Multi-Agent Task Manager
A CrewAI system with hierarchical delegation: a Planner agent breaks a request down and routes work to dedicated Slack, Jira, and Memory agents, backed by Weaviate semantic memory, with retries, approval workflows, and both a FastAPI and Streamlit interface.
User request
|
Planner agent
|-------------------|-------------------|
Slack agent Jira agent Memory agent
| | |
Slack API Jira API Weaviate (vector DB)
|-------------------|-------------------|
Final resultProblem
Turning a single natural-language request — "create a ticket and notify the team" — into coordinated, trackable actions across Slack and Jira, with the system remembering context across requests instead of treating each one in isolation.
Architecture
A Planner agent decomposes the request and hierarchically delegates to specialized Slack and Jira agents, while a Memory agent reads and writes semantic context to Weaviate for sub-second recall. Background task execution with status tracking, retries, and approval gates sits behind both a FastAPI service and a Streamlit UI.
My Contribution
Designed the hierarchical delegation model and built all four agents (Planner, Slack, Jira, Memory), the background task/retry system, and both interfaces.
Outcome
Automates multi-step Slack/Jira workflows that previously required manually creating tickets and cross-posting status updates by hand.