KV KisanVaani Voice · Satellite · RSK

Code for Communities · Track 4

Smart water, crop and advisory decisions over a missed call.

A voice-and-SMS platform for small farmers that recommends crop portfolios, detects monsoon dry-spell risk, and routes uncertain crop-health cases to Rythu Seva Kendras with plot evidence attached.

Selected solution

A complete field-to-RSK operating system

Farmer access

Missed call, IVR, SMS and WhatsApp voice

No app dependency. A farmer can call, use a neighbor's smartphone, or ask an RSK operator to log the case.

Agronomy brain

Crop portfolio + water-risk engine

Ranks safe, balanced and high-return crops using soil, satellite, rainfall, groundwater and market signals.

RSK operations

Expert queue and district early warning

Low-confidence photo/voice cases go to RSK staff with plot history, then return as voice advice and training labels.

End-to-end web form

One farmer, one plot, one advisory workflow

Fill the form once. The page generates the IVR transcript, crop portfolio, dry-spell alert, SMS fallback, crop-health triage and RSK queue without redirecting anywhere.

Farmer intake

Voice-first advisory form

Ready

Farmer: Ee varsham lo ye pantalu veyali?

KisanVaani: Submit the form to generate plot-specific advice.

SMS fallback

Advice summary will appear here for 2G and feature-phone users.

Monsoon radar sentinel

Sentinel-1/Sentinel-2 plot evidence

Wetness low
Low wetnessHealthy cropWater body

Smart crop engine

Risk portfolio recommendation

Explainable
Soil N lowpH 7.8Rainfall -34%GW deepMSP protected
86
Red gramSafe crop · low water demand · MSP support
71
MaizeBalanced · sow after 10 mm effective rain
38
PaddyAvoid · high irrigation and seed-loss risk

Advisory and dry-spell alert

Actionable economics

78% risk

No effective rain is forecast for 7 days. Delay sowing until the next 10 mm rain event and save about Rs 3,800 in seed risk.

Crop stage
Sowing
ET deficit
18 mm
Alert rule
2/3 signals

RSK human-in-the-loop

Expert triage queue

2 open
OpenYellow leaves · red gram · flowering

AI confidence 42%. Cannot separate nutrient stress from early pest damage.

Cluster5 nearby yellow-leaf reports

Possible early warning near Kothapalli and Ramapuram within 5 km.

Google Cloud deployable architecture

Built for a real district pilot, not a one-off app

KisanVaani architecture diagram

Access

Exotel/Twilio missed call, IVR, SMS, WhatsApp voice, and RSK kiosk.

AI

Bhashini/AI4Bharat ASR/TTS, Vertex AI/Cloud Run scoring, confidence-gated diagnosis.

Data

Soil Health Card, IMD/CHIRPS, Sentinel-1/2 via Earth Engine, CGWB, Agmarknet, MSP.

Ops

Cloud SQL/PostGIS, Pub/Sub alerts, static hosting, and Cloud Monitoring.

Production implementation

From prototype to one-district pilot

Week 1

District data bootstrap

Load village boundaries, Soil Health Card records, crop calendars, RSK roster, IMD/CHIRPS history and Agmarknet/MSP baselines.

Week 2

Voice channel integration

Connect Exotel/Twilio missed-call IVR, SMS templates, Bhashini/AI4Bharat ASR/TTS and dialect phrase dictionaries.

Week 3

Remote sensing pipeline

Generate village/plot wetness, NDVI/NDWI and rainfall anomaly features from Sentinel-1, Sentinel-2 and CHIRPS.

Week 4

RSK closed-loop pilot

Train RSK users, start confidence-gated tickets, track callback completion and publish district pest/stress heatmaps.

Deployment package

Everything is visible on this page

RunbookUse the form, generate advice, attach crop-health note, close RSK ticket.
BlueprintCrop engine, dry-spell alert, diagnosis and expert loop are embedded in the UI.
Cloud pathCloud Run APIs, Pub/Sub events, Cloud SQL/PostGIS and Earth Engine features.
CI/CDStatic site is deployed automatically by the repository workflow after each push.