RediSprint

# RediSprint An A2A (Agent-to-Agent) Sprint Planning System that automates Jira ticket creation from Google Sheets using AI-powered context enrichment. Built for the [Agents with Superpowers Context Engineering Hackathon](https://sf.aitinkerers.org/p/agents-with-superpowers-context-engineering-hackathon-w-redis) in San Francisco (November 2024) - Winner of the one-day hackathon. ## Overview This system uses three collaborative agents that communicate via Redis Streams: 1. Sprint Reader Agent - Reads sprint planning data from Google Sheets 2. Context Enricher Agent - Searches previous tickets and codebase to enrich ticket descriptions with relevant context using Redis vector search 3. Jira Creator Agent - Creates comprehensive Jira tickets with AI-generated descriptions ## Key Features - AI-powered context enrichment using GPT-4o - Semantic search over past tickets and codebase using vector embeddings - Priority-based ticket creation based on client impact - Asynchronous agent communication via Redis Streams - Automatic latest sprint detection - Web UI for easy execution and real-time monitoring ## Configuration ### Google Sheet Format Your Google Sheet should have these columns (in order): | Sprint | Type | Title | Client Impact | Effort | Owner | Status | Assignee | |--------|------|-------|---------------|--------|-------|--------|----------| | Sprint 5 | Story | Add user authentication | High | 5 days | John | Planned | john@company.com | ### Advanced Configuration Additional settings in `a2a_agents.py` (usually don't need to change): - `QUEUE_PREFIX`: Redis queue prefix (default: "a2a:sprint:") - Search parameters: Number of tickets and code files to retrieve - Agent polling intervals and timeouts ### Priority Mapping The system automatically maps Client Impact to Jira Priority: | Client Impact | Jira Priority | |---------------|---------------| | Critical | Highest | | High | High | | Medium | Medium | | Low | Low | ## How It Works 1. Agent 1 reads your Google Sheet and identifies the latest sprint 2. For each ticket in the latest sprint: - Agent 2 performs semantic search on past tickets (top 5 results) - Agent 2 performs semantic search on codebase (top 3 files) - Agent 2 uses GPT-4o to generate a comprehensive ticket description - Agent 2 forwards enriched data to Agent 3 3. Agent 3 creates the Jira ticket with all enriched information ## Acknowledgments Built for the Agents with Superpowers Context Engineering Hackathon hosted by AI Tinkerers SF in collaboration with Redis. Winner of the one-day hackathon (November 2024).