A small Mastra-based agent + static UI that can:

• Fetch top 3 Product Hunt products for any given day.
• Search products using the Product Hunt Algolia index.
• Answer questions about launching on Product Hunt.
• Launch confetti on the front end.
• Host a dummy Product Hunt–style page on GitHub Pages and embed the agent there.

This folder contains:

• src/ — Agent, tools, a lightweight server (src/server.ts).
• web/ — Static site that looks like a PH page, with a chat widget and confetti.

Live demo: cometchat.github.io/ai-agent-mastra-examples/product-hunt-agent/web

Prereqs: Node 20+.

1. Install deps:

   • If needed, run npm install in the repo root.

2. Build the server code:

   • npm run build:server

3. Start the API server:

   • npm run serve
   • It listens on http://localhost:8787.

4. Open the static UI:

   • Open web/index.html in your browser (or serve web/ with any static server).
   • The UI calls the local API by default.

For real Product Hunt post data and agent answers:

• PRODUCTHUNT_API_TOKEN — Bearer token for Product Hunt GraphQL v2 (https://api.producthunt.com/v2/api/graphql). Required for the /api/top* endpoints to return live posts; if omitted, those endpoints respond with empty arrays.
• OPENAI_API_KEY — For the agent’s Q&A about launching.

Search note: Product search uses the public Product Hunt Algolia index with fixed public credentials (from the PH wiki) and does not require any env vars.

• GET /api/health → { ok: true }
• GET /api/top?limit=3 → { posts: PHPost[], first: number, order: 'VOTES' }
  • Top all-time by votes. limit (aka first) is clamped to 1–10.
• GET /api/top-week?limit=3&days=7 → { posts, first, days, order: 'RANKING', window: 'rolling-week' }
• GET /api/top-range?timeframe=today&tz=America/New_York&limit=3 → { posts, first, timeframe, tz, order: 'RANKING' }
  • Timeframes supported: today, yesterday, this-week, last-week, this-month, last-month, YYYY-MM-DD, or ranges like from:2024-08-01 to:2024-08-15.
• GET /api/search?q=term&limit=10 → { hits: Hit[], q, limit }
  • Backed by Product Hunt’s public Algolia index using this request:
    • URL: https://0h4smabbsg-dsn.algolia.net/1/indexes/Post_production?query={q}&hitsPerPage={limit}
    • Headers: X-Algolia-API-Key: 9670d2d619b9d07859448d7628eea5f3, X-Algolia-Application-Id: 0H4SMABBSG
• POST /api/chat → { reply: string } with JSON body { message: string }

CORS is enabled for demo usage.

• Push the web/ directory to your public repo and enable GitHub Pages (source: /(root) or /docs, or an action). If you use web/ as Pages root, set that in the repo settings.

• In web/index.html, set window.PH_AGENT_API to your deployed API server URL (Render/Vercel/Fly/etc.). Example:

  <script>
     window.PH_AGENT_API = 'https://your-api.example.com';
  </script>

• The page includes a chat widget with a Product Hunt–style icon, and a Launch button that fires confetti.

• Without PRODUCTHUNT_API_TOKEN, top posts endpoints return empty arrays; search still works using the public Algolia index.
• The chat agent uses OpenAI via @ai-sdk/openai and Mastra’s Agent. Provide OPENAI_API_KEY for real answers.
• This is a minimal demo; harden auth, rate limits, and error handling before production.

• npm run build:server — TS → dist/ for the server.
• npm run serve — Start the server at :8787.
• npm run dev / npm run start — Mastra CLI (if you prefer Mastra’s workflow).

This app is self-contained. Make the repo public (or copy this folder to a new public repo) to share the source code.

• Guide: docs/mastra-product-hunt-agent.md — Step-by-step on building, running, wiring frontend actions, and connecting with CometChat.