Background

Debbane Agri, a leader in agricultural innovation across Lebanon and the MENA region, needed a solution to improve how farmers and agronomists access practical agricultural knowledge. Farmers often struggle with:

• Identifying plant diseases accurately and on time.
• Accessing expert recommendations for treatment and crop management in a simple, reliable way.
• Bridging the knowledge gap between decades of accumulated agricultural expertise at Debbane Agri and farmers on the ground.

Traditional methods of diagnosis and advice were either time-consuming, required physical visits, or depended on scattered documentation that was difficult for end users to navigate. Debbane Agri sought an intelligent, scalable, and user-friendly solution to make expert guidance accessible instantly.

Proposed Solution

Nebulane designed and implemented a multimodal agricultural chatbot powered by Amazon Bedrock and AWS serverless services.

The solution works as follows:

1. Text-based Queries
• When users send a text-only prompt, it is forwarded to a Bedrock Agent.
• The agent uses Nova Pro as its foundation model to query a knowledge base built from Debbane Agri’s agricultural expertise and resources.
• The system then provides a clear and contextually relevant response to the farmer.
2. Image-based Queries
• When users upload an image (e.g., a tomato leaf with symptoms), it is sent to an AWS Lambda function.
• The Lambda function generates an embedding of the image and compares it against a vector database of plant disease embeddings.
• The most relevant disease match is returned to the Bedrock Agent.
• The agent then provides the user with the disease diagnosis and a recommended treatment or crop management practice.

This combination of text + image processing creates a comprehensive virtual agricultural assistant, empowering farmers to quickly diagnose problems and apply the right solutions.

Metrics for Success

The success of the chatbot solution is measured through:

1. Accuracy & Relevance
• High precision in matching crop disease images to the correct diagnosis.
• Reliable and contextually accurate answers to text-based agricultural questions.
2. User Adoption & Engagement
• Number of farmers actively using the chatbot.
• Frequency of repeat usage (trust and reliability of the tool).
3. Operational Efficiency
• Reduction in time spent by agronomists on repetitive queries.
• Faster response times compared to traditional consultation methods.
4. Impact on Farming Outcomes
• Improved speed of disease detection and treatment.
• Increased crop health and productivity through timely interventions.

Conclusion

With Nebulane’s implementation of a multimodal chatbot using Amazon Bedrock and AWS serverless services, Debbane Agri has made expert agricultural knowledge instantly accessible to farmers. The solution improves disease diagnosis, streamlines agronomist support, and boosts farming outcomes, bridging the gap between decades of expertise and real-time field needs across Lebanon and the MENA region.