feat: Add Mistral AI model support through OpenAIChatCompletionClient

python/packages/autogen-ext/src/autogen_ext/models/openai/_model_info.py

@@ -46,6 +46,15 @@
     "llama-3.3-70b": "Llama-3.3-70B-Instruct",
     "llama-4-scout": "Llama-4-Scout-17B-16E-Instruct-FP8",
     "llama-4-maverick": "Llama-4-Maverick-17B-128E-Instruct-FP8",
+    # Mistral AI models
+    "mistral-large-latest": "mistral-large-2411",
+    "mistral-medium-latest": "mistral-medium-2505",
+    "mistral-small-latest": "mistral-small-2503",
+    "codestral-latest": "codestral-2501",
+    "pixtral-large-latest": "pixtral-large-2411",
+    "ministral-3b-latest": "ministral-3b-2410",
+    "ministral-8b-latest": "ministral-8b-2410",
+    "open-mistral-nemo": "open-mistral-nemo-2407",
 }
 
 _MODEL_INFO: Dict[str, ModelInfo] = {
@@ -441,6 +450,87 @@
         "structured_output": True,
         "multiple_system_messages": True,
     },
+    # Mistral AI models
+    "mistral-large-2411": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MISTRAL,
+        "structured_output": True,
+        "multiple_system_messages": True,
+    },
+    "mistral-medium-2505": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MISTRAL,
+        "structured_output": True,
+        "multiple_system_messages": True,
+    },
+    "mistral-small-2503": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MISTRAL,
+        "structured_output": True,
+        "multiple_system_messages": True,
+    },
+    "codestral-2501": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.CODESRAL,
+        "structured_output": True,
+        "multiple_system_messages": True,
+    },
+    "open-codestral-mamba": {
+        "vision": False,
+        "function_calling": False,
+        "json_output": True,
+        "family": ModelFamily.OPEN_CODESRAL_MAMBA,
+        "structured_output": False,
+        "multiple_system_messages": True,
+    },
+    "pixtral-large-2411": {
+        "vision": True,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.PIXTRAL,
+        "structured_output": True,
+        "multiple_system_messages": True,
+    },
+    "pixtral-12b-2409": {
+        "vision": True,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.PIXTRAL,
+        "structured_output": False,
+        "multiple_system_messages": True,
+    },
+    "ministral-3b-2410": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MINISTRAL,
+        "structured_output": False,
+        "multiple_system_messages": True,
+    },
+    "ministral-8b-2410": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MINISTRAL,
+        "structured_output": False,
+        "multiple_system_messages": True,
+    },
+    "open-mistral-nemo-2407": {
+        "vision": False,
+        "function_calling": True,
+        "json_output": True,
+        "family": ModelFamily.MISTRAL,
+        "structured_output": False,
+        "multiple_system_messages": True,
+    },
 }
 
 _MODEL_TOKEN_LIMITS: Dict[str, int] = {
@@ -491,11 +581,38 @@
     "Llama-3.3-70B-Instruct": 128000,
     "Llama-4-Scout-17B-16E-Instruct-FP8": 128000,
     "Llama-4-Maverick-17B-128E-Instruct-FP8": 128000,
+    # Mistral AI models
+    "mistral-large-2411": 131072,
+    "mistral-medium-2505": 131072,
+    "mistral-small-2503": 131072,
+    "codestral-2501": 262144,
+    "open-codestral-mamba": 262144,
+    "pixtral-large-2411": 131072,
+    "pixtral-12b-2409": 131072,
+    "ministral-3b-2410": 131072,
+    "ministral-8b-2410": 131072,
+    "open-mistral-nemo-2407": 131072,
 }
 
 GEMINI_OPENAI_BASE_URL = "https://generativelanguage.googleapis.com/v1beta/openai/"
 ANTHROPIC_OPENAI_BASE_URL = "https://api.anthropic.com/v1/"
 LLAMA_API_BASE_URL = "https://api.llama.com/compat/v1/"
+MISTRAL_API_BASE_URL = "https://api.mistral.ai/v1/"
+
+# Mistral model name prefixes used for auto-detection.
+_MISTRAL_MODEL_PREFIXES = (
+    "mistral-",
+    "codestral-",
+    "open-codestral-",
+    "pixtral-",
+    "ministral-",
+    "open-mistral-",
+)
+
+
+def _is_mistral_model(model: str) -> bool:
+    """Check if the model name matches a Mistral AI model prefix."""
+    return any(model.startswith(prefix) for prefix in _MISTRAL_MODEL_PREFIXES)
 
 
 def resolve_model(model: str) -> str:

python/packages/autogen-ext/src/autogen_ext/models/openai/_openai_client.py

@@ -1480,6 +1480,15 @@ def __init__(self, **kwargs: Unpack[OpenAIClientConfiguration]):
                 copied_args["base_url"] = _model_info.LLAMA_API_BASE_URL
             if "api_key" not in copied_args and "LLAMA_API_KEY" in os.environ:
                 copied_args["api_key"] = os.environ["LLAMA_API_KEY"]
+        if _model_info._is_mistral_model(copied_args["model"]):
+            if "base_url" not in copied_args:
+                copied_args["base_url"] = _model_info.MISTRAL_API_BASE_URL
+            if "api_key" not in copied_args and "MISTRAL_API_KEY" in os.environ:
+                copied_args["api_key"] = os.environ["MISTRAL_API_KEY"]
+            # Mistral API rejects the 'name' field in messages (HTTP 422),
+            # so disable it by default unless the user explicitly set it.
+            if "include_name_in_message" not in kwargs:
+                include_name_in_message = False
 
         client = _openai_client_from_config(copied_args)
         create_args = _create_args_from_config(copied_args)

python/packages/autogen-ext/tests/models/test_openai_model_client.py

@@ -24,7 +24,13 @@
 from autogen_core.models._model_client import ModelFamily
 from autogen_core.tools import BaseTool, FunctionTool
 from autogen_ext.models.openai import AzureOpenAIChatCompletionClient, OpenAIChatCompletionClient
-from autogen_ext.models.openai._model_info import resolve_model
+from autogen_ext.models.openai._model_info import (
+    MISTRAL_API_BASE_URL,
+    _is_mistral_model,
+    get_info,
+    get_token_limit,
+    resolve_model,
+)
 from autogen_ext.models.openai._openai_client import (
     BaseOpenAIChatCompletionClient,
     calculate_vision_tokens,
@@ -191,6 +197,150 @@ async def test_openai_chat_completion_client_with_gemini_model() -> None:
     assert client
 
 
+@pytest.mark.asyncio
+async def test_openai_chat_completion_client_with_mistral_model() -> None:
+    """Test that Mistral models can be instantiated via OpenAIChatCompletionClient."""
+    client = OpenAIChatCompletionClient(model="mistral-small-latest", api_key="api_key")
+    assert client
+
+
+@pytest.mark.asyncio
+async def test_openai_chat_completion_client_with_mistral_model_variants() -> None:
+    """Test that various Mistral model variants are recognized."""
+    mistral_models = [
+        "mistral-large-latest",
+        "mistral-medium-latest",
+        "mistral-small-latest",
+        "codestral-latest",
+        "pixtral-large-latest",
+        "ministral-3b-latest",
+        "ministral-8b-latest",
+        "open-mistral-nemo",
+    ]
+    for model_name in mistral_models:
+        client = OpenAIChatCompletionClient(model=model_name, api_key="api_key")
+        assert client, f"Failed to create client for model: {model_name}"
+
+
+def test_is_mistral_model() -> None:
+    """Test the _is_mistral_model helper function."""
+    # Positive cases
+    assert _is_mistral_model("mistral-large-latest")
+    assert _is_mistral_model("mistral-small-2503")
+    assert _is_mistral_model("codestral-latest")
+    assert _is_mistral_model("codestral-2501")
+    assert _is_mistral_model("open-codestral-mamba")
+    assert _is_mistral_model("pixtral-large-2411")
+    assert _is_mistral_model("pixtral-12b-2409")
+    assert _is_mistral_model("ministral-3b-2410")
+    assert _is_mistral_model("ministral-8b-latest")
+    assert _is_mistral_model("open-mistral-nemo")
+    assert _is_mistral_model("open-mistral-nemo-2407")
+    # Negative cases
+    assert not _is_mistral_model("gpt-4o")
+    assert not _is_mistral_model("gemini-1.5-flash")
+    assert not _is_mistral_model("claude-3-5-sonnet-20241022")
+    assert not _is_mistral_model("Llama-4-Scout-17B-16E-Instruct-FP8")
+
+
+def test_mistral_model_info() -> None:
+    """Test that Mistral model info is correctly registered."""
+    # mistral-large
+    info = get_info("mistral-large-latest")
+    assert info["family"] == ModelFamily.MISTRAL
+    assert info["function_calling"] is True
+    assert info["json_output"] is True
+    assert info["vision"] is False
+    assert info["structured_output"] is True
+
+    # pixtral (vision model)
+    info = get_info("pixtral-large-latest")
+    assert info["family"] == ModelFamily.PIXTRAL
+    assert info["vision"] is True
+    assert info["function_calling"] is True
+
+    # codestral
+    info = get_info("codestral-latest")
+    assert info["family"] == ModelFamily.CODESRAL
+    assert info["function_calling"] is True
+
+    # open-codestral-mamba (no function calling)
+    info = get_info("open-codestral-mamba")
+    assert info["family"] == ModelFamily.OPEN_CODESRAL_MAMBA
+    assert info["function_calling"] is False
+
+    # ministral
+    info = get_info("ministral-8b-latest")
+    assert info["family"] == ModelFamily.MINISTRAL
+    assert info["function_calling"] is True
+
+    # open-mistral-nemo
+    info = get_info("open-mistral-nemo")
+    assert info["family"] == ModelFamily.MISTRAL
+    assert info["function_calling"] is True
+
+
+def test_mistral_model_token_limits() -> None:
+    """Test that Mistral model token limits are correctly registered."""
+    assert get_token_limit("mistral-large-latest") == 131072
+    assert get_token_limit("mistral-small-latest") == 131072
+    assert get_token_limit("codestral-latest") == 262144
+    assert get_token_limit("open-codestral-mamba") == 262144
+    assert get_token_limit("pixtral-large-latest") == 131072
+    assert get_token_limit("ministral-3b-latest") == 131072
+    assert get_token_limit("open-mistral-nemo") == 131072
+
+
+def test_mistral_resolve_model() -> None:
+    """Test that Mistral model pointers resolve correctly."""
+    assert resolve_model("mistral-large-latest") == "mistral-large-2411"
+    assert resolve_model("mistral-medium-latest") == "mistral-medium-2505"
+    assert resolve_model("mistral-small-latest") == "mistral-small-2503"
+    assert resolve_model("codestral-latest") == "codestral-2501"
+    assert resolve_model("pixtral-large-latest") == "pixtral-large-2411"
+    assert resolve_model("ministral-3b-latest") == "ministral-3b-2410"
+    assert resolve_model("ministral-8b-latest") == "ministral-8b-2410"
+    assert resolve_model("open-mistral-nemo") == "open-mistral-nemo-2407"
+    # Versioned models should resolve to themselves
+    assert resolve_model("mistral-large-2411") == "mistral-large-2411"
+    assert resolve_model("codestral-2501") == "codestral-2501"
+
+
+@pytest.mark.asyncio
+async def test_mistral_auto_base_url() -> None:
+    """Test that Mistral models automatically get the correct base_url."""
+    client = OpenAIChatCompletionClient(model="mistral-small-latest", api_key="test_key")
+    assert client._raw_config.get("base_url") == MISTRAL_API_BASE_URL
+
+
+@pytest.mark.asyncio
+async def test_mistral_auto_api_key(monkeypatch: pytest.MonkeyPatch) -> None:
+    """Test that MISTRAL_API_KEY env var is auto-detected for Mistral models."""
+    monkeypatch.setenv("MISTRAL_API_KEY", "test_mistral_key")
+    client = OpenAIChatCompletionClient(model="mistral-small-latest")
+    assert client._raw_config.get("api_key") == "test_mistral_key"
+
+
+@pytest.mark.asyncio
+async def test_mistral_include_name_disabled_by_default() -> None:
+    """Test that include_name_in_message defaults to False for Mistral models.
+
+    This addresses the HTTP 422 error from Mistral API when the 'name' field
+    is included in messages (see issue #6147).
+    """
+    client = OpenAIChatCompletionClient(model="mistral-small-latest", api_key="test_key")
+    assert client._include_name_in_message is False
+
+
+@pytest.mark.asyncio
+async def test_mistral_include_name_can_be_overridden() -> None:
+    """Test that include_name_in_message can be explicitly set for Mistral models."""
+    client = OpenAIChatCompletionClient(
+        model="mistral-small-latest", api_key="test_key", include_name_in_message=True
+    )
+    assert client._include_name_in_message is True
+
+
 @pytest.mark.asyncio
 async def test_openai_chat_completion_client_serialization() -> None:
     client = OpenAIChatCompletionClient(model="gpt-4.1-nano", api_key="sk-password")