VIDYA_AI

Vidya AI is a Hinglish-first voice and chat senior buddy that can become a tutor when the student asks for help.

The repo now includes a working Next.js app shell, typed API routes, browser voice input, spoken replies, an adaptive conversation orchestrator, Motion specialist content packs for relevant questions, GPT-generated Motion visuals with static fallback, and an offline autoresearch-style research loop for improving prompts and learning flows.

Current Product Docs

• Full blueprint: [docs/VIDYA_AI_MVP_Blueprint.md](C:\Users\ADITYA GUPTA\VIDYA_AI\docs\VIDYA_AI_MVP_Blueprint.md)
• PRD summary: [docs/PRD.md](C:\Users\ADITYA GUPTA\VIDYA_AI\docs\PRD.md)
• Architecture summary: [docs/ARCHITECTURE.md](C:\Users\ADITYA GUPTA\VIDYA_AI\docs\ARCHITECTURE.md)

Run It

npm install
npm run dev

Useful checks:

npm test
npm run build
node scripts/research-runner.js

LiveKit Realtime Voice

This project now includes a LiveKit realtime voice path:

• Frontend token route: src/app/api/livekit/token/route.ts
• Voice and chat UI: src/features/tutor/client/TutorExperience.tsx
• Python agent worker: agents/livekit/vidya_agent.py
• Setup guide: [docs/LIVEKIT.md](C:\Users\ADITYA GUPTA\VIDYA_AI\docs\LIVEKIT.md)

Required environment variables:

LIVEKIT_URL=
LIVEKIT_API_KEY=
LIVEKIT_API_SECRET=
OPENAI_API_KEY=
VISUAL_GENERATION_PROVIDER=openai
OPENAI_IMAGE_MODEL=gpt-image-1.5

Core Files

• AGENTS.md - repo-specific instructions for Codex and other coding agents
• RULES.md - non-negotiable engineering and safety rules
• WORKING-CONTEXT.md - active sprint context, assumptions, and next actions
• .mcp.json - MCP server configuration placeholder
• .env.example - environment variable template
• research/ - offline prompt and lesson iteration loop inspired by karpathy/autoresearch

Project Layout

.
|-- AGENTS.md
|-- RULES.md
|-- WORKING-CONTEXT.md
|-- README.md
|-- .env.example
|-- .gitignore
|-- .mcp.json
|-- docs/
|-- scripts/
|-- src/
|   |-- app/
|   |-- core/
|   |-- features/
|   |-- lib/
|   `-- types/
`-- tests/
    |-- integration/
    `-- unit/

Next Steps

1. Replace mock provider adapters with real model, STT, and TTS integrations.
2. Expand the adaptive conversation engine from deterministic specialist packs to structured model outputs.
3. Add real student-session telemetry and acceptance testing.
4. Use research/ to improve buddy-style explanations and specialist chapter packs.