Python SDK for M3 Chat

Official Python SDK for M3 Chat – Stream and fetch AI-generated responses from models like LLaMA, Mistral, Gemma, and more.

Features

• Simple interface to call M3 Chat models
• Stream or fetch full responses
• Batch support (non-streaming)
• Model validation
• Typed API and clean structure
• CLI-friendly

Installation

pip install m3chat_sdk

or

python3 -m pip install m3chat_sdk

Quick Start

from m3chat_sdk import M3ChatClient

client = M3ChatClient(stream=False)
response = client.get_response(model="mistral", content="Hello, how are you?")
print("Response:", response)

Usage

Initialize Client

from m3chat_sdk import M3ChatClient

client = M3ChatClient(stream=True) # or False

Get a Response

response = client.get_response(model="llama3:8b", content="Explain quantum physics.")
print(response)

Batch Requests (Non-Streaming)

prompts = [
"What is the capital of Germany?",
"Tell me a joke.",
"What's 12 * 8?",
]

responses = client.batch_requests(messages=prompts, model="gemma")
for i, res in enumerate(responses):
print(f"Response {i + 1}: {res}")

Supported Models

[
"llama3:8b",
"llama2-uncensored",
"gemma3",
"gemma",
"phi3:mini",
"mistral",
"gemma:2b",
"gemma:7b",
"qwen:7b",
"qwen2.5-coder",
"qwen3",
"deepseek-coder:6.7b",
"deepseek-v2:16b",
"dolphin-mistral:7b",
"dolphin3",
"starcoder2:7b",
"magistral",
"devstral",
]

Testing

Run the client test with:

python3 tests/test_client.py

License

Apache License 2.0 — see LICENSE

Contributing

Issues and PRs welcome! https://github.com/m3-chat/python-sdk