ContentHero

ContentHero – Project Story

🌟 Inspiration

While consulting with multiple businesses I kept rewriting the same briefs so every blog post, email, or ad sounded “on-brand.” Re-explaining the brand voice and target persona each time was painful. I imagined a one-click engine that remembers a company’s DNA, understands its customers, and spits out channel-ready copy on demand.

What I Learned

Deep personas outperform surface demographics – mapping pains, goals, and aspirations drives sharper hooks.
A two-step outline ➜ draft chain delivers higher-quality long-form copy than single prompts.
Marketers crave version history; “Save as Draft” became a top-used feature.
Even elite copywriters share structural DNA (headline rhythm, CTA cadence); templating those patterns boosts consistency.

How I Built It

Layer	Tech & Approach
Front-end	React + Vite, Tailwind for UI, Ant Design components
Auth & Data	Supabase (Postgres, Row-Level Security)
AI Pipeline	OpenAI o4-mini functions 1️⃣ Generate structured outline JSON 2️⃣ Expand to full draft with brand-voice tokens
Content Assistant	Retrieval-augmented chat referencing saved outlines/drafts
Hosting / CI	Netlify continuous deployment

Challenges

Issue	Fix
Token overflow from verbose brand briefs	Extractive summarizer before prompt
Tone drift in early drafts	Weighted persona sentiment + seeded style examples
Multi-tenant isolation	Granular Supabase RLS + service-role edge functions
Generation latency > 12 s	Batched prompts & enabled streaming (now < 7 s)