Closed yuriiIykyknow closed 14 hours ago
I get this in different configurations with the tools I use: PDFSearchTool always works, DOCXSearchTool always fails, and SerperSearchTool sometimes works sometimes not. When it works, the agents send the correct inputs:
## Tool Input:
"{\"search_query\": \"Whatever terms\"}"
And always when failing, the wrong tool input looks the same as yours:
## Tool Input:
"{\"name\": \"Something relevant\", \"description\": \"Some text to be analyzed.\", \"args_schema\": \"None\"}"
No idea where this is decided, when to use the good search_query and when the wrong name/description/args_schema, but tools definitely don't like the latter.
PS: maybe related to https://github.com/crewAIInc/crewAI/issues/1344 ?
When I set up CrewAI locally from the GitHub repository, the Dalli tool worked as expected. So, I assume the work on that is done but has not been pushed to PyPI. i see changes in agent.py
@yuriiIykyknow Thank you so much for reporting this!
Looking into this right now. I think we'll be able to push the new version that will address this issue later tonight or first thing tomorrow.
I'll update you and close this as soon as the new version get's cut!
@yuriiIykyknow I think this is related to what i found in #1511, which should be addressed in new releases
Yeah I'm seeing this bug in my Crew
Hey @yuriiIykyknow! The new version just dropped. Please update to the latest version of crewAI (0.76.9) and this issue should go away.
If anything else comes up, please let me know!
Description
Agent provides wrong input value to Dalli tool. Also, crewAi can't provide right arguments to Dalli tool because args_schema is wrong. Excpect 'image_description' but got 'description' key
Steps to Reproduce
Expected behavior
Get link to image)
Screenshots/Code snippets
Operating System
Windows 11
Python Version
3.12
crewAI Version
0.76.2
crewAI Tools Version
0.13.2
Virtual Environment
Venv
Evidence
Agent: Artist Thought: Thought: I need to ensure I am using the right key name in the Action Input for the Dall-E Tool. I should properly encapsulate the attributes in the correct format. Using tool: Dall-E Tool Tool Input: "{\"name\": \"Cat with Wings\", \"description\": \"A cute and friendly cat with colorful butterfly-like wings, big sparkling eyes, and a fluffy tail. The cat is posed playfully, and the background is whimsical and colorful, adding to its charm.\", \"args_schema\": {}, \"return_direct\": true}" Tool Output: Error: the Action Input is not a valid key, value dictionary.
Agent: Artist Thought: Thought: I have been using repeated descriptions without success. I'll simplify the action to just focus on the essential details for creating the cat with wings. Using tool: Dall-E Tool Tool Input: "{\"name\": \"Cat with wings\", \"description\": \"A cute cat with vibrant, fluffy wings, looking friendly and playful. The cat is sitting with a joyful expression, set against a cheerful, whimsical background.\", \"args_schema\": \"BaseModel\"}" Tool Output: Image description is required.
Possible Solution
Additional context
None