💡 Inspiration Most group plans don't die because of a lack of interest—they die in the coordination graveyard. Research shows that if a group chat hits 20 messages without a decision, the plan is 70% likely to fail. We built Alfredo to end the "When/Where" spiral for dating, business outings, and group vacations.
🚀 What it does Alfredo is an autonomous agentic pipeline that lives in your Discord. By running a single /alfredo command, the agent:
- Reads the Vibe: Analyzes the last 5 messages for context (cuisine, budget, neighborhood).
- Syncs Calendars: Autonomously DMs every participant to find the only 90-minute window where everyone is free.
- Books Autonomously: Queries restaurants against group dietary restrictions, picks the best fit, and uses browser automation to book on OpenTable.
- Voice Fallback: If the venue isn't online, Alfredo uses Voice AI to call the restaurant and confirm the booking manually.
🛠️ How we built it We focused on building a "Ship to Prod" architecture using federated infra: -WunderGraph Cosmo: We implemented 3 federated subgraphs (Users, Restaurants, and Hotels). The Cosmo Router handles parallel data fetching, allowing us to merge user dietary data with real-time restaurant availability in a single request. -TinyFish: Powering our "Agentic Operator." It navigates complex SPAs like OpenTable to perform bookings that traditional APIs can't reach. -Ghost: We managed two Postgres instances (Users & Sessions) entirely via MCP, allowing us to iterate on schema design directly from our agent logic. -Vapi: Our "fail-safe" layer. If browser automation times out, Vapi triggers an outbound AI call to ensure the reservation is secured.
📈 SMB & Social Impact Alfredo isn't just for users; it’s for the Small Business. Restaurant and hotel owners lose thousands to "no-shows" and last-minute dietary surprises. Our Operator Dashboard provides SMBs with live session data and party-specific dietary flags before the group arrives, allowing them to prep better and maximize table turnover.
🧠 Challenges & Learnings The biggest challenge was the parallelization of intent. Coordinating 10 different calendars while simultaneously querying multiple subgraphs for "vibes" required a robust federation layer. We learned that WunderGraph Cosmo is the secret sauce for making multi-agent systems feel "instant" rather than sluggish.
🛠️ What's next? We are expanding "Hotel Mode" to handle multi-city group travel coordination and integrating with more SMB-focused POS systems to provide even deeper operational insights for restaurant owners.
Built With
- ghost
- tinyfish
- vapi
- wondergraph