feat(groq): support lc image types

libs/partners/groq/langchain_groq/chat_models.py

@@ -40,12 +40,16 @@
     ToolCall,
     ToolMessage,
     ToolMessageChunk,
+    is_data_content_block,
 )
 from langchain_core.messages.ai import (
     InputTokenDetails,
     OutputTokenDetails,
     UsageMetadata,
 )
+from langchain_core.messages.block_translators.openai import (
+    convert_to_openai_data_block,
+)
 from langchain_core.output_parsers import JsonOutputParser, PydanticOutputParser
 from langchain_core.output_parsers.base import OutputParserLike
 from langchain_core.output_parsers.openai_tools import (
@@ -1263,6 +1267,29 @@ def _is_pydantic_class(obj: Any) -> bool:
 #
 # Type conversion helpers
 #
+def _format_message_content(content: Any) -> Any:
+    """Format message content for Groq API.
+
+    Converts LangChain image content blocks to Groq's expected image_url format.
+
+    Args:
+        content: The message content (string or list of content blocks).
+
+    Returns:
+        Formatted content suitable for Groq API.
+    """
+    if content and isinstance(content, list):
+        formatted: list = []
+        for block in content:
+            # Handle LangChain standard data content blocks (image, audio, file)
+            if isinstance(block, dict) and is_data_content_block(block):
+                formatted.append(convert_to_openai_data_block(block))
+            else:
+                formatted.append(block)
+        return formatted
+    return content
+
+
 def _convert_message_to_dict(message: BaseMessage) -> dict:
     """Convert a LangChain message to a dictionary.
 
@@ -1277,7 +1304,10 @@ def _convert_message_to_dict(message: BaseMessage) -> dict:
     if isinstance(message, ChatMessage):
         message_dict = {"role": message.role, "content": message.content}
     elif isinstance(message, HumanMessage):
-        message_dict = {"role": "user", "content": message.content}
+        message_dict = {
+            "role": "user",
+            "content": _format_message_content(message.content),
+        }
     elif isinstance(message, AIMessage):
         # Translate v1 content
         if message.response_metadata.get("output_version") == "v1":

libs/partners/groq/tests/integration_tests/test_standard.py

@@ -5,7 +5,6 @@
 import pytest
 from langchain_core.language_models import BaseChatModel
 from langchain_core.rate_limiters import InMemoryRateLimiter
-from langchain_core.tools import BaseTool
 from langchain_tests.integration_tests import (
     ChatModelIntegrationTests,
 )
@@ -24,12 +23,6 @@ def chat_model_class(self) -> type[BaseChatModel]:
     def chat_model_params(self) -> dict:
         return {"model": "llama-3.3-70b-versatile", "rate_limiter": rate_limiter}
 
-    @pytest.mark.xfail(reason="Not yet implemented.")
-    def test_tool_message_histories_list_content(
-        self, model: BaseChatModel, my_adder_tool: BaseTool
-    ) -> None:
-        super().test_tool_message_histories_list_content(model, my_adder_tool)
-
     @pytest.mark.xfail(
         reason="Groq models have inconsistent tool calling performance. See: "
         "https://github.com/langchain-ai/langchain/discussions/19990"

libs/partners/groq/tests/unit_tests/test_chat_models.py

@@ -22,6 +22,7 @@
     _convert_chunk_to_message_chunk,
     _convert_dict_to_message,
     _create_usage_metadata,
+    _format_message_content,
 )
 
 if "GROQ_API_KEY" not in os.environ:
@@ -945,3 +946,64 @@ def test_combine_llm_outputs_with_missing_details() -> None:
 def test_profile() -> None:
     model = ChatGroq(model="openai/gpt-oss-20b")
     assert model.profile
+
+
+def test_format_message_content_string() -> None:
+    """Test that string content is passed through unchanged."""
+    content = "hello"
+    assert content == _format_message_content(content)
+
+
+def test_format_message_content_none() -> None:
+    """Test that None content is passed through unchanged."""
+    content = None
+    assert content == _format_message_content(content)
+
+
+def test_format_message_content_empty_list() -> None:
+    """Test that empty list is passed through unchanged."""
+    content: list = []
+    assert content == _format_message_content(content)
+
+
+def test_format_message_content_text_and_image_url() -> None:
+    """Test that existing image_url format is passed through unchanged."""
+    content = [
+        {"type": "text", "text": "What is in this image?"},
+        {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}},
+    ]
+    assert content == _format_message_content(content)
+
+
+def test_format_message_content_langchain_image_base64() -> None:
+    """Test that LangChain image blocks with base64 are converted."""
+    content = {"type": "image", "base64": "<base64 data>", "mime_type": "image/png"}
+    expected = [
+        {
+            "type": "image_url",
+            "image_url": {"url": "data:image/png;base64,<base64 data>"},
+        }
+    ]
+    assert expected == _format_message_content([content])
+
+
+def test_format_message_content_langchain_image_url() -> None:
+    """Test that LangChain image blocks with URL are converted."""
+    content = {"type": "image", "url": "https://example.com/image.jpg"}
+    expected = [
+        {"type": "image_url", "image_url": {"url": "https://example.com/image.jpg"}}
+    ]
+    assert expected == _format_message_content([content])
+
+
+def test_format_message_content_mixed() -> None:
+    """Test that mixed content with text and image is handled correctly."""
+    content = [
+        {"type": "text", "text": "Describe this image"},
+        {"type": "image", "base64": "<data>", "mime_type": "image/png"},
+    ]
+    expected = [
+        {"type": "text", "text": "Describe this image"},
+        {"type": "image_url", "image_url": {"url": "data:image/png;base64,<data>"}},
+    ]
+    assert expected == _format_message_content(content)