Forcing the use of format causes an immediate error if { is included in the system prompt.

[Ayn1631]

Checked other resources

• I added a very descriptive title to this question.
• I searched the LangChain documentation with the integrated search.
• I used the GitHub search to find a similar question and didn't find it.

Commit to Help

• I commit to help with one of those options 👆

Example Code

@tool
def Test_Tool(code_str:str, code_input:str) -> str:
    '''
    这是一个测试Tool
    '''
    
    return True
functions_call = chat.bind_tools(tools=[Test_Tool]).kwargs['functions']  # 获得工具列表
systemPrompt = f"""
你可以调用工具, 这是工具列表
{functions_call}
"""
prompt = ChatPromptTemplate.from_messages(
    [
        (
            "system",
            systemPrompt,
        ),
        ("user", "{input}"),
    ]
)

agent_debug = (
    prompt
    | chat
)

agent_debug.invoke({'input': '计算斐波那契数列第100项'})



```cmd
---------------------------------------------------------------------------
KeyError                                  Traceback (most recent call last)
Cell In[699], line 1
----> 1 agent_debug.invoke({'input': '计算斐波那契数列第100项'})

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\runnables\base.py:3022, in RunnableSequence.invoke(self, input, config, **kwargs)
   3020 context.run(_set_config_context, config)
   3021 if i == 0:
-> 3022     input = context.run(step.invoke, input, config, **kwargs)
   3023 else:
   3024     input = context.run(step.invoke, input, config)

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\prompts\base.py:208, in BasePromptTemplate.invoke(self, input, config, **kwargs)
    206 if self.tags:
    207     config["tags"] = config["tags"] + self.tags
--> 208 return self._call_with_config(
    209     self._format_prompt_with_error_handling,
    210     input,
    211     config,
    212     run_type="prompt",
    213     serialized=self._serialized,
    214 )

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\runnables\base.py:1927, in Runnable._call_with_config(self, func, input, config, run_type, serialized, **kwargs)
   1923     context = copy_context()
   1924     context.run(_set_config_context, child_config)
   1925     output = cast(
   1926         Output,
-> 1927         context.run(
   1928             call_func_with_variable_args,  # type: ignore[arg-type]
   1929             func,  # type: ignore[arg-type]
   1930             input,  # type: ignore[arg-type]
   1931             config,
   1932             run_manager,
   1933             **kwargs,
   1934         ),
   1935     )
   1936 except BaseException as e:
   1937     run_manager.on_chain_error(e)

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\runnables\config.py:396, in call_func_with_variable_args(func, input, config, run_manager, **kwargs)
    394 if run_manager is not None and accepts_run_manager(func):
    395     kwargs["run_manager"] = run_manager
--> 396 return func(input, **kwargs)

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\prompts\base.py:182, in BasePromptTemplate._format_prompt_with_error_handling(self, inner_input)
    181 def _format_prompt_with_error_handling(self, inner_input: dict) -> PromptValue:
--> 182     _inner_input = self._validate_input(inner_input)
    183     return self.format_prompt(**_inner_input)

File D:\Mypower\Anaconda3\env\torch124\Lib\site-packages\langchain_core\prompts\base.py:176, in BasePromptTemplate._validate_input(self, inner_input)
    170     example_key = missing.pop()
    171     msg += (
    172         f"\nNote: if you intended {{{example_key}}} to be part of the string"
    173         " and not a variable, please escape it with double curly braces like: "
    174         f"'{{{{{example_key}}}}}'."
    175     )
--> 176     raise KeyError(
    177         create_message(message=msg, error_code=ErrorCode.INVALID_PROMPT_INPUT)
    178     )
    179 return inner_input

KeyError: 'Input to ChatPromptTemplate is missing variables {"\'name\'"}.  Expected: ["\'name\'", \'input\'] Received: [\'input\']\nNote: if you intended {\'name\'} to be part of the string and not a variable, please escape it with double curly braces like: \'{{\'name\'}}\'.\nFor troubleshooting, visit: https://python.langchain.com/docs/troubleshooting/errors/INVALID_PROMPT_INPUT '

### Description

What I'm Doing
I'm using LangChain to enable LLMs to call tools.
Since I can't use OpenAI models and don't have an OpenAI API key, I need an alternative approach.

To achieve this, I generate a list of callable tools using the convert_to_openai_tool function, format the output to mimic OpenAI's function call structure, and provide this to my LLM so it can use the tools effectively.

Error Description
An error occurred while building the Chain.
I suspect that when invoking the model, the enforced use of format is causing issues. Specifically, the { characters in the tool list are being interpreted as format placeholders, leading to a failure in execution.

Desired Solutions (Any One of These)
Fix the error
Find a way to enable function calling without using OpenAI Chat

Timestamp: March 3, 2025, 21:48:35 Hui

### System Info


# pip freeze | findstr langchain
langchain==0.3.13
langchain-cohere==0.3.4
langchain-community==0.3.13
langchain-core==0.3.28
langchain-experimental==0.3.4
langchain-openai==0.2.14
langchain-text-splitters==0.3.4

[Ayn1631]

I consulted GPT but did not receive a solution.
The methods proposed by GPT include but are not limited to:

Method 1: JSON processing for {} to avoid f"" misinterpretation

Method 2: Print functions_call to ensure the tool list is correct
<functions_call is correct, here is the output>

[{'name': 'Test_Tool',
  'description': '这是一个测试Tool',
  'parameters': {'properties': {'code_str': {'type': 'string'},
    'code_input': {'type': 'string'}},
   'required': ['code_str', 'code_input'],
   'type': 'object'}}]

Method 3: Ensure ChatPromptTemplate variable matching and remove name if unnecessary
<name was not defined by me; it was misdetected>

Method 4: Check @tool parsing to ensure the tool is correctly bound