litechat 0.0.54

automated huggingchat openai style fastapi inference

LiteChat 🚀

LiteChat is a lightweight, OpenAI-compatible interface for running local LLM inference servers. It provides seamless integration with various open-source models while maintaining OpenAI-style API compatibility.

Features ✨

• 🔄 OpenAI API compatibility
• 🌐 Web search integration
• 💬 Conversation memory
• 🔄 Streaming responses
• 🛠️ Easy integration with HuggingFace models
• 📦 Compatible with both litellm and OpenAI clients
• 🎯 Type-safe model selection

Installation 🛠️

pip install litechat playwright
playwright install

Available Models 🤖

LiteChat supports the following models:

• Qwen/Qwen2.5-Coder-32B-Instruct: Specialized coding model
• Qwen/Qwen2.5-72B-Instruct: Large general-purpose model
• meta-llama/Llama-3.3-70B-Instruct: Latest Llama 3 model
• CohereForAI/c4ai-command-r-plus-08-2024: Cohere's command model
• Qwen/QwQ-32B-Preview: Preview version of QwQ
• nvidia/Llama-3.1-Nemotron-70B-Instruct-HF: NVIDIA's Nemotron model
• meta-llama/Llama-3.2-11B-Vision-Instruct: Vision-capable Llama model
• NousResearch/Hermes-3-Llama-3.1-8B: Lightweight Hermes model
• mistralai/Mistral-Nemo-Instruct-2407: Mistral's instruction model
• microsoft/Phi-3.5-mini-instruct: Microsoft's compact Phi model

Model Selection Helpers 🎯

LiteChat provides helper functions for type-safe model selection:

from litechat import litechat_model, litellm_model

# For use with LiteChat native client
model = litechat_model("Qwen/Qwen2.5-72B-Instruct")

# For use with LiteLLM
model = litellm_model("Qwen/Qwen2.5-72B-Instruct")  # Returns "openai/Qwen/Qwen2.5-72B-Instruct"

Quick Start 🚀

Starting the Server

You can start the LiteChat server in two ways:

1. Using the CLI:

litechat_server

2. Programmatically:

from litechat import litechat_server

if __name__ == "__main__":
    litechat_server(host="0.0.0.0", port=11437)

Using with OpenAI Client

import os
from openai import OpenAI

os.environ['OPENAI_BASE_URL'] = "http://localhost:11437/v1"
os.environ['OPENAI_API_KEY'] = "key123" # required, but not used

client = OpenAI()
response = client.chat.completions.create(
    model=litechat_model("NousResearch/Hermes-3-Llama-3.1-8B"),
    messages=[
        {"role": "system", "content": "You are a helpful assistant"},
        {"role": "user", "content": "What is the capital of France?"}
    ]
)
print(response.choices[0].message.content)

Using with LiteLLM

import os

from litellm import completion
from litechat import OPENAI_COMPATIBLE_BASE_URL, litellm_model

os.environ["OPENAI_API_KEY"] = "key123"

response = completion(
    model=litellm_model("NousResearch/Hermes-3-Llama-3.1-8B"),
    messages=[{"content": "Hello, how are you?", "role": "user"}],
    api_base=OPENAI_COMPATIBLE_BASE_URL
)
print(response)

Using LiteChat's Native Client

from litechat import completion, genai, pp_completion
from litechat import litechat_model

# Basic completion
response = completion(
    prompt="What is quantum computing?",
    model="nvidia/Llama-3.1-Nemotron-70B-Instruct-HF",
    web_search=True  # Enable web search
)

# Stream with pretty printing
pp_completion(
    prompt="Explain the theory of relativity",
    model="Qwen/Qwen2.5-72B-Instruct",
    conversation_id="physics_chat"  # Enable conversation memory
)

# Get direct response
result = genai(
    prompt="Write a poem about spring",
    model="meta-llama/Llama-3.3-70B-Instruct",
    system_prompt="You are a creative poet"
)

Advanced Features 🔧

Web Search Integration

Enable web search to get up-to-date information:

response = completion(
    prompt="What are the latest developments in AI?",
    web_search=True
)

Conversation Memory

Maintain context across multiple interactions:

response = completion(
    prompt="Tell me more about that",
    conversation_id="unique_conversation_id"
)

Streaming Responses

Get token-by-token streaming:

for chunk in completion(
    prompt="Write a long story",
    stream=True
):
    print(chunk.choices[0].delta.content, end="", flush=True)

API Reference 📚

LiteAI Client

from litechat import LiteAI, litechat_model

client = LiteAI(
    api_key="optional-key",  # Optional API key
    base_url="http://localhost:11437",  # Server URL
    system_prompt="You are a helpful assistant",  # Default system prompt
    web_search=False,  # Enable/disable web search by default
    model=litechat_model("nvidia/Llama-3.1-Nemotron-70B-Instruct-HF")  # Default model
)

Completion Function Parameters

• messages: List of conversation messages or direct prompt string
• model: HuggingFace model identifier (use litechat_model() for type safety)
• system_prompt: System instruction for the model
• temperature: Control randomness (0.0 to 1.0)
• stream: Enable streaming responses
• web_search: Enable web search
• conversation_id: Enable conversation memory
• max_tokens: Maximum tokens in response
• tools: List of available tools/functions

Contributing 🤝

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

License 📄

This project is licensed under the MIT License - see the LICENSE file for details.

Support 💬

For support, please open an issue on the GitHub repository or reach out to the maintainers.