Inventory schema
Homepage overview
Finances dashboard
Insights chat interface

Equinox - AI-Powered Enterprise Resource Planning & Autovisualization agent with AI voice agent outbound calling

Inspiration

Witnessing businesses struggle with inefficient inventory management systems that required manual data entry, complex SQL queries for basic insights, delayed supplier communications causing stock-outs, and repetitive phone calls for price negotiations. We envisioned transforming Enterprise Resource Planning (ERP) from a data storage system into an intelligent business partner where owners could simply ask inventory questions out loud and get instant, intelligent responses.

What it does

Equinox is an AI-powered ERP system that revolutionizes inventory management through:

Automated supplier negotiations: AI voice agents call suppliers and negotiate prices autonomously
Predictive analytics: Prevent stock-outs before they happen with intelligent forecasting
Natural language SQL generation: Convert business questions into database queries without technical knowledge
Real-time dashboards: Dynamic visualizations that update automatically with business intelligence

The system transforms complex business operations into simple conversations while maintaining human oversight for critical decisions.

How we built it

Foundation: React + Vite frontend with Supabase backend for real-time capabilities and authentication

AI-Powered Query Engine:

Google Gemini 2.5 Flash converts natural language to SQL queries
Multi-layered validation prevents dangerous operations
Comprehensive 67-column inventory schema covers SKU tracking to advanced analytics

Voice Interface:

Vapi.ai processes spoken inventory questions
Function calling enables AI agents to execute real database operations
Two-step verification for destructive operations

Business Automation:

Twilio + 11labs power automated supplier negotiation calls
Multi-agent architecture with specialized roles (Supplier, Analysis, Decision, Transaction agents)
Smart deal analysis and recommendation engines

Intelligence Layer:

Chart.js visualizations generated dynamically by Gemini
Real-time updates via Supabase subscriptions
Predictive analytics for inventory optimization

Challenges we ran into

AI Reliability: AI services generated invalid SQL or inconsistent responses. Solved with multi-layered validation and query sanitization.

Voice Recognition Accuracy: Business operations need perfect accuracy. Implemented confirmation workflows and two-step verification for critical actions.

Real-Time Data Sync: Multiple users caused data conflicts. Used Supabase's real-time subscriptions with optimistic UI updates and conflict resolution.

Security Risks: AI-generated SQL posed security threats. Implemented comprehensive query validation, parameterized queries, and Row Level Security (RLS).

Performance at Scale: Voice queries on 10k+ items caused timeouts. Optimized with pagination, indexed queries, and intelligent data aggregation.

Accomplishments that we're proud of

10x faster inventory analysis through voice queries vs traditional interfaces
Automated supplier negotiations saving hours of manual phone calls
Predictive insights that prevent stock-outs before they occur
Zero SQL knowledge required - natural language interface for all users
Multi-agent negotiation system that operates autonomously while keeping humans in control

Successfully proved that business software can be reimagined using AI and voice technologies to create more intelligent, conversational systems.

What we learned

AI Integration Complexity: Mastered guardrail engineering, function calling patterns, fallback systems, and context management across multi-turn conversations.

Real-Time Data Management: Learned Row Level Security for multi-tenant isolation, optimistic UI updates, data validation layers, and caching strategies for large datasets.

Voice Interface Design: Understood natural language processing complexity, error handling for speech recognition, and user feedback systems for confirming AI actions.

Service-Oriented Architecture: Adopted clear separation of concerns across voice interfaces, AI responses, database updates, and UI components.

What's next for Equinox

Multi-language support for global businesses
Advanced ML models for more sophisticated demand forecasting
Integration marketplace connecting with popular business tools (Shopify, QuickBooks, etc.)
Mobile app with offline voice capabilities
Industry-specific modules tailored for retail, manufacturing, and logistics
Blockchain integration for supply chain transparency and automated smart contracts
Advanced analytics dashboard with custom KPI tracking and automated reporting