xai[patch]: Fix 400 error for AIMessages with tool_calls but no content #34229

arrdel · 2025-12-06T05:15:18Z

xAI API requires all messages to have at least an empty content field. When using LangGraph agents with tool calling, AIMessage objects can have tool_calls but no content, causing 400 errors.

This fix ensures content is set to empty string instead of None for AIMessages with tool_calls.

…tead of None xAI's API is stricter than OpenAI's and requires all messages to have at least an empty content field. When using LangGraph agents with tool calling, AIMessages can have tool_calls but no content, which causes xAI to return a 400 error: 'Invalid request content: Each message must have at least one content element.' This fix overrides the message conversion logic in ChatXAI to ensure that AIMessages with tool_calls or function_call always have at least an empty string for content, rather than None (which is acceptable for OpenAI but not xAI). Changes: - Added _convert_message_to_dict_xai() function that wraps the OpenAI converter and ensures content is never None for AIMessages with tool calls - Overrode _get_request_payload() to use the xAI-specific converter - Added necessary imports from langchain_openai.chat_models.base Fixes langchain-ai#34140

codspeed-hq · 2025-12-06T05:19:31Z

CodSpeed Performance Report

Merging #34229 will not alter performance

_{Comparing arrdel:fix-chatxai-empty-content-34140 (35c1c1f) with master (80c3970)}

Summary

✅ 1 untouched
⏩ 33 skipped¹

33 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

Copilot

Pull request overview

This PR fixes a 400 error that occurs when using xAI's API with AIMessages that have tool_calls but no content. The xAI API differs from OpenAI in that it requires the content field to be an empty string rather than None for such messages.

Key Changes:

Introduces _convert_message_to_dict_xai function to wrap OpenAI's message converter and ensure content is empty string instead of None for AIMessages with tool_calls/function_call
Overrides _get_request_payload to use the xAI-specific message converter for the chat/completions API path

Copilot · 2025-12-06T05:20:04Z

libs/partners/xai/langchain_xai/chat_models.py

+        if self._use_responses_api(payload):
+            if self.use_previous_response_id:
+                last_messages, previous_response_id = _get_last_messages(messages)
+                payload_to_use = last_messages if previous_response_id else messages
+                if previous_response_id:
+                    payload["previous_response_id"] = previous_response_id
+                payload = _construct_responses_api_payload(payload_to_use, payload)
+            else:
+                payload = _construct_responses_api_payload(messages, payload)


The responses API path doesn't use the xAI-specific message converter. When _construct_responses_api_payload is called, it internally uses _construct_responses_api_input which calls _convert_message_to_dict from langchain_openai, not _convert_message_to_dict_xai. This means AIMessages with tool_calls will still have content=None instead of content="" when using the responses API, potentially causing the same 400 error this PR aims to fix.

Consider either:

Creating an xAI-specific version of _construct_responses_api_payload that uses _convert_message_to_dict_xai, or

Overriding _construct_responses_api_input to use the xAI converter, or

If xAI doesn't support the responses API, explicitly disable it by overriding _use_responses_api to return False

Copilot · 2025-12-06T05:20:04Z

libs/partners/xai/langchain_xai/chat_models.py

+def _convert_message_to_dict_xai(
+    message: BaseMessage,
+    api: Literal["chat/completions", "responses"] = "chat/completions",
+) -> dict:
+    """Convert a LangChain message to dictionary format expected by xAI.
+    
+    xAI's API requires that all messages have at least an empty content field,
+    unlike OpenAI which allows None for messages with tool_calls.
+    
+    Args:
+        message: The LangChain message to convert
+        api: The API format to use (default: "chat/completions")
+        
+    Returns:
+        Dictionary representation of the message compatible with xAI's API
+    """
+    message_dict = _openai_convert_message_to_dict(message, api=api)
+
+    # xAI requires content to be at least an empty string, not None
+    # This is especially important for AIMessages with tool_calls but no content
+    if isinstance(message, AIMessage) and message_dict.get("content") is None:
+        if "tool_calls" in message_dict or "function_call" in message_dict:
+            message_dict["content"] = ""
+
+    return message_dict


Missing test coverage for the new _convert_message_to_dict_xai function. There should be unit tests that verify:

AIMessage with tool_calls but no content gets content="" instead of content=None

AIMessage with function_call but no content gets content="" instead of content=None

AIMessage with both tool_calls and content preserves the content

Other message types are not affected

The existing tests in test_chat_models.py import _convert_message_to_dict from langchain_openai, but don't test the xAI-specific converter. Consider adding tests similar to the existing message conversion tests but using the xAI version.

Copilot · 2025-12-06T05:20:04Z

libs/partners/xai/langchain_xai/chat_models.py

+    
+    xAI's API requires that all messages have at least an empty content field,
+    unlike OpenAI which allows None for messages with tool_calls.
+    
+    Args:
+        message: The LangChain message to convert
+        api: The API format to use (default: "chat/completions")
+        
+    Returns:
+        Dictionary representation of the message compatible with xAI's API


The docstring is missing type information in the Args and Returns sections. Following Google-style docstring conventions, the Args should include types and the Returns section should specify the dict structure or reference the type hint.

Example:

Args: message: The LangChain message to convert. api: The API format to use. Defaults to "chat/completions". Returns: Dictionary representation of the message compatible with xAI's API.

Note: Types should be in the function signature (which they are), so the docstring can focus on describing the parameters rather than repeating types.

Suggested change

xAI's API requires that all messages have at least an empty content field,

unlike OpenAI which allows None for messages with tool_calls.

Args:

message: The LangChain message to convert

api: The API format to use (default: "chat/completions")

Returns:

Dictionary representation of the message compatible with xAI's API

xAI's API requires that all messages have at least an empty content field,

unlike OpenAI which allows None for messages with tool_calls.

Args:

message: The LangChain message to convert.

api: The API format to use. Defaults to "chat/completions".

Returns:

Dictionary representation of the message compatible with xAI's API.

Copilot · 2025-12-06T05:20:04Z

libs/partners/xai/langchain_xai/chat_models.py

+        
+        Overrides the base implementation to use xAI-specific message conversion
+        that ensures all messages have at least an empty content field.


The docstring is incomplete. It's missing the Args and Returns sections that are present in similar functions throughout the codebase. Following Google-style docstring conventions:

"""Prepare the request payload for xAI's API. Overrides the base implementation to use xAI-specific message conversion that ensures all messages have at least an empty content field. Args: input_: The input to convert into messages. stop: List of stop sequences. Defaults to None. **kwargs: Additional keyword arguments to pass to the API. Returns: Dictionary containing the request payload for xAI's API.

Suggested change

Overrides the base implementation to use xAI-specific message conversion

that ensures all messages have at least an empty content field.

Overrides the base implementation to use xAI-specific message conversion

that ensures all messages have at least an empty content field.

Args:

input_: The input to convert into messages.

stop: List of stop sequences. Defaults to None.

**kwargs: Additional keyword arguments to pass to the API.

Returns:

Dictionary containing the request payload for xAI's API.

Copilot · 2025-12-06T05:20:05Z

libs/partners/xai/langchain_xai/chat_models.py

+    _convert_from_v1_to_chat_completions,
+    _convert_message_to_dict as _openai_convert_message_to_dict,
+    _get_last_messages,
+    _use_responses_api,


Import of '_use_responses_api' is not used.

Suggested change

_use_responses_api,

Copilot AI review requested due to automatic review settings December 6, 2025 05:15

arrdel requested review from ccurme and mdrxy as code owners December 6, 2025 05:15

github-actions bot added integration Related to a provider partner package integration xai labels Dec 6, 2025

Copilot started reviewing on behalf of arrdel December 6, 2025 05:15 View session

Copilot AI reviewed Dec 6, 2025

View reviewed changes

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

xai[patch]: Fix 400 error for AIMessages with tool_calls but no content #34229

xai[patch]: Fix 400 error for AIMessages with tool_calls but no content #34229

arrdel commented Dec 6, 2025

Uh oh!

codspeed-hq bot commented Dec 6, 2025

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Dec 6, 2025

Uh oh!

Copilot AI Dec 6, 2025

Uh oh!

Copilot AI Dec 6, 2025

Uh oh!

Copilot AI Dec 6, 2025

Uh oh!

Copilot AI Dec 6, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant


		Overrides the base implementation to use xAI-specific message conversion
		that ensures all messages have at least an empty content field.

xai[patch]: Fix 400 error for AIMessages with tool_calls but no content #34229

Are you sure you want to change the base?

xai[patch]: Fix 400 error for AIMessages with tool_calls but no content #34229

Conversation

arrdel commented Dec 6, 2025

Uh oh!

codspeed-hq bot commented Dec 6, 2025

CodSpeed Performance Report

Merging #34229 will not alter performance

Summary

Footnotes

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Uh oh!

Copilot AI Dec 6, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 6, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 6, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 6, 2025

Choose a reason for hiding this comment

Uh oh!

Copilot AI Dec 6, 2025

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant