How to dynamically load tool descriptions

[Huyueeer]

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

In the old version of langchain, I used the following method to register the tool:

tools = from_function({
    execute_sql_tao: get_prompt_template_prompty("tools", "execute_sql-gpt4o-v1")
})

def execute_sql_tao(sql: str):
    print('\n', '【execute_sql_tool_node】')
    sql = sql.replace('sql', "").replace('```', "")
    data = postgres_raw_sql_query(sql, db_key='GP')
    df = pd.DataFrame(data)
    data = df.to_markdown()
    return data

After I upgraded to the new version, I found that I could not load my prompt configuration when I registered using @tool:
```python
@tool
def new_execute_sql_tao(sql: str):
    print('\n', '【execute_sql_tool_node】')
    sql = sql.replace('```sql', "").replace('```', "")
    data = postgres_raw_sql_query(sql, db_key='PG')
    df = pd.DataFrame(data)
    data = df.to_markdown()
    return data

### Description

I studied the implementation of @tool and found that it gets the description by loading __doc__. Is there any way I can continue as before?

### System Info

none

[feijoes]

As stated in the documentation, the Tool decorator uses the function's docstring as the tool's description. Therefore, a docstring must be provided.

@tool
def search(query: str) -> str:
    """Look up things online."""
    return "LangChain"
print(search.description) # search(query: str) -> str - Look up things online.

EDIT:
Try using StructuredTool instead. This should achieve the desired result.

from langchain.tools import  StructuredTool
tool =  StructuredTool.from_function(
    name="<name>",
    func=execute_sql_tao
    description=get_prompt_template_prompty("tools", "execute_sql")
)

[Huyueeer]

Oh no, I think it's some kind of loop, hahaha, I tested it and found an error:

  File "D:\py_project\bot\venv\lib\site-packages\langchain_core\tools\convert.py", line 304, in tool
    return _create_tool_factory(name_or_callable.__name__)(name_or_callable)
  File "D:\py_project\bot\venv\lib\site-packages\langchain_core\tools\convert.py", line 249, in _tool_factory
    return StructuredTool.from_function(
  File "D:\py_project\bot\venv\lib\site-packages\langchain_core\tools\structured.py", line 190, in from_function
    raise ValueError(msg)
ValueError: Function must have a docstring if description not provided.

So I actually found that assigning description and using state: Annotated[dict, InjectedState] cannot be done in the same version, or in the same way as registering a tool.

[feijoes]

@Huyueeer description=get_prompt_template_prompty("tools", "execute_sql") is a empty string? Is a sync Function? Maybe is a versión issue? Try upgrade the package and check if description is none or something

I can't right now, but in a few hours, I'll try to create a similar tool and attempt to recreate the error. However, it's strange because structureTool should have all the necessary components.

[feijoes]

@Huyueeer Hi, I found the issue. Just as I mentioned, I get the same error if get_prompt_template_prompty returns None or an empty string, but if it returns a string, it works fine.

[Huyueeer]

Thank you very much for your patience and assistance, I checked it according to your idea, and then modified the code and found that it was OK. My final modification was:

def new_execute_sql_tao(sql: str, state: Annotated[dict, InjectedState]):
    print('\n', '【execute_sql_tool_node】')
    sql = sql.replace('```sql', "").replace('```', "")
    data = postgres_raw_sql_query(sql, db_key=selectSiteDB(state['site']))
    df = pd.DataFrame(data)
    data = df.to_markdown()
    return data

from langchain.tools import  StructuredTool
tools =  [StructuredTool.from_function(
    name="<name>",
    func=execute_sql_tao
    description=get_prompt_template_prompty("tools", "execute_sql")
)]