I have created a crew with:
- 4 tasks do some analytics about a conversation
- 1 task, summarizes those analytics
- 2 agents: 1 for crew manager, 1 for some analytics
Each task has structured outputs.
I have created a crew like so:
Crew(
tasks=[
ConversationSentimentTask,
ConversationSummarizationTask,
ReputationTask,
ServiceQualityTask,
SalesRecommendationTask,
ConversationAnalyticsReportTask,
],
agents=[SalesRecommendationAgent],
manager_agent=ConversationAnalyticsAgent,
process=Process.sequential,
verbose=True,
)
Each task except the last one could be run async. And the last task contains all previous tasks as context
.
There are some questions:
- Is this approach correct?
- Often, some tasks fail to generate desired structured output and the run fails. Then the crew stuck on “Task Completion\n Task Completed” but never outputs the final answer. What could be the issue?