[BUG] Custom tools are not called by the crewAI agent

[sreanik]

Description

Custom tools are not called by the crewAI agent (even when the agent decides to invoke the tool). This issue is seen in the version 0.114.0. Works fine in the version 0.108.0.

Steps to Reproduce

Create a custom tool and assign it to an agent. The task description should necessitate the agent to invoke the tool.

Expected behavior

The custom tool is called by the agent when it decides to invoke.

Screenshots/Code snippets

N/A

Operating System

Ubuntu 20.04

Python Version

3.12

crewAI Version

0.114.0

crewAI Tools Version

0.40.1

Virtual Environment

Venv

Evidence

I ran the same piece of code in both 0.114.0 and 0.108.0 version of crewAI and noticed that tool was called only in 0.108.0.

Possible Solution

Probably, comparison of the two versions (0.114.0 and 0.108.0) to identify the delta code change that introduced the issue.

Additional context

N/A

[Vidit-Ostwal]

Can you set the verbose=True and share all the logs once

[lorenzejay]

Can you share the custom tool code if you can ?

[sreanik]

Hi, here is a sample tool and I notice that the API tool is getting called but the logs do not show that the tool is invoked (In the earlier version of crewAI the logs capture that the tool is invoked and also shows the tool input and output).
Here, the statement print("---------Tool Invoked---------")
Code and the logs below.

`
SAMPLE_API_BASE_URL = "https://jsonplaceholder.typicode.com"

class SampleAPITool(BaseTool):
"""Tool for interacting with the sample JSONPlaceholder API."""

name: str = "sample_api_caller"
description: str = "Useful for fetching or creating sample data using the JSONPlaceholder API."

def _run(self, task: str, **kwargs) -> str:
    """Executes the API call based on the task and provided arguments."""
    try:

        print("---------Tool Invoked---------")
        endpoint = "/todos"
        params = kwargs.get("params", {})
        response = requests.get(f"{SAMPLE_API_BASE_URL}{endpoint}", params=params)
        response.raise_for_status()
        return json.dumps(response.json(), indent=2)

    except requests.exceptions.RequestException as e:
        return f"Error communicating with the Sample API: {e}"
    except json.JSONDecodeError:
        return f"Error decoding JSON response from the Sample API."
    except Exception as e:
        return f"An unexpected error occurred: {e}"

sample_api_tool = SampleAPITool()
`

───────────────────────────────────────────────────────── Crew Execution Started ──────────────────────────────────────────────────────────╮
│ │
│ Crew Execution Started │
│ Name: crew │
│ ID: 2c270d56-465c-416d-bc8b-3be1c757f9e8 │
│ │
│ │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

Provider List: https://docs.litellm.ai/docs/providers

🚀 Crew: crew
└── 📋 Task: e294f771-b90f-4659-87da-cb0d724f509e
Status: Executing Task...

Provider List: https://docs.litellm.ai/docs/providers

🚀 Crew: crew
└── 📋 Task: e294f771-b90f-4659-87da-cb0d724f509e
Status: Executing Task...
└── 🤖 Agent: API User
Status: In Progress

Agent: API User

Task: Use the 'sample_api_caller' tool to fetch a list of all todo items from the API.

🤖 Agent: API User
Status: In Progress
└── 🧠 Thinking...

2025-04-24 14:35:56,517 - 6242250752 - init.py-init:111 - WARNING: Invalid type NoneType for attribute 'gen_ai.response.service_tier' value. Expected one of ['bool', 'str', 'bytes', 'int', 'float'] or a sequence of those types
🤖 Agent: API User
Status: In Progress

Agent: API User

Final Answer:

{"todos": [{"userId": 1, "id": 1, "title": "delectus aut autem", "completed": false}, {"userId": 1, "id": 2, "title": "quis ut nam facilis et officia qui", "completed": false}, {"userId": 1, "id": 3, "title": "fugiat veniam minus", "completed": false}, {"userId": 1, "id": 4, "title": "et porro tempora", "completed": true}, {"userId": 1, "id": 5, "title": "laboriosam mollitia et enim quasi adipisci quia provident illum", "completed": false}, {"userId": 1, "id": 6, "title": "qui ullam ratione quibusdam voluptatem quia omnis", "completed": false}, {"userId": 1, "id": 7, "title": "illo expedita consequatur quia in", "completed": false}, {"userId": 1, "id": 8, "title": "quo adipisci enim quam ut ab", "completed": true}, {"userId": 1, "id": 9, "title": "molestiae perspiciatis ipsa", "completed": false}, {"userId": 1, "id": 10, "title": "illo est ratione doloremque quia maiores aut", "completed": true}]}

🚀 Crew: crew
└── 📋 Task: e294f771-b90f-4659-87da-cb0d724f509e
Status: Executing Task...
└── 🤖 Agent: API User
Status: ✅ Completed

🚀 Crew: crew
└── 📋 Task: e294f771-b90f-4659-87da-cb0d724f509e
Assigned to: API User
Status: ✅ Completed
└── 🤖 Agent: API User
Status: ✅ Completed
╭───────────────────────────────────────────────────────────── Task Completion ──────────────────────────────────────────────────────────────╮
│ │
│ Task Completed │
│ Name: e294f771-b90f-4659-87da-cb0d724f509e │
│ Agent: API User │
│ │
│ │
╰────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────╯

╭───────────────────────────────────────────────────────────── Crew Completion ──────────────────────────────────────────────────────────────╮
│ │
│ Crew Execution Completed │
│ Name: crew │
│ ID: 2c270d56-465c-416d-bc8b-3be1c757f9e8

[Vidit-Ostwal]

Okay, so the tool is getting called but logs aren't showing up?
Can you share your entire crew, along with agents ?

[lucasgomide]

Could you test following this tutorial?

[sreanik]

@Vidit-Ostwal. Here is the sample crew. You can even try with your example.

api_interaction_agent = Agent(
role="API User",
goal="Interact with the sample JSONPlaceholder API to retrieve and manage todo items.",
backstory="An expert in consuming RESTful APIs to fetch and manipulate data.",
verbose=True,
allow_delegation=False,
tools=[sample_api_tool],
llm=crewai_llm
)

get_all_todos_task = Task(
description="Use the 'sample_api_caller' tool to fetch a list of all todo items from the API.",
agent=api_interaction_agent,
expected_output="get all todo items"
)

api_crew = Crew(
agents=[api_interaction_agent],
tasks=[get_all_todos_task],
verbose=True
)

app = Flask(name)

@app.route('/kickoff', methods=['POST'])
def kickoff():
try:
print("Starting the Crew...")
result = api_crew.kickoff()
print("\nCrew Completed!")
print("\nFinal Results:")
print(result)
return 200
except Exception as e:
return jsonify({"status": "error", "message": str(e)}), 500

if name == 'main':
app.run(debug=True,port=3000, host='0.0.0.0')

[Vidit-Ostwal]

@sreanik Can you add the tool as well?

[sreanik]

Hi Vidit, The tool code is shared in the earlier comment along with the logs. Thanks!

…

On Fri, 25 Apr 2025 at 6:40 PM, Vidit Ostwal ***@***.***> wrote: *Vidit-Ostwal* left a comment (crewAIInc/crewAI#2664) <#2664 (comment)> @sreanik <https://github.com/sreanik> Can you add the tool as well? — Reply to this email directly, view it on GitHub <#2664 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ARKUOPYFBI4CBV73LXOEKPT23IX5PAVCNFSM6AAAAAB3UEUIUSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMZQGM4TCMZUHE> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[Vidit-Ostwal]

Hi Vidit,

The tool code is shared in the earlier comment along with the logs.

Thanks!
…

On Fri, 25 Apr 2025 at 6:40 PM, Vidit Ostwal @.***> wrote:
Vidit-Ostwal left a comment (crewAIInc/crewAI#2664)
<#2664 (comment)>

@sreanik https://github.com/sreanik Can you add the tool as well?

—
Reply to this email directly, view it on GitHub
<#2664 (comment)>,
or unsubscribe
https://github.com/notifications/unsubscribe-auth/ARKUOPYFBI4CBV73LXOEKPT23IX5PAVCNFSM6AAAAAB3UEUIUSVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDQMZQGM4TCMZUHE
.
You are receiving this because you were mentioned.Message ID:
@.***>

Yes, sorry I missed that.
Thanks.

[lucasgomide]

@sreanik I'm missing args_schema in your Tool. Could you please try this tutorial? or using the @tool decorator?

[sreanik]

@lucasgomide thanks for your response. I had initially tried with the @tool decorator and it had failed.

[Vidit-Ostwal]

@lucasgomide thanks for your response. I had initially tried with the @tool decorator and it had failed.

Are you using args_schema this helps generating the type of input response which is required for calling the tool.