Goofre

What inspired us :bulb: {#inspiration}

Canadian SMBs are currently caught in a "Silo Trap." With a slowing economy and shifting trade dynamics, merchants need to amplify their business, but they are overwhelmed by bloated tech stacks. The shift towards ==Agentic Commerce==—where AI agents execute tasks across discovery and purchase—is transforming retail, but adoption is nearly impossible for a local bootstrapped startup without a massive engineering team.

"Merchants don't need another dashboard with 50 charts. They need a Chief of Staff."

Watching my son build his own Dino World game, I realized that managing any ecosystem requires seamless orchestration. We wanted to build a bridge to Google's Unified Commerce Profile (UCP) [^1] that was as simple as asking a question.

What it does :gear: {#what-it-does}

Goofre App is an enterprise-grade intelligence interface facilitating Merchant adoption of the Google Unified Commerce Profile (UCP). It bridges the gap between raw Google retail datasets and actionable agentic commerce through a multi-stage orchestration engine and a voice-first Command Center.

Instead of logging into Google Analytics, Search Console, Merchant Center, and an inventory PoS separately, a merchant simply taps a microphone and asks: "Hey Go Free, what is the status?"

The app connects to the Goofre Open Source engine, synthesizes live data across Sales, Marketing, Inventory, and Customer Service, and delivers a concise, 15-second audio briefing from a veteran AI persona.

How we built it

How we built it :hammer_and_wrench: {#architecture}

For the 1^st^ iteration of our engine, we architected a 5-step data and reasoning pipeline:

We feed the LLM a clean, strictly-formatted payload to ensure predictability:

{
  "sales": "Dino World AR Playset conversions dropped to zero.",
  "analytics": "Local Cambridge web traffic spiked 40% for sustainable toys.",
  "inventory": "Warehouse B POS sync is delayed by 245 minutes.",
  "service": "High churn risk detected for 3 VIP customers."
}

## Challenges we ran into
The "Midnight Micro-Services Collision." At hour 40 of the hackathon, we attempted to layer 4 distinct AI personas over a ~~massive 12-point mock data matrix~~ bloated React state. The state managers and Gemini intent-routing collided, resulting in microphone abort loops.

We made a surgical pivot at midnight. We gutted the multi-persona architecture and consolidated it into a Single AI Manager. We locked Gemini into a strict prompt to prevent hallucinated markdown characters from breaking the Text-to-Speech engine. Predictability beat complexity.

## Accomplishments that we're proud of
We successfully decoupled a complex Google protocol and made it accessible via a zero-friction interface.

Our biggest accomplishment is the Economic Multiplier Effect. By splitting this project into an Open Source Engine and a Merchant App, we are creating two distinct impacts: stabilizing merchants with enterprise-grade AI, and creating a massive sandbox for Canadian developers to build local Agentic integrations.

## What we learned

We learned that in Agentic Commerce, reducing the number of tools is more important than adding new features. We can model the value of this orchestration mathematically by looking at the reduction in cognitive load ((C)):$$C_{total} = \sum_{i=1}^{n} (T_i \times D_i)$$Where (T) is the time spent and (D) is the decision fatigue per software tool. By reducing (n) (the number of disparate retail dashboards) down to 1 voice interface, we exponentially decrease the merchant's cognitive burden, freeing them to focus on actual growth.

What's next for Goofre OS :rocket: {#roadmap}
[x] Pivot to the Single AI Manager architecture for demo stability.

[ ] Onboard 3,000 developers to the Goofre open-source ACO by mid-2026.

[ ] Beta test the PWA with local SMBs in Cambridge to refine PoS latency syncs.

[ ] Upgrade the AI from "Read-Only" (briefings) to "Read/Write" actions.

Built With

• antigravity
• api
• elevenlabs
• firebase
• gemini
• google-cloud
• javascript
• opensource
• pwa
• react
• typescript
• vite