Best CrewAI Approach for Building a Ticketing Chatbot with API Workflows

CrewAI Community Support
1

I want to build an AI chatbot using CrewAI to handle ticketing tasks via API interactions:

  1. Update Ticket
  2. Create a New Ticket
  3. Close Ticket
  4. Fetch Ticket Details

Workflow:

  1. Receive user question
  2. Classify user intent and route to the correct CrewAI Flow
  3. Generate API request using user query, API schema, and few-shot examples
  4. Trigger API request and fetch response
  5. Generate a response for the user

Since tasks are workflow-oriented and interdependent, I plan to use CrewAI Flows for structured execution. Additionally, memory, state management, and caching are crucial for seamless conversation handling.

Key Considerations:

  • Implement custom tools & functions for API interactions
  • Maintain chatbot memory & state for continuity
  • Ensure robust error handling & logging

What is the best CrewAI implementation strategy for achieving this? Any best practices or guidelines would be appreciated!

2

Hi. Any answer?

I have been working on a similar chatbot which connects to the Atlassian MCP Server to raise/update IT Tickets. I’ve been stuck at a dead end.

What I am trying to do it I want an ITSM chatbot that converses naturally with the user until it gathers all the information about the issue that the user is facing. For this, I am trynig to include the human in the loop.I’ve defined a manager agent and a hierarchical process to do this which delegates to the other downsteam agents until it raises an issue with JIRA. I am getting a lot of errors where the manager agent is unable to delegate tasks to subsequent agents even though I have allowed delegation= True.

Also, the supervisor agent is asking for approval everytime it passes onto the downstram agent. This is not how it is supposed to work.