Description
Tool calling is working very unreliably, and unpredictably with Agno and Groq. Often tool calls fail with error code 400, failed to call a function. Also many LLM calls are generated by the Agno agent which causes API resource usage issues.
Steps to Reproduce
To reproduce I have modified an Agno documentation example team_workflow.py to use Groq instead of OpenAI, so model=Groq(id="llama-3.3-70b-versatile").
Agent Configuration (if applicable)
reddit_researcher = Agent( name="Reddit Researcher", role="Research a topic on Reddit", model=Groq(id="llama-3.3-70b-versatile"), tools=[searxng_tool], add_name_to_instructions=True, instructions=dedent(""" You are a Reddit researcher. You will be given a topic to research on Reddit. You will need to find the most relevant posts on Reddit. """), )
Expected Behavior
Tool call succeeding without error
Actual Behavior
Outputs: Working...ERROR Error calling Groq API: Error code: 400 - {'error': {'message': "Failed to call a function. Please adjust your prompt. See 'failed_generation' for more details.", 'type':
'invalid_request_error', 'code': 'tool_use_failed', 'failed_generation': '### Top Stories from Reddit\nAs the Reddit Researcher, I have found the following top stories on Reddit
Screenshots or Logs (if applicable)
No response
Environment
### Environment
MacOS 15.5
Agno 1.7.7
Possible Solutions (optional)
No response
Additional Context
- Error happens in both single agent and team of agents.
- Other LLMs like Groq meta-llama/llama-4-maverick-17b-128e-instruct, gemma2-9b-it fail as well in tool calling.
- Sometimes tool call works, sometimes it doesn't. Two successive runs of same example with same input prompt can result in one succeeding and one failing.
- Use of other frameworks like LangChain do not generate this error using Groq for inference
Description
Tool calling is working very unreliably, and unpredictably with Agno and Groq. Often tool calls fail with error code 400, failed to call a function. Also many LLM calls are generated by the Agno agent which causes API resource usage issues.
Steps to Reproduce
To reproduce I have modified an Agno documentation example team_workflow.py to use Groq instead of OpenAI, so model=Groq(id="llama-3.3-70b-versatile").
Agent Configuration (if applicable)
reddit_researcher = Agent( name="Reddit Researcher", role="Research a topic on Reddit", model=Groq(id="llama-3.3-70b-versatile"), tools=[searxng_tool], add_name_to_instructions=True, instructions=dedent(""" You are a Reddit researcher. You will be given a topic to research on Reddit. You will need to find the most relevant posts on Reddit. """), )Expected Behavior
Tool call succeeding without error
Actual Behavior
Outputs: Working...ERROR Error calling Groq API: Error code: 400 - {'error': {'message': "Failed to call a function. Please adjust your prompt. See 'failed_generation' for more details.", 'type':
'invalid_request_error', 'code': 'tool_use_failed', 'failed_generation': '### Top Stories from Reddit\nAs the Reddit Researcher, I have found the following top stories on Reddit
Screenshots or Logs (if applicable)
No response
Environment
### Environment MacOS 15.5 Agno 1.7.7Possible Solutions (optional)
No response
Additional Context