This project introduces AgriGen, a multilingual agricultural Q&A system powered by prompt-tuned, lightweight generative AI models such as Phi-2 and TinyLLaMA. Designed with a vision toward future offline deployment, AgriGen currently operates in a hybrid mode and is being optimized to run on resource-constrained devices without requiring continuous internet connectivity. This approach aims to support farmers in remote or low-connectivity areas by enabling access to critical information on crop planning, fertilizer usage, yield improvement, and disease management through natural language interactions. Farmers can ask region-specific questions such as “What should I grow in Gujarat in July?” or “How do I treat yellow spots on chili leaves?” and receive accurate, context-sensitive responses in their chosen language, including English, Hindi, or Telugu. AgriGen's prompt-tuned local language models ensure that answers are tailored to local agricultural conditions. Additional features like speech output and translation support further enhance accessibility and usability. AgriGen builds on our prior work, “Smart Agriculture Advisor: A Machine Learning Approach to Precision Farming” (IEEE AIIoT 2025), which focused on predictive analytics for crop selection, fertilizer recommendation, and disease detection. By integrating generative AI, multilingual support, and plans for offline capability, AgriGen advances this foundation to enable more intelligent and inclusive interactions in agriculture. In real-world rural contexts, solutions must be affordable, adaptable, and easy to use. AgriGen addresses these challenges by merging state-of-the-art generative AI techniques with practical, underserved farming needs. This aligns with current trends in low-resource language models, AI for social good, and multimodal systems, positioning AgriGen as a significant step toward democratizing smart farming worldwide.