LocalIQ — AI Market Intelligence for Small Businesses

Built for RLC Hacks 2026 · Track: Small Business AI · Theme: Optimize the future of technology

The Problem

Small businesses don't lose to big chains because of worse products — they lose because they have no data. Walmart runs demand forecasting; Amazon predicts trending demand weeks ahead. A local coffee shop is left guessing. LocalIQ closes that gap with instant, free market intelligence.

What It Does

Enter a business type and city. LocalIQ returns, in seconds:

Local search trends — what people in your area are actively searching for right now
Competitor weaknesses — gaps in nearby competitors you can exploit this week
AI-written social media — ready-to-post content for Instagram, Facebook, and Twitter
A weekly action playbook — concrete, measurable next steps with a KPI to track

Try it: localiq12.vercel.app — e.g., coffee in Austin.

How We Built It

Frontend (React 18 + Vite + Tailwind + TypeScript): 2-state UI (form → results) with a responsive insight grid, history & comparison modes, PDF/JSON export, light/dark theme, full PWA support, and keyboard-accessible focus rings.
Backend (Vercel serverless functions): three stateless /api endpoints — trends, competitors, insight — each with security headers, CORS, input validation, and graceful mock fallback. No persistent server to scale, no sleeping free-tier backend.
AI Integration: GLM 5.1 via the HCNSEC endpoint (model glm-5.1, falling back to auto → routes models like stepfun-ai/step-3.7-flash) for market summaries, social copy, and action items. If the model call fails, realistic mock data takes over so the demo never crashes.

Architecture Note

The whole app deploys as a single Vercel project: the React SPA is the static output and the Express-routed backend was converted to native Vercel serverless functions under /api/*. Same origin, no CORS gymnastics, no sleeping backend.

What I Learned

Security-first: every endpoint was threat-modeled against OWASP — security headers, rate limiting, input sanitization, and generic (non-leaky) error messages.
Zero-budget resilience: free-tier APIs + mock-first design mean the app works even when every external API is down or rate-limited.
Stateless functions: converting a long-lived Express app to per-request serverless handlers forced clean request isolation.

Math / Metrics

Average insight latency: ~2.3s (GLM) / <0.8s (mock fallback)
Endpoint mock coverage: 100% (all 3 routes gracefully degrade)
Security layers: 5 (security headers, CORS, input validation, generic errors, max-duration cap)

AI Disclosure (per RLC Hacks rules)

This project uses GLM 5.1 (https://api.hcnsec.cn/v1/chat/completions, with auto fallback) to generate market summaries, social media captions, and action recommendations. Claude Code (Anthropic) assisted in building the project. Mock data fallbacks are clearly labeled via a source field in every API response.