from crewai import Agent, Task, Crew, Process
Query Generator Agent
Query_generator_agent = Agent(
role=“Eligibility Query Generator”,
goal=“Convert structured JSON input response into a natural language query that retains all field values, ensuring a complete, readable, and contextually rich format.”,
backstory=“”"
You are an expert at processing structured data and turning it into a single, well-formed eligibility query.
Your queries help decision-making systems—both automated and human reviewers—determine eligibility without missing any key information.
“”",
llm=llm_model,
verbose=True,
max_rpm=6,
respect_context_window=True,
allow_delegation=False
)
Query Generation Task
Query_generate_task = Task(
description=“”"
Transform structured JSON input response into a single, well-formed natural language query that retains all field values while assessing eligibility.
The final query must be neutral and present all available details for eligibility assessment.
“”",
expected_output=“return Query generated in str format”
)
Admission Evaluation Agent
Admission_Evaluation_agent = Agent(
role=“Admission Evaluation Agent”,
goal=“Evaluate marks for users received in the query.”,
backstory=“”"
You are responsible for assessing user details against predefined admission criteria stored in your tool.
You analyze applications, verify eligibility, and determine if a candidate meets admission requirements.
“”",
llm=llm_model,
verbose=True,
max_rpm=6,
respect_context_window=True,
tools=[rag_tool],
allow_delegation=False
)
Evaluation Task
Eval_agent_task = Task(
description=“”"
Process the details received from the query generator response and use the available knowledge resource
to evaluate the user for admission eligibility only. Do NOT evaluate for insurance. Strictly act on information you have.
“”",
expected_output=“return Yes or No”,
context=[Query_generate_task]
)
Manager Agent
manager = Agent(
role=“Eligibility Assessment Orchestrator”,
goal=“Efficiently manage the eligibility assessment pipeline by analyzing responses and determining the appropriate evaluation path.”,
backstory=“”"
The Planner Agent orchestrates the eligibility verification process by analyzing user responses and routing them through the correct evaluation path.
It ensures logical execution, data accuracy, and process efficiency.
“”",
llm=llm_model,
verbose=True,
max_rpm=6,
respect_context_window=True,
allow_delegation=True
)
Crew Definition
Crew_three = Crew(
agents=[Query_generator_agent, Admission_Evaluation_agent],
task=[Query_generate_task, Eval_agent_task],
verbose=True,
manager_agent=manager,
process=Process.hierarchical
)
Start Crew Process
Crew_three.kickoff(inputs={“response”: raw_results})
##Error:
[Flow._execute_single_listener] Error in method third_method: Something went wrong. Kickoff should return only one task output.
Traceback (most recent call last):
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\flow\flow.py”, line 363, in _execute_single_listener
listener_result = await self._execute_method(
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\flow\flow.py”, line 306, in _execute_method
else method(*args, **kwargs)
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\exploreOnPOC\OGFolder\all_agents.py”, line 279, in third_method
second_results = Crew_three.kickoff(inputs={“response”:raw_results})
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\crew.py”, line 555, in kickoff
result = self._run_hierarchical_process()
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\crew.py”, line 665, in _run_hierarchical_process
return self._execute_tasks(self.tasks)
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\crew.py”, line 770, in _execute_tasks
return self._create_crew_output(task_outputs)
File “C:\Users\francis.nadirsha\Desktop\GENVIZ\Genviz\lib\site-packages\crewai\crew.py”, line 874, in _create_crew_output
raise ValueError(
ValueError: Something went wrong. Kickoff should return only one task output.
None
Error is arising when I am using agent orchestration.