Official Python SDK for M3 Chat – Stream and fetch AI-generated responses from models like LLaMA, Mistral, Gemma, and more.
- 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
pip install m3chat_sdkor
python3 -m pip install m3chat_sdkfrom m3chat_sdk import M3ChatClient
client = M3ChatClient(stream=False)
response = client.get_response(model="mistral", content="Hello, how are you?")
print("Response:", response)from m3chat_sdk import M3ChatClient
client = M3ChatClient(stream=True) # or Falseresponse = client.get_response(model="llama3:8b", content="Explain quantum physics.")
print(response)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}")[
"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",
]Run the client test with:
python3 tests/test_client.pyApache License 2.0 — see LICENSE
Issues and PRs welcome! https://github.com/m3-chat/python-sdk
