Building an autonomous AI agent is easy. Keeping it alive 24/7 for four months on a ₹0 budget is the real engineering problem

Description and use

A 24/7 autonomous AI newsroom built on free-tier services with fallback LLMs and serverless backend ran continuously for four months without cost.

Problem

Maintaining a fully autonomous AI news generation pipeline on free-tier services without human intervention.

Outcome

The redesigned XeL Studio pipeline operated continuously for four months on free-tier services, demonstrating that combining multiple free services with fallback logic can sustain a production AI agent.

Workflow

• 1. Every 30 minutes, scrape tech news via Searchwala.
• 2. Use two LLMs (Groq and Cerebras) behind a fallback router to generate article text.
• 3. Generate cover image with Cloudflare.
• 4. Store article in Supabase (PostgreSQL) and JSON for pipeline state.
• 5. Frontend reads from Firebase via Vercel-hosted Next.js app.
• 6. Daily cleanup job runs on Supabase free tier.
• 7. Heavy AI and scraping tasks run on GitHub Actions scheduled workflow to avoid Vercel timeouts.

Evidence

Two LLMs, one pipeline Groq and Cerebras run behind a fallback router. If one hits a rate limit or goes down, the other takes over.

Source

Vercel for the frontend, GitHub Actions for the heavy lifting The Next.js + TypeScript frontend is hosted on Vercel. But serverless timeouts kill long jobs, so AI generation and Python scraping run on GitHub Actions as a scheduled backend.

Source

Three databases, each with one job Firebase: frontend reads. I built the UI layer around it because the load speed is excellent. Supabase (PostgreSQL): long-term article storage on its own free tier, so the archive can keep growing. JSON: the news pipeline only.

Source

Media at ₹0 Cloudflare generates the cover images. Cloudinary stores and delivers them.

Source