Looking for feedback on OpenVidya: an open-source AI classroom layer for NCERT/CBSE [R]

OpenVidya is an open-source fork of OpenMAIC, designed to adapt multi-agent AI classroom generation specifically for Indian education. Instead of treating learning as a generic slide/chat experience, it grounds lessons in NCERT/CBSE curricula using structured JSON registries. Key features include concept dependency graphs that ensure prerequisite-aware lesson sequencing, board-style questions with varying difficulty, traps, and explanations, and a lab experiment registry that tracks apparatus, objectives, and common mistakes. The system operates in five distinct pedagogy modes: Teacher Narration, Story Quest, Exam Dojo, Lab Without Walls, and Rapid Revision, each with mode-specific prompting for outline generation, slide creation, and runtime narration.

The project’s thesis is that an AI tutor for India must understand exam patterns, local examples, curriculum structure, and how students revise, practice, and get stuck—not just translate content. The developer is actively seeking feedback on architecture (whether this grounding approach is optimal for lesson generation), product focus (which user segment—students, teachers, coaching centers, or edtech builders—should be prioritized first), evaluation metrics (how to measure improvement over generic AI tutors), and dataset expansion (suggestions for open Indian curriculum/question resources). The project is early-stage and open to contributions.