feat: Turn Mistral to use llm_http_handler (#12245)

* Enhance Mistral API: Add support for parallel tool calls and refine name handling in tool messages. Plus, introduce a new test for parallel tool calls in the Mistral model.

* tests

* make mypy happy

* Refine name handling in Mistral chat transformation: clarify conditions for removing the 'name' field based on message role and content.

* refactor: streamline Mistral integration by removing deprecated references and adding a new handler

- Removed "mistral" from the list of compatible providers in constants.
- Updated the completion function in main.py to utilize the new Mistral handler.
- Deleted outdated Mistral chat and embedding files.
- Introduced a new handler for Mistral chat completions, implementing the llm_http_handler pattern.
- Added integration tests for the Mistral handler to ensure proper API base and key handling.

* lint

* fix: remove unneeded handler object

* add tests

* Addres PR comments

litellm/__init__.py

@@ -1046,7 +1046,7 @@ from .llms.groq.chat.transformation import GroqChatConfig
 from .llms.voyage.embedding.transformation import VoyageEmbeddingConfig
 from .llms.infinity.embedding.transformation import InfinityEmbeddingConfig
 from .llms.azure_ai.chat.transformation import AzureAIStudioConfig
-from .llms.mistral.mistral_chat_transformation import MistralConfig
+from .llms.mistral.chat.transformation import MistralConfig
 from .llms.openai.responses.transformation import OpenAIResponsesAPIConfig
 from .llms.azure.responses.transformation import AzureOpenAIResponsesAPIConfig
 from .llms.openai.chat.o_series_transformation import (

litellm/constants.py

@@ -397,7 +397,6 @@ openai_compatible_endpoints: List = [
 
 openai_compatible_providers: List = [
     "anyscale",
-    "mistral",
     "groq",
     "nvidia_nim",
     "cerebras",

litellm/llms/mistral/chat.py

@@ -1,5 +0,0 @@
-"""
-Calls handled in openai/
-
-as mistral is an openai-compatible endpoint.
-"""

litellm/llms/mistral/chat/transformation.py

@@ -174,7 +174,7 @@ class MistralConfig(OpenAIGPTConfig):
 
     def _get_openai_compatible_provider_info(
         self, api_base: Optional[str], api_key: Optional[str]
-    ) -> Tuple[Optional[str], Optional[str]]:
+    ) -> Tuple[str, Optional[str]]:
         # mistral is openai compatible, we just need to set this to custom_openai and have the api_base be https://api.mistral.ai
         api_base = (
             api_base

litellm/llms/mistral/embedding.py

@@ -1,5 +1,4 @@
 """
 Calls handled in openai/
-
 as mistral is an openai-compatible endpoint.
-"""
+"""

litellm/main.py

@@ -1841,7 +1841,6 @@ def completion(  # type: ignore # noqa: PLR0915
             or custom_llm_provider == "sambanova"
             or custom_llm_provider == "volcengine"
             or custom_llm_provider == "anyscale"
-            or custom_llm_provider == "mistral"
             or custom_llm_provider == "openai"
             or custom_llm_provider == "together_ai"
             or custom_llm_provider == "nebius"
@@ -1930,6 +1929,33 @@ def completion(  # type: ignore # noqa: PLR0915
                     additional_args={"headers": headers},
                 )
 
+        elif custom_llm_provider == "mistral":
+            api_key = api_key or litellm.api_key or get_secret("MISTRAL_API_KEY")
+            api_base = (
+                api_base
+                or litellm.api_base
+                or get_secret("MISTRAL_API_BASE")
+                or "https://api.mistral.ai/v1"
+            )
+            
+            response = base_llm_http_handler.completion(
+                model=model,
+                messages=messages,
+                api_base=api_base,
+                custom_llm_provider=custom_llm_provider,
+                model_response=model_response,
+                encoding=encoding,
+                logging_obj=logging,
+                optional_params=optional_params,
+                timeout=timeout,
+                litellm_params=litellm_params,
+                acompletion=acompletion,
+                stream=stream,
+                api_key=api_key,
+                headers=headers,
+                client=client,
+                provider_config=provider_config,
+            )
         elif (
             "replicate" in model
             or custom_llm_provider == "replicate"

tests/test_litellm/llms/mistral/test_mistral_chat_transformation.py

@@ -8,7 +8,7 @@ sys.path.insert(
     0, os.path.abspath("../..")
 )  # Adds the parent directory to the system path
 
-from litellm.llms.mistral.mistral_chat_transformation import MistralConfig
+from litellm.llms.mistral.chat.transformation import MistralConfig
 
 
 @pytest.mark.asyncio

tests/test_litellm/llms/mistral/test_mistral_completion.py

@@ -0,0 +1,160 @@
+import pytest
+import litellm
+
+
+@pytest.fixture(autouse=True)
+def add_mistral_api_key_to_env(monkeypatch):
+    """Add Mistral API key to environment for testing."""
+    monkeypatch.setenv("MISTRAL_API_KEY", "fake-mistral-api-key-12345")
+
+
+@pytest.fixture
+def mistral_api_response():
+    """Mock response data for Mistral API calls."""
+    return {
+        "id": "chatcmpl-mistral-123",
+        "object": "chat.completion",
+        "created": 1677652288,
+        "model": "mistral-medium-latest",
+        "choices": [
+            {
+                "index": 0,
+                "message": {
+                    "role": "assistant",
+                    "content": "Hello from Mistral! How can I help you today?",
+                },
+                "finish_reason": "stop",
+            }
+        ],
+        "usage": {"prompt_tokens": 10, "completion_tokens": 15, "total_tokens": 25},
+    }
+
+
+@pytest.fixture
+def mistral_api_response_with_empty_content():
+    """Mock response data for Mistral API calls with empty content that should be converted to None."""
+    return {
+        "id": "chatcmpl-mistral-123",
+        "object": "chat.completion",
+        "created": 1677652288,
+        "model": "mistral-medium-latest",
+        "choices": [
+            {
+                "index": 0,
+                "message": {
+                    "role": "assistant",
+                    "content": "",  # Empty string that should be converted to None
+                },
+                "finish_reason": "stop",
+            }
+        ],
+        "usage": {"prompt_tokens": 10, "completion_tokens": 0, "total_tokens": 10},
+    }
+
+
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_mistral_basic_completion(sync_mode, respx_mock, mistral_api_response):
+    """Test basic Mistral completion functionality."""
+    litellm.disable_aiohttp_transport = True
+    
+    model = "mistral/mistral-medium-latest"
+    messages = [{"role": "user", "content": "Hello, how are you?"}]
+    
+    # Mock the Mistral API endpoint
+    respx_mock.post("https://api.mistral.ai/v1/chat/completions").respond(
+        json=mistral_api_response
+    )
+    
+    if sync_mode:
+        response = litellm.completion(model=model, messages=messages)
+    else:
+        response = await litellm.acompletion(model=model, messages=messages)
+    
+    # Verify response
+    assert response.choices[0].message.content == "Hello from Mistral! How can I help you today?"
+    assert response.model == "mistral-medium-latest"
+    assert response.usage.total_tokens == 25
+
+
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_mistral_transform_response_empty_content_conversion(sync_mode, respx_mock, mistral_api_response_with_empty_content):
+    """
+    Test that Mistral's transform_response method is being called by verifying
+    the specific behavior of converting empty string content to None.
+    
+    This test verifies that the _handle_empty_content_response method in
+    MistralConfig.transform_response is being applied.
+    """
+    litellm.disable_aiohttp_transport = True
+    
+    model = "mistral/mistral-medium-latest"
+    messages = [{"role": "user", "content": "Generate an empty response"}]
+    
+    # Mock the Mistral API endpoint with empty content
+    respx_mock.post("https://api.mistral.ai/v1/chat/completions").respond(
+        json=mistral_api_response_with_empty_content
+    )
+    
+    if sync_mode:
+        response = litellm.completion(model=model, messages=messages)
+    else:
+        response = await litellm.acompletion(model=model, messages=messages)
+    
+    # Verify that the transform_response method was called by checking that
+    # empty string content was converted to None (Mistral-specific behavior)
+    assert response.choices[0].message.content is None
+    assert response.model == "mistral-medium-latest"
+    assert response.usage.total_tokens == 10
+
+
+@pytest.mark.parametrize("sync_mode", [True, False])
+@pytest.mark.asyncio
+async def test_mistral_transform_request_name_field_removal(sync_mode, respx_mock, mistral_api_response):
+    """
+    Test that Mistral's transform_request method is being called by verifying
+    the specific behavior of removing the 'name' field from non-tool messages.
+    
+    This test verifies that the _handle_name_in_message method in
+    MistralConfig._transform_messages is being applied.
+    """
+    litellm.disable_aiohttp_transport = True
+    
+    model = "mistral/mistral-medium-latest"
+    # Include a message with 'name' field that should be removed for non-tool messages
+    messages = [
+        {"role": "user", "content": "Hello", "name": "should_be_removed"},
+        {"role": "assistant", "content": "Hi there!"},
+        {"role": "user", "content": "How are you?"}
+    ]
+    
+    # Mock the Mistral API endpoint
+    respx_mock.post("https://api.mistral.ai/v1/chat/completions").respond(
+        json=mistral_api_response
+    )
+    
+    if sync_mode:
+        response = litellm.completion(model=model, messages=messages)
+    else:
+        response = await litellm.acompletion(model=model, messages=messages)
+    
+    # Verify the response works (if transform_request wasn't called, the API would reject the request)
+    assert response.choices[0].message.content == "Hello from Mistral! How can I help you today?"
+    assert response.model == "mistral-medium-latest"
+    
+    # Verify that the request was made (if transform_request failed, this would fail)
+    assert len(respx_mock.calls) == 1
+    
+    # Get the actual request that was made
+    request = respx_mock.calls[0].request
+    import json
+    request_data = json.loads(request.content.decode('utf-8'))
+    
+    # Verify that the 'name' field was removed from the user message
+    # (Mistral API only supports 'name' in tool messages)
+    user_message = request_data["messages"][0]
+    assert user_message["role"] == "user"
+    assert user_message["content"] == "Hello"
+    assert "name" not in user_message  # The 'name' field should have been removed
+