Zenit

Zenit is an edtech web app: marketing site, authenticated student and teacher dashboards, AI-assisted study tools (chat, roadmaps, quizzes, podcasts, visualizer videos), gamification (coins, XP, study pass), and Razorpay checkout. It is a Next.js 16 monolith with MongoDB persistence and NextAuth (Google).

Stack

Layer	Choice
Framework	Next.js 16 (App Router), React 19, TypeScript
Styling	Tailwind CSS v4, Radix UI, Framer Motion
Data	MongoDB (mongodb driver)
Auth	NextAuth v4, Google OAuth
AI	Google Gemini (@google/genai / @google/generative-ai), OpenAI where configured
Payments	Razorpay
Voice	Vapi (mentor calls), ElevenLabs (TTS)

Project conventions and AI agent notes live in AGENTS.md (this repo targets a newer Next.js than classic tutorials).

Prerequisites

• Node.js 20+ (recommended)
• npm (or compatible client)
• MongoDB URI (Atlas or local)
• Google OAuth client (Web application) for sign-in

Quick start

npm install

Create .env.local in the repo root. Set at least the variables in Required below, then:

npm run dev

Open http://localhost:3000.

npm run build
npm start

Production build and start.

npm run lint

Run ESLint.

Environment variables

Create .env.local in the repo root (never commit secrets). Grouped by concern; only Required is needed for a minimal local run with auth + DB + core AI chat.

Required (typical local dev)

Variable	Purpose
MONGODB_URI	MongoDB connection string
MONGODB_DB_NAME	Database name (many routes default to lr_hacks if unset; align with your cluster)
NEXTAUTH_SECRET	NextAuth signing secret
NEXTAUTH_URL	App origin, e.g. http://localhost:3000
GOOGLE_CLIENT_ID	Google OAuth client ID
GOOGLE_CLIENT_SECRET	Google OAuth client secret
GEMINI_API_KEY	Google AI / Gemini (chat, roadmaps, visualizer, many API routes)

Recommended for full URLs and client features

Variable	Purpose
NEXT_PUBLIC_APP_URL	Public site URL (metadata, some API headers)

Payments (shop)

Variable	Purpose
NEXT_PUBLIC_RAZORPAY_KEY_ID	Razorpay key id (client)
RAZORPAY_KEY_SECRET	Razorpay secret (server)

AI mentor (phone calls via Vapi)

Variable	Purpose
VAPI_PRIVATE_API_KEY	Vapi server key
VAPI_PHONE_NUMBER_ID	Vapi outbound number id
NEXT_PUBLIC_VAPI_GENERATE_ASSISTANT_ID	Default assistant id (optional override from client)

Optional / feature-specific

Examples used across the codebase: OPENAI_API_KEY, DEEPSEEK_API_KEY, OPENROUTER_API_KEY, ELEVENLABS_API_KEY, GEMINI_MODEL, GEMINI_VEO_MODEL. Routes fail gracefully or return errors if a key is missing for that feature.

Project layout (short)

Path	Role
src/app/	App Router: (marketing), dashboard/*, teacher/*, auth/*, api/*
src/components/	UI, student-dashboard/, ui/ primitives
src/lib/	Mongo client, auth config, AI helpers, quiz/roadmap logic
src/config/	Copy, nav, tool flags (dashboard-chat-tools, dashboard-chat-persistence, etc.)
src/models/	User-oriented types aligned with Mongo documents
public/	Static curriculum assets (roadmaps, cheat sheets, visualizer media)

Dashboard chat (study chat) — in detail

Route: /dashboard/chat. Primary UI: src/components/student-dashboard/DashboardChatView.tsx. Backend for plain Q&A: POST /api/dashboard-chat (src/app/api/dashboard-chat/route.ts), using Gemini with a system prompt tuned for students and Zenit’s Visualizer (so the model does not treat random YouTube links as the main answer when an in-app clip is appropriate).

Composer and layout

• Hero state when there are no messages: headline, Claude-style composer (src/components/ui/claude-style-chat-input.tsx) with attachments, optional Do vs Chat mode, and idea carousel presets (summarize, quiz me, study plan, “I want to learn …”, etc.).
• Thread state after the first message: scrollable transcript, same composer docked at the bottom, Zenit-branded bubbles.

Markdown and rich bubbles

Assistant and user bubbles render Markdown (react-markdown, GFM, remark-breaks, rehype-highlight, GitHub-dark code styling). Rich assistant cards exist for roadmaps, podcasts (audio + link to library), generated videos (inline player + optional topic breakdown from /api/generate-topic-info), and quizzes (deep link into AI Quiz).

Roadmap wizard (inline, no separate page required)

If the user’s text matches “learn X” / roadmap-style intents, the chat runs a short agent flow: fetch three focus paths (/api/roadmap-focus-options), then inline chips for focus → level → (optional) JS/TS stack for web topics, then /api/ai-generate-roadmap and /api/save-roadmap, and finally a roadmap card linking to /dashboard/roadmap.

Quick actions (toolbar chips)

From the composer row users can trigger (with coin checks where applicable, aligned with the rest of the app):

Chip	Behavior
Roadmap	Prefills “I want to learn …” in the composer.
Podcast	Requires a PDF attachment; spends coins → /api/generate-podcast → podcast bubble.
Quiz	PDF or typed topic; coins → /api/generate-quiz → quiz bubble.
Video	PDF or typed topic; coins (visualizer) → /api/generate-animation with videoIntent → video bubble + optional topic panel.
My videos	Opens a gallery dialog (GET /api/visualizer/videos); picking a row inserts a saved video message (no extra coin spend).
Mentor (end chip)	Opens a dialog with phone + /api/call (Vapi), same as /dashboard/ai-mentor.

“Agentic” video from natural language

If the latest user message matches configurable regex intents (src/config/dashboard-chat-tools.ts → agenticVideo), the chat skips a generic text reply and instead builds a merged prompt from recent user lines + recent assistant explanations, then runs the same Visualizer pipeline (coins + animation + optional topic info). Example intent: asking to visualize with a video after discussing a topic.

AI mentor call from chat text

Config-driven flow (mentorCallFromChat): phrases like “call me” / “I have a doubt” → assistant asks for E.164 number → next message triggers POST /api/call and XP award, mirroring the dedicated mentor page.

Persistence (MongoDB)

• Collection (default): dashboard_chat_sessions — see src/config/dashboard-chat-persistence.ts (listLimit, maxMessagesStored, autoLoadLatestOnMount, saveDebounceMs).
• API: GET/POST /api/dashboard-chat/sessions, GET/PATCH /api/dashboard-chat/sessions/[id] — scoped by NextAuth user email.
• UI: top bar with session picker (recent titles + relative time), New chat (creates an empty session), loading state while the list hydrates. Messages are debounced to the server after edits.

Config knobs (high level)

File	Role
src/config/dashboard-chat-tools.ts	Mentor call copy/cancel regex, agentic video patterns and copy, Visualizer gallery toggle, topic panel after video, mentor assistant id (NEXT_PUBLIC_VAPI_GENERATE_ASSISTANT_ID).
src/config/dashboard-chat-persistence.ts	Mongo collection name, caps, auto-load latest, save debounce.

---

Other features (brief)

Marketing and auth

• Landing (src/app/(marketing)): public home, pricing-style sections, FAQ, navbar/footer.
• Auth: Google sign-in via NextAuth; role selection and protected dashboard routes (src/app/auth, src/lib/auth-config.ts).

Student dashboard (beyond chat)

• Home (/dashboard): stats, quick links, activity graph, entry to tools.
• Roadmap (/dashboard/roadmap): full roadmap experience linked from chat-generated paths.
• AI Notes (/dashboard/ai-notes): AI Quiz and AI Podcasts flows (upload / generate, libraries).
• Visualizer (/dashboard/visualizer): prompt → coin-gated educational video (/api/generate-animation), topic info panel, gallery of saved clips per user.
• AI Mentor (/dashboard/ai-mentor): phone number + Vapi outbound call UI.
• AR Learning (/dashboard/ar-learning): AR-oriented chat/experiences where configured.
• Online code / interview prep (/dashboard/onlineCode, nested routes): tracks, questions, system-design markdown hubs under public/, flat markdown readers.
• Shop (/dashboard/shop): Razorpay orders, coin packs, study pass tiers (src/config shop + study pass).
• Leaderboard, job automation entry, and other dashboard links as wired in src/config/student-dashboard.ts and the shell.

Teacher area

• /teacher/*: tools such as diagrams, slides, question papers, stats, prompt enhancement — backed by dedicated /api/teacher/* routes and Gemini where applicable.

Gamification and user data

• Coins, XP, study pass checks appear in API routes (e.g. spend-coins, award-xp) and Mongo-backed user documents (src/models/User.ts patterns).

Static and generated assets

• Large public curriculum (e.g. CN, DBMS, system-design, roadmap JSON) for interview prep and learning paths.
• Podcasts, PPT, PDFs, animations produced through various /api/generate-* and storage patterns (Cloudinary, Mongo metadata, etc., depending on feature).

Realtime and integrations

• Socket.io client where quiz battles or live features need it.
• ElevenLabs, DeepSeek/OpenAI routes where enabled by env.

License

Private project ("private": true in package.json). Adjust if you open-source the repo.