Rhetro: AI Podcast Application

A cutting-edge AI podcast application that simplifies content creation. It converts text into high-quality speech and generates eye-catching thumbnails using advanced AI models. Perfect for both seasoned podcasters and beginners, Rhetro streamlines the production process, letting you focus on delivering engaging content.

• Next.js 14 (App Router)
• TailwindCSS
• Shadcn UI
• Google TTS
• Gemini
• Clerk
• Convex
• Sentry
• Databuddy (analytics)

• Convert your written content into high-quality, natural-sounding speech, making podcast production faster and more efficient.
• Generate visually appealing and professional-looking thumbnails for your podcast with the help of advanced AI models, ensuring your podcast stands out.

Rhetro is split between a Next.js frontend and a Convex backend-as-a-service. There is no custom server — all server-side logic runs as Convex functions (queries, mutations, actions, HTTP routes).

Browser ──► Next.js (App Router)                   ──► Convex (DB + Storage + Actions) ──► Google APIs (TTS / Gemini)
                  │                                          ▲
                  │                                          │
                  └── Clerk (auth + hosted UI) ──► svix webhook ──► Convex /clerk

Frontend — app/

• Route groups: (auth) for /signin, /signup; (root) for the authenticated shell (create, discover, podcasts, profile).
• middleware.ts protects every non-public route via Clerk.
• Root layout wires up ConvexClerkProvider (Convex client authenticated with Clerk JWTs), AudioProvider, and Databuddy analytics.

Backend — convex/

• schema.ts — podcasts (with search indexes on author, title, description) and users tables.
• podcasts.ts, users.ts — CRUD, search, view counting, top-creator queries.
• files.ts — issues pre-signed upload URLs for Convex Storage (audio + thumbnail blobs).
• googleTTS.ts — Convex action calling Google Cloud TTS for MP3 synthesis.
• gemini.ts — Convex action calling Gemini gemini-3-pro-image-preview for thumbnails.
• http.ts — POST /clerk HTTP route that verifies svix signatures and mirrors Clerk user.created/updated/deleted events into the users table.
• auth.config.ts — trusts the Clerk JWT issuer so ctx.auth.getUserIdentity() works inside queries/mutations.

Observability — Sentry is wired through next.config.mjs (withSentryConfig) plus sentry.{client,server,edge}.config.ts and instrumentation.ts.

Pushing to main deploys the Next.js app to Vercel. The Convex backend is deployed separately with npx convex deploy (typically wired into the same CI step).

Two processes need to run side-by-side: the Next.js dev server and the Convex dev server.

npm install
npx convex dev   # in one terminal — pushes convex/ to your dev deployment, regenerates _generated/
npm run dev      # in another terminal — starts Next.js on http://localhost:3000

Copy .env.example to .env.local and fill in:

Convex actions also need GOOGLE_GENERATIVE_AI_API_KEY — set this in the Convex dashboard (Environment Variables), not in .env.local, since the actions execute on the Convex backend, not in Next.js.