[langchain-deepseek] Missing `reasoning_content` in request payload when using deepseek-reasoner with tool calling

[yefengzi7]

Checked other resources

• This is a bug, not a usage question.
• I added a clear and descriptive title that summarizes this issue.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).
• This is not related to the langchain-community package.
• I posted a self-contained, minimal, reproducible example. A maintainer can copy it and run it AS IS.

Package (Required)

• langchain
• langchain-openai
• langchain-anthropic
• langchain-classic
• langchain-core
• langchain-cli
• langchain-model-profiles
• langchain-tests
• langchain-text-splitters
• langchain-chroma
• langchain-deepseek
• langchain-exa
• langchain-fireworks
• langchain-groq
• langchain-huggingface
• langchain-mistralai
• langchain-nomic
• langchain-ollama
• langchain-perplexity
• langchain-prompty
• langchain-qdrant
• langchain-xai
• Other / not sure / general

Example Code (Python)

### Checked other resources

- [x] I added a very descriptive title to this issue.
- [x] I searched the LangChain documentation with the integrated search.
- [x] I used the GitHub search to find a similar question and didn't find it.
- [x] I am sure that this is a bug in LangChain rather than my code.
- [x] The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

### Example Code


from langchain_deepseek import ChatDeepSeek
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage
from langchain_core.tools import tool

@tool
def calculator(a: float, b: float) -> float:
    """Add two numbers."""
    return a + b

llm = ChatDeepSeek(model="deepseek-reasoner")
llm_with_tools = llm.bind_tools([calculator])

# First call - works fine
messages = [HumanMessage(content="What is 1 + 2?")]
response = llm_with_tools.invoke(messages)
# response.additional_kwargs contains 'reasoning_content'

# Simulate tool execution
messages.append(response)
messages.append(ToolMessage(content="3", tool_call_id=response.tool_calls[0]["id"]))

# Second call - fails with 400 error
response2 = llm_with_tools.invoke(messages)  # ERROR!

Error Message and Stack Trace (if applicable)

Error code: 400 - {'error': {'message': 'Missing `reasoning_content` field in the assistant message at message index 1. For more information, please refer to https://api-docs.deepseek.com/guides/thinking_with_tools', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}}

Description

Checked other resources

• I added a very descriptive title to this issue.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.
• I am sure that this is a bug in LangChain rather than my code.
• The bug is not resolved by updating to the latest stable version of LangChain (or the specific integration package).

Example Code

from langchain_deepseek import ChatDeepSeek
from langchain_core.messages import HumanMessage, AIMessage, ToolMessage
from langchain_core.tools import tool

@tool
def calculator(a: float, b: float) -> float:
    """Add two numbers."""
    return a + b

llm = ChatDeepSeek(model="deepseek-reasoner")
llm_with_tools = llm.bind_tools([calculator])

# First call - works fine
messages = [HumanMessage(content="What is 1 + 2?")]
response = llm_with_tools.invoke(messages)
# response.additional_kwargs contains 'reasoning_content'

# Simulate tool execution
messages.append(response)
messages.append(ToolMessage(content="3", tool_call_id=response.tool_calls[0]["id"]))

# Second call - fails with 400 error
response2 = llm_with_tools.invoke(messages)  # ERROR!

### System Info

System Information
OS: Windows
OS Version: 10.0.22631
Python Version: 3.12.12

Package Information
langchain_core: 1.1.0
langchain: 1.1.0
langchain_deepseek: 1.0.1
langchain_openai: 1.1.0
langgraph: 1.0.3
openai: 2.8.1

[iantinney]

I'll take a look into this, shouldn't be too hard to fix. Would appreciate if someone assigns it, thanks

[Jiweipy]

Bug Fix Proposal: ChatDeepSeek Class Missing reasoning_content Field for Reasoning Models

Issue Description

When using DeepSeek reasoning models (e.g., deepseek-reasoner) with tool calling functionality, the ChatDeepSeek class throws a 400 error:

BadRequestError: Error code: 400 - {'error': {'message': 'Missing `reasoning_content` field in the assistant message at message index 2', 'type': 'invalid_request_error', 'param': None, 'code': 'invalid_request_error'}}

Root Cause

The DeepSeek reasoning model API requires assistant messages to include a reasoning_content field, even if it's empty. The current ChatDeepSeek._get_request_payload() method does not properly handle this `requirement.

Reproduction Steps

• Create a ChatDeepSeek instance with the deepseek-reasoner model

• Use tool calling functionality

• After the model generates a response containing tool calls, continue the conversation

• The second assistant message lacks the reasoning_content field, causing the API to return a 400 error

Solution

Modify the ChatDeepSeek._get_request_payload() method to ensure all assistant messages include the reasoning_content field.

Specific Changes

file_path_class: site-packages/langchain_deepseek/chat_models.py => class ChatDeepSeek(BaseChatOpenAI)

def _get_request_payload(
    self,
    input_: LanguageModelInput,
    *,
    stop: list[str] | None = None,
    **kwargs: Any,
) -> dict:
    payload = super()._get_request_payload(input_, stop=stop, **kwargs)
    
    # Ensure assistant messages include reasoning_content field
    for i, message in enumerate(payload["messages"]):
        if message["role"] == "assistant":
            # DeepSeek API expects assistant content to be a string, not a list
            if isinstance(message["content"], list):
                text_parts = [
                    block.get("text", "")
                    for block in message["content"]
                    if isinstance(block, dict) and block.get("type") == "text"
                ]
                message["content"] = "".join(text_parts) if text_parts else ""
            
            # Ensure each assistant message includes reasoning_content field
            # Get reasoning content from additional_kwargs, or set to empty string if not present
            if "reasoning_content" not in message:
                # Try to get reasoning content from original message
                if isinstance(input_, list) and i < len(input_):
                    original_msg = input_[i]
                    if hasattr(original_msg, 'additional_kwargs'):
                        reasoning = original_msg.additional_kwargs.get("reasoning_content", "")
                        message["reasoning_content"] = reasoning
                    else:
                        message["reasoning_content"] = ""
                else:
                    message["reasoning_content"] = ""
        
        # Handle tool messages
        elif message["role"] == "tool" and isinstance(message["content"], list):
            message["content"] = json.dumps(message["content"])
    
    return payload