APIError as e:
    print(f"API error: {e}")

🏗️ Architecture

uniai/
├── client.py            # Main UniAI class
├── exceptions.py        # Custom exceptions
├── core/
│   ├── base.py          # Abstract BaseProvider
│   ├── config.py        # Pydantic config models
│   └── types.py         # Type definitions
├── context/
│   └── memory.py        # Conversation memory
└── providers/
    ├── openai.py        # OpenAI provider
    ├── deepseek.py      # DeepSeek provider
    ├── gemini.py       # Gemini provider

🗺️ Roadmap

• v0.1 - Core functionality (Unified interface, OpenAI & DeepSeek, Context management, Streaming)
• v0.1.3 - Gemini support
• v0.2 - Extended providers (Claude, Ollama)
• v0.3 - Advanced features (Async support, Function calling, Token counting)
• v1.0 - Agent framework (Multi-model collaboration, Tool integration)

🤝 Contributing

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

📄 License

MIT License © 2024 UniAI Contributors