Ungraded Lab: Email Assistant Workflow- prompt to delete email by subject

Course Q&A Agentic AI
,
1

Module 3: Tool use

Ungraded Lab: Email Assistant Workflow

The last task in this lab, “6.5. Targeted action: Delete the “Happy Hour” email” has a prompt to delete the email using just the text.

The default prompt did not work for me, even after several runs.

I am sharing the prompt that actually worked for me

Delete the email with subject happy hour:white_check_mark:

Delete the "Happy Hour” email:cross_mark:

2

It seems that using the word “subject” might be confusing the LLM in this case and affecting its understanding of the prompt.

3

Did your failure result in spew starting like this?

ValidationError                           Traceback (most recent call last)
File /usr/local/lib/python3.11/site-packages/aisuite/utils/tools.py:271, in Tools.execute_tool(self, tool_calls)
    270 try:
--> 271     validated_args = param_model(**arguments)
    272     result = tool_func(**validated_args.model_dump())

File /usr/local/lib/python3.11/site-packages/pydantic/main.py:253, in BaseModel.__init__(self, **data)
    252 __tracebackhide__ = True
--> 253 validated_self = self.__pydantic_validator__.validate_python(data, self_instance=self)
    254 if self is not validated_self:
4

The spew above was from the provided prompt (“Delete the happy hour email”). A prompt that worked for me:
“Find the happy hour email and delete it”

My sense is that the real fix would be in the docstring.

5

I was also able to resolve the issue by changing the definition of the tool to (pardon the accursed formatting):

def find_emails_matching_query(query: str) → list:
“”"
Find emails containing the query in subject, body, or sender.

Args:
    query (str): A keyword or phrase to search for.

Returns:
    List[dict]: A list of emails matching the query string.
"""
return requests.get(f"{BASE_URL}/emails/search", params={"q": query}).json()