Farmer.CHAT - A GPT-powered, multi-lingual WhatsApp bot
summary
A WhatsApp chatbot pulls Ministry of Agriculture videos and PDFs plus Digital Green transcripts into a vector database, transcribes regional audio with Bashini/AI for Bharat, translates to English for GPT, then summarizes and returns localized text and audio via Google Text-to-Speech. Demo handles voice notes, detects out-of-scope queries, and shows source citations. Tested by about 50 extension agents (not yet wide farmer rollout); main problems are hallucinations, unreliable video URL delivery, and search relevance when conversation history grows.
AI-written summary · about
chapters (11)
- 0:00Farmer.CHAT project
- 1:53Vision for WhatsApp bot
- 3:17Live demo
- 5:15Technical architecture
- 8:42Prompt engineering
- 11:11Testing with agents
- 13:22Video retrieval challenges
- 16:05Language navigation issues
- 17:10Context-sensitive advice
- 18:13Context limitations
- 19:37Customer strategy
related talks
28mScrew the Browser! Build ChatGPT via SMS
36mAI for Government Efficiency with ChatGPT: From Hype to Reality
20mBoost Your Mobile App with ChatGPT
54melevating customer experience w/ conversational AI
24mVirtual assistant for hiring last-mile workforce
35mExtensionBot: A Domain Specific Gen AI Tool for the Cooperative Extension Service
39mBuilding Your Own Bot with GPT: Harnessing the Power of AI...
26mUsing Conversational AI to combat the impact of COVID-19