KrishiSaathi AI
Link to open source: https://github.com/krabhi75/LIAA.git
KrishiSaathi AI is a voice first agricultural support and CRM platform that helps farmers access agricultural services through natural conversations in Hindi, Hinglish and English.
Instead of asking farmers to type information or navigate complex applications, KrishiSaathi allows them to simply speak about their problem. The AI understands the conversation, remembers the context, asks dynamic follow up questions, automatically captures farmer responses into structured CRM data, retrieves relevant real time information such as weather, and provides the appropriate next action.
When the AI is uncertain or the problem requires human expertise, KrishiSaathi automatically creates a structured agricultural case and escalates it to an expert, ensuring the farmer does not have to repeat the entire problem.
Core Workflow
Farmer Voice → AI Understanding → Dynamic Questions → Automatic Data Capture → Weather and External Data → AI Assessment → Action → Expert Escalation → CRM → Follow Up
Key Capabilities
- Natural real time voice conversations
- Hindi, Hinglish and English support
- Code switched conversations
- Session level conversational memory
- Dynamic questioning based on farmer responses
- Automatic extraction of farmer and crop information
- Location based weather integration
- Agricultural information retrieval
- AI confidence and uncertainty handling
- Inbound farmer calls
- Outbound farmer campaigns
- Automatic response capture during calls
- Structured case creation
- Agricultural expert escalation
- Farmer CRM and conversation timeline
- Callback and follow up workflows
Example
A farmer says:
"Meri wheat crop ke leaves yellow ho rahe hain."
Instead of immediately giving a diagnosis, KrishiSaathi asks contextual questions such as when the problem started, which part of the field is affected and whether any fertilizer or spray was recently used.
The system automatically captures this information in the CRM, retrieves weather conditions for the farmer's location and uses the information as additional context.
If the AI does not have enough confidence to provide a safe recommendation, it creates a structured expert case containing the farmer's information, crop details, symptoms, conversation history, weather context and AI assessment.
