RSS Ingest Pipeline
48 automated runs per day. URL-based category detection, domain-level overrides (e.g. mirchi9.com → Cinema), AI cascade: Cerebras → Gemini → SambaNova.
Back to ProductsTelugu News. Audio-First.
Case Study · Kshana Vartha
Live · Android · Telangana & Andhra PradeshTelugu News. Audio-First.
Swipe to Read.
Kshana Vartha is a hyperlocal Telugu news app we built — AI-summarized articles, TTS audio narration, a WhatsApp reporter newsroom, and a SwipeReader UX. Running on ₹0/month infrastructure.
2,000+
Articles Live
9
Categories Covered
48
AI Runs / Day
₹0
Monthly Infra Cost
6 Wks
Idea to Launch
Sector
Regional Media · Telugu News
Deliverable
Android App · AI Pipeline · Admin Panel
Audience
Telugu-speaking Andhra Pradesh & Telangana
Status
Live · 48 AI runs / day
The Challenge
Telugu speakers left behind?
National news apps optimise for Hindi and English. Local Telugu news — especially at village and mandal level — barely exists online.
The Problem
- Telugu speakers ignored by national apps focused on Hindi and English
- Long articles nobody reads — no time, no context
- WhatsApp forwards: unverified, misleading, no source
- No audio — millions consume news by listening, not reading
- Building a news platform costs ₹50,000+ per month in cloud servers
- Village and mandal-level news gets zero coverage
Kshana Vartha's Answer
- Dedicated Telugu-only feed — hyperlocal, district-level coverage
- 2-sentence AI summaries — most interesting fact leads every article
- Human reporters file via WhatsApp, verified before publishing
- TTS audio for every article — listen while commuting or farming
- Runs entirely on free-tier infrastructure — ₹0 per month
- 9 categories including Village, Farming, Weather, Jobs, and Schemes
What We Built
Eight systems. One seamless product.
From AI pipeline to Android app — everything built from scratch, in 6 weeks.
AI Summarization
Hook-first Telugu summaries — most interesting fact leads. English content forced into Telugu script. Lean prompt when category is confident; full prompt otherwise.
TTS Audio
Every article gets an audio clip. Text sanitized before generation so narration sounds natural. Articles deduplicated by content hash — no repeat processing.
WhatsApp Newsroom
Human reporters file stories via WhatsApp using the official Meta Cloud API. Admin panel receives, reviews, and approves before articles go live.
Fast Track Queue
Breaking news queue with one-click approval. Verified articles bypass the standard pipeline and go live in seconds — no waiting for the next ingest run.
SwipeReader App
Vertical swipe UX — like Reels but for news. Inline audio player. Category and location filters. Archive going back 30 days. Built with React + Capacitor for Android.
FCM Push Notifications
2-tier system: Tier 1 (Politics, Schemes, Farming, Weather, Jobs, Village) gets push alerts. Tier 2 (Cinema, Sports, General) stays silent. 3-hour gap guard prevents notification fatigue.
Admin Panel
FastAPI backend. Draft/verify workflow, R2 media upload, reporter management, market price entry, WhatsApp webhook integration, and sync to Cloudflare R2 CDN.
Two Interfaces
Reader-first app.
Reporter-first admin.
The SwipeReader serves the audience. The admin panel serves the newsroom. Both built from scratch.
Kshana Vartha SwipeReader
Android · Free download
- Vertical swipe through articles — like Reels for news
- Inline audio player — tap to listen to the AI-narrated summary
- 9 category filters including Village, Farming, and Local Schemes
- Geolocation filter — news from your district first
- Archive going back 30 days — never miss a story
- FCM push notifications — breaking news delivered instantly
- Offline-ready — bundled local libraries, zero CDN requests at startup
Kshana Vartha Admin Panel
Web · For editors and reporters
- WhatsApp-based reporter intake — no special app needed for reporters
- Draft review and one-click article verification
- Media upload to Cloudflare R2 with automatic CDN delivery
- Reporter profile management and byline tracking
- Fast Track queue for breaking news — approve in seconds
- Market price entry for commodity and mandi rates
- Feed sync: push articles to R2 → app refreshes instantly
📱 Reporters file news from their phones via WhatsApp — no training or special app required.
Infrastructure
A full media platform.
₹0 per month.
Most news platforms spend ₹50,000+ per month on servers, CDN, and messaging. Kshana Vartha is built entirely on free-tier services — without compromising on speed, reliability, or scale.
| Service | Role | Monthly Cost |
|---|---|---|
| Cloudflare R2 | Media CDN + feed delivery | Free (10 GB/month) |
| Firebase Spark | Auth + Firestore + FCM push | Free tier |
| Cerebras API | Primary AI summarization | Free (open access) |
| Google Gemini | Secondary AI summarization | Free tier |
| SambaNova | Fallback AI summarization | Free tier |
| cron-job.org | Ingest trigger (48× per day) | Free |
| GitHub Actions | Pipeline execution | Free (2,000 min/month) |
| Capacitor / Play Store | Android app delivery | Free tier |
| Total monthly infrastructure cost | ₹0 | |
Why Kshana Vartha
How it compares
National news apps aren't built for Telugu audiences. Kshana Vartha is.
| Feature | National News App | Kshana Vartha |
|---|---|---|
| Language | Hindi / English only | Telugu only — no dilution |
| Audio Narration | None | Every article, TTS-generated |
| Local Coverage | State / national level only | Village & mandal level |
| Monthly Infra Cost | ₹50,000+ | ₹0 (free-tier stack) |
| AI Summaries | None | Hook-first, 2 sentences in Telugu |
| Push Notifications | Generic breaking news | Category-targeted, 2-tier |
| Human Newsroom | Wire copy only | WhatsApp reporter network |
| Time to Launch | 6+ months | 6 weeks |
Build Timeline
Idea to live app in 6 weeks
AI pipeline, admin panel, Android app, push notifications — all built and shipped.
1
Week 1
Architecture
Free-tier stack chosen: R2, Firebase, Capacitor, cron-job.org
2
Week 2
Ingest Pipeline
RSS parsing, AI cascade, category detection, TTS generation
3
Week 3
Admin Panel
WhatsApp newsroom, draft/verify workflow, R2 media sync
4
Week 4
SwipeReader App
Capacitor Android build, audio player, vertical swipe UX
5
Week 5
Push & Polish
FCM 2-tier notifications, geolocation filter, archive loader
Week 6
Launch
APK live on Android, 48 AI runs/day, reporters onboarded
Founder's Note
“We wanted to reach Telugu-speaking audiences with verified, audio-first local news — not more WhatsApp rumours. The challenge was doing it without a server bill. By combining free-tier AI APIs, Cloudflare R2, and Firebase, we built a full media platform for ₹0 per month. The AI pipeline runs 48 times a day, and every article reaches the audience in Telugu audio within minutes of publishing.
— Founder, Kshana Vartha · Data Strive Labs
Want a hyperlocal news app for your region?
We can build a similar platform — in any language, for any geography — on free-tier infrastructure. Tell us your audience.
Start a conversation