AgroYachay: An open-source IoT and large-language-model platform supporting agronomic and economic decision-making for Andean smallholders

Torres-Cruz F.; Vilca-Solorzano R.A.; Yana-Yucra D.M.; Ibañez-Quispe V.; Fuentes-Navarro E.L. · SoftwareX (2026) · DOI: 10.1016/j.softx.2026.103037

Resumen

Low-cost agricultural IoT is increasingly affordable, yet turning raw telemetry into practical decisions for smallholders remains a challenge, especially in Andean regions with limited technical support and language access. We present AgroYachay, an open-source platform integrating IoT sensing, large language models and explainable analytics: continuous field monitoring, AI-assisted pest and disease diagnosis, crop-specific recommendations, weather-informed scheduling, and transparent estimation of crop productivity and economic returns—through a multilingual interface (Spanish, Quechua, Aymara and English) that lowers the language barrier for indigenous farmers. Built on ESP32 nodes, a Flask/PostgreSQL backend, a self-hosted open vision model and a cloud LLM layer, it is the open cloud counterpart to AgroCommish. On 509 laboratory and field potato-leaf images the pest module detected disease with 100 % sensitivity but only 62 % specificity and macro-F1 of 0.55, supporting its use as a field triage aid rather than a fine-grained diagnostic. Released under the MIT licence with an archived DOI.

Instituciones: Universidad Nacional Agraria La Molina; Universidad Nacional del Altiplano