Skip to content

ankan1811/virtual_study_group

BranchesTags

Repository files navigation

Virtual Study Group

A full-stack web application for creating virtual study group spaces with real-time video conferencing, live chat, AI-powered doubt solving, companion messaging, and collaborative study tools.

Tech Stack

Layer Technology
Backend Node.js, Express, TypeScript
Frontend React 18, TypeScript, Vite
Database MongoDB (Mongoose)
Real-time Socket.IO
Video Calls Agora RTC SDK
AI Switchable: Google Gemini 2.5 Flash (default) / xAI Grok
RAG / Embeddings Gemini text-embedding-004 (768-dim vectors, free tier)
Cloud Storage Cloudflare R2 (S3-compatible, free tier) for summary persistence
State Management Redux Toolkit
Whiteboard @excalidraw/excalidraw (MIT, client-side, lazy-loaded)
Study Radio SomaFM internet radio streams (ambient/chill/electronic)
Podcasts Listen Notes API (300 free calls/month) — curated across 5 topics, JSON file cache, node-cron bi-weekly refresh
Styling Tailwind CSS, shadcn/ui, Framer Motion
Auth Stateless OTP (HMAC-SHA256) + Google OAuth + JWT + nodemailer
Rate Limiting express-rate-limit (per-user + per-IP, all thresholds via env vars)
News Feed Mock articles (AI / Tech / Productivity categories, 30-min cache)

Features

Authentication (OTP + Google OAuth)

  • Two sign-in methods: passwordless OTP and single-click Google OAuth
  • Google OAuth — "Continue with Google" button on the auth page. Uses @react-oauth/google on frontend to get an access token, backend verifies it against Google's userinfo API (POST /auth/google). Works for both new registrations and existing logins. Stores googleId on the User model for account linking. Google Client ID via VITE_GOOGLE_CLIENT_ID env var.
  • Passwordless OTP auth — no passwords stored. Users verify email ownership via a 6-digit OTP on every login/register
  • Stateless OTP — HMAC-SHA256 signed hash (no OTP stored in DB). Server generates OTP + hash, emails OTP to user, client sends back OTP + hash for verification
  • Configurable expiry via OTP_EXPIRY_MINUTES env var (default: 5 minutes)
  • Two-step frontend flow: enter email → receive OTP → enter OTP → authenticated
  • Resend OTP with 30-second cooldown, "Change email" back button
  • JWT token-based session after successful OTP verification
  • Protected routes for authenticated users
  • Persistent auth state via Redux + localStorage
  • JWT rehydration on page refresh (no re-login needed)
  • Logout — clears JWT from localStorage, disconnects socket, resets Redux state, redirects to login

Personal Rooms & Invite System

  • Every user has a personal room (user_{userId}) — no room IDs to remember
  • Shareable invite link — "Invite" button in the room call page copies a permanent link (/join/{roomId}) to clipboard with animated feedback. Anyone with the link can join (logged-in users only; unauthenticated users are redirected to login and auto-joined after auth)
  • Only study companions can invite each other directly via real-time socket invites
  • Real-time invite notifications with accept/decline overlay
  • "Enter My Room" one-click access from the home page and sidebar

Study Companions

  • Send, accept, and decline companion requests (friend system)
  • Search for companions by name or email
  • Real-time presence tracking (online/offline status)
  • Companion bar on home page with avatar initials and status rings
  • Companion popover with quick actions: Message, Invite to Room
  • Real-time companion request notifications via overlays

User Profile

  • Profile page accessible from the profile avatar dropdown ("My Profile")
  • Gradient banner with large avatar circle, name, and "Student" label
  • Default avatar picker — 5 themed avatars (Cool Guy, Scholar, Scientist, Artist, Astronaut) with emoji + gradient backgrounds. Camera icon overlay opens a modal picker with spring animations. "Use Initials Instead" option to revert to letter-based avatar
  • Avatar selection saved instantly to backend and Redux store. Persists across refreshes via API fetch on JWT rehydration (avatar not stored in JWT)
  • Avatar updates reflect live across all UI — Navbar top-right button, profile dropdown header, and profile page all read from Redux
  • Shared avatar constants in frontend/src/utils/avatars.ts (used by ProfilePage + Navbar)
  • Editable name and bio with inline edit/save mode
  • Stats row showing companion count and email
  • Education section — degree, institution, and year with inline editing. Displayed as a styled card with bold degree and secondary institution/year line
  • Projects showcase — up to 2 projects with title, description, and clickable project link (opens in new tab). Add/remove projects in edit mode with dashed "Add Project" button. Each project rendered as a mini card with "View" link pill
  • Work Experience — single entry with company name, role (displayed as indigo badge), duration, and description. Full inline editing with 4 input fields
  • All new sections use consistent Tailwind styling with section icons (GraduationCap, FolderGit2, Briefcase), Framer Motion stagger animations, and full dark mode support
  • Profile updates re-issue JWT and sync Redux state instantly
  • Extended User model with bio, avatar, education, projects[], and workExperience fields

Settings

  • Settings page at /settings accessible from the profile avatar dropdown ("Settings")
  • Appearance section — functional dark mode toggle (uses existing useDarkMode hook), display preferences link
  • Notifications section — push notifications toggle, sound effects toggle
  • General section — language, privacy & security links
  • About section — app version info
  • Account actions — Edit Profile button (navigates to /profile), Log Out button
  • Premium design with staggered Framer Motion entry animations, decorative background blobs, grouped setting cards with dividers
  • Auth-guarded — redirects to login if not authenticated (with JWT rehydration race condition handling)

Direct Messaging (DM)

  • Slide-in DM panel from companion avatars
  • Real-time messaging powered by Socket.IO
  • Deterministic socket room naming (dm_{sortedUserIds})
  • Message history stored in MongoDB (last 50 loaded on open)
  • Full message delivery pipeline: pending → delivered → read
    • 🕐 Clock = sending (optimistic, before server ack)
    • ✓ Single tick = delivered (saved to DB)
    • ✓✓ Double tick = read (recipient opened the panel)
  • Unread badge (green ring on avatar) persists across refreshes — restored from GET /dm/unread-counts on mount
  • dm:markRead socket event instantly notifies sender when recipient opens the conversation

Chats Page (Recent Conversations)

  • WhatsApp-style recent chats list accessible from the sidebar ("Chats" nav item)
  • Shows all DM conversations sorted by most recent message (descending)
  • Each row displays: companion avatar (initials), name, last message preview, relative timestamp, unread badge count
  • Bold styling for unread conversations, muted for read
  • Search bar to filter conversations by companion name
  • Clicking a chat row opens the DM panel as a slide-in overlay
  • Real-time updates: new messages bump conversations to the top via dm:receive socket event
  • Aggregation pipeline on backend groups messages by conversation partner

Persistent Notifications

  • Bell icon in navbar with real-time unread badge
  • Stored in MongoDB with a 10-day TTL — notifications survive page refreshes and logouts
  • Three notification types:
    • companion_request — someone sent you a companion request
    • companion_accepted — your companion request was accepted
    • room_invite — a companion invited you to their study room (including missed invites when offline)
  • Per-notification Read/Unread state — click to mark read, "Mark all read" bulk action
  • Real-time push via notification:new socket event when online
  • Fetched via GET /notifications on app load when offline delivery occurred
  • Animated dropdown with relative timestamps, per-type icons, and inline delete

AI Whiteboard

  • Full-page collaborative whiteboard — opens as a dedicated route (/whiteboard/:roomId) when the Whiteboard tab is clicked in the call room. Full drawing tools (shapes, text, freehand, arrows, etc.) powered by @excalidraw/excalidraw
  • Completely free — the whiteboard library is MIT-licensed and runs entirely client-side, AI analysis via Gemini free tier
  • Lazy-loaded via React.lazy() — whiteboard bundle only downloads when first opened, keeping initial page load fast
  • Real-time collaboration — all room participants see the same whiteboard live via Socket.IO
    • Drawing changes debounced at 200ms (client) and throttled at 100ms (server)
    • Echo-loop prevention ensures incoming sync doesn't re-trigger outgoing broadcasts
  • Built-in AI Assist sidebar — collapsible right panel (360px, Framer Motion spring) with:
    • "Explain This" button to analyze the entire whiteboard
    • Custom question input to ask specific questions about the drawing
    • Q&A history with teal-accented chat bubbles and Framer Motion animations
  • Toolbar — top bar with Back to Room, Clear whiteboard, Summary (generates + saves whiteboard summary to R2 + MongoDB), and AI Assist toggle
  • Smart AI payloads — sends compact text descriptions (element type, text content, dimensions) rather than raw JSON

Podcasts

  • Podcast discovery page at /podcasts accessible from the sidebar (Mic2 icon)
  • 5 topic tabs: Trending, AI, Tech, Business, Productivity & Tools — each lazily fetched on first tab visit
  • Listen Notes API integration (GET /podcasts/:topic) — uses best_podcasts endpoint for genre-based topics and search for AI
  • JSON file cache in backend/cache/podcasts/{topic}.json — persists across server restarts
    • Cache validity: checks if fetchedAt >= last Tuesday or Saturday midnight (no cron needed for correctness)
    • Refresh chain: fresh cache → Listen Notes API → stale cache → 3 mock fallback podcasts per topic
    • source field in response: "cache" | "api" | "stale-cache" | "mock" — frontend shows amber notice on stale/mock
  • Scheduled refresh via node-cron — runs at 02:00 UTC every Tuesday and Saturday (5 topics × 2 = 10 API calls/week ≈ 40/month, well within the 300/month free tier). Registered in server.ts on startup.
  • Sequential topic fetching with 1.5s gap to avoid API burst limits
  • Podcast cards: thumbnail, title, publisher, 3-line description, listen score badge, episode count, "Listen Now" external link → Listen Notes
  • Refresh banner: pulsing dot + "Fresh drops every Tue & Sat — stay ahead of the curve."
  • Animated tab bar with Framer Motion layoutId sliding gradient highlight per topic
  • Skeleton loading grid (8 ghost cards) while fetching; error state with retry button

Study Radio

  • Full-page radio player at /radio — browse and play curated internet radio channels from SomaFM
  • 8 curated channels — Groove Salad, Drone Zone, Lush, Space Station Soma, DEF CON Radio, Boot Liquor, Fluid, Groove Salad Classic (ambient, chill, electronic, Americana)
  • Global audio context — RadioContext with useReducer, manages HTMLAudioElement + Web Audio API AnalyserNode for real-time visualizations. State persisted to sessionStorage
  • Canvas-based audio visualizer — frequency bar animation with fallback to CSS animated bars
  • Floating MiniPlayer — bottom-right mini player visible on all pages except /radio. Play/pause, volume, mute, expand, and close controls
  • Accessible from sidebar — "Study Radio" nav item with headphones icon

Video Calling (Agora RTC)

  • Opt-in video calls — Agora RTC only starts when the user clicks "Start Video Call" in the room lobby. Chat, AI, and whiteboard tools are available immediately without consuming Agora hours
  • Agora App ID loaded from environment variable (VITE_AGORA_APP_ID)
  • Multi-user video grid supporting up to 5 participants
  • Camera toggle (on/off), microphone toggle (mute/unmute)
  • End call button (red PhoneOff) in the video grid to leave the call without leaving the room
  • Automatic remote user subscription

Real-time Chat

  • Live in-room messaging powered by Socket.IO
  • Uses a singleton socket instance shared across all components
  • Visual distinction between your messages, others' messages, and bot messages
  • Bot welcome message on room join
  • Linkify helper — URLs in bot messages render as clickable links (e.g., summary download links)
  • Summary bot messages get special styling (violet background, rounded border) distinct from regular bot messages
  • Opt-in chat persistence — chat messages are ephemeral by default (fresh every session). Users can save them to MongoDB on demand:
    • Inline "Save" button in the chat input bar — saves all messages to the database. Disabled when no unsaved messages exist; re-enables when new messages arrive after a save. Shows a brief "Saved" confirmation with checkmark
    • "Save your chats?" exit prompt — animated modal appears when leaving the room with unsaved messages. Options: "Save & Exit" (persists then navigates) or "Exit without saving" (discards). Prompt is skipped entirely if the user already saved all messages via the inline button
    • Browser tab close protection — native beforeunload dialog warns when unsaved messages exist
    • React Router navigation blocking — useBlocker intercepts sidebar/back button navigation and shows the save prompt
    • Messages are bulk-saved via POST /chat/bulk-save with a sessionId (UUID) grouping messages per save operation. Capped at 500 messages per request
    • Bot messages are excluded from saves and unsaved message counts

AI Doubt Solver

  • In-room AI panel accessible via the "AI Doubt" tab in the room lobby (no video call required)
  • Switchable AI provider via AI_PROVIDER env variable:
    • Gemini 2.5 Flash (default) — Google's free tier (250 requests/day, resets daily)
    • Grok 3 Mini — xAI's API (fallback option)
  • Both providers use the OpenAI-compatible SDK — no code changes needed to switch
  • Text input with full conversation history
  • Voice input via Web Speech API (browser mic button)
  • Styled Q&A cards with generic "AI Assistant" branding
  • Full dark mode support — input field, Q&A bubbles, loading indicators, mic button, and all backgrounds properly themed

Summaries

  • Dedicated Summaries page at /summaries accessible from sidebar ("Summaries" nav item)
  • Three sub-tabs: Room Chat, DM Chat, Whiteboard — each shows persistent summaries filtered by type
  • Summaries stored in MongoDB (Summary model) alongside Cloudflare R2 HTML uploads
  • Summary cards with type badge, title, date, expandable content, and delete button (with ownership check)
  • Generate summary buttons placed contextually throughout the app:
    • Room chat: Summary button in bottom action bar of RoomCallPage (visible only on Chat tab, disabled when no user messages)
    • Whiteboard: Summary button in WhiteboardPage toolbar (between Clear and AI Assist)
    • DM: Summary icon button (FileText) in DmPanel header
  • All use shared generateAndSaveSummary() utility for consistent generate-then-save flow
  • DM Summary — POST /ai/dm-summary fetches up to 100 DMs between user and companion, formats as chat transcript, generates AI summary
  • Save to Cloudflare R2 — uploads summary as styled HTML document to R2 (S3-compatible storage)
    • Generates a presigned download URL valid for 7 days
    • After saving, a VSG Bot message is broadcast to the room chat with the download link
    • Monthly upload quota: configurable via R2_MAX_UPLOADS_PER_MONTH env var (default: 10 per user)
  • RAG-powered Q&A — ask natural language questions across all saved summaries (e.g., "What did we discuss about React?", "What topics this week?")
    • Proper vector embeddings using Gemini text-embedding-004 (768-dim, free tier)
    • Summaries are embedded at save time, stored as vectors in MongoDB
    • At query time: question is embedded, cosine similarity finds the top 5 most relevant summaries, which are passed as context to Gemini 2.5 Flash for answering
    • AI answers cite which summary and date the information comes from
    • Collapsible Q&A panel on the Summaries page with suggestion chips, loading states, and source citation badges
    • Three-layer rate limiting to protect the free tier:
      • Per-user: SUMMARY_QA_MAX_PER_USER (default 10) per SUMMARY_QA_WINDOW_MIN (default 15 min)
      • Global daily: EMBEDDING_DAILY_MAX (default 400) embedding API calls/day across all users, tracked via EmbeddingCounter collection
      • Save-time embeddings naturally limited by existing R2 upload quota
    • Backfill script (backend/src/scripts/backfillEmbeddings.ts) for generating embeddings on existing summaries
    • Full technical documentation with flowcharts — detailed step-by-step breakdown of the entire RAG pipeline, summary generation flows, embedding mechanics, cosine similarity, and rate limiting architecture

Home Page

  • Instagram-style companion presence bar with online/offline indicators
  • "Let's Study Together" gradient CTA card
  • Inshorts-style news feed with mock AI/Tech/Productivity articles
  • Category filter chips (All / AI / Tech / Productivity)
  • Animated card layout with accent color badges

Navigation

  • Collapsible sidebar (not a top bar) with framer-motion spring animation
  • Hamburger toggle at top-left
  • Sidebar items: Home, Chats, Summaries, My Room, Ask AI, Study Radio, Podcasts, Contact us
  • "My Room" menu item dispatches room entry and navigates to call
  • Profile avatar dropdown (shows emoji avatar when set, initials otherwise) with: My Profile, Settings, My Room, Ask AI, Logout
  • Active route highlighting

Rate Limiting & Security

  • All rate limit thresholds configurable via env vars — every value in RATE_LIMIT_CONFIG reads from a corresponding env var with sensible defaults. No code changes needed to tune limits
  • Dual-layer protection on auth endpoints — both per-email and per-IP limits applied simultaneously:
    • Login & Register: AUTH_MAX_PER_EMAIL (default 7) / AUTH_MAX_PER_IP (default 15) per AUTH_WINDOW_MIN (default 15 min)
    • Send OTP: OTP_MAX_PER_EMAIL (default 5) / OTP_MAX_PER_IP (default 7) per OTP_WINDOW_MIN (default 15 min)
  • Per-userId rate limiting on authenticated endpoints:
    • AI Doubt Solver & Session Summary: AI_MAX_PER_USER (default 20) per AI_WINDOW_MIN (default 15 min)
    • Summary Q&A: SUMMARY_QA_MAX_PER_USER (default 10) per SUMMARY_QA_WINDOW_MIN (default 15 min) — tighter limit since each Q&A call makes both an embedding call and a chat completion call
    • User Search: SEARCH_MAX_PER_USER (default 30) per SEARCH_WINDOW_MIN (default 1 min)
  • R2 summary upload quota: R2_MAX_UPLOADS_PER_MONTH (default 10) per user per calendar month, tracked via MongoDB UploadCounter collection
  • Embedding daily cap: EMBEDDING_DAILY_MAX (default 400) embedding API calls per day across all users, tracked via MongoDB EmbeddingCounter collection. Protects Gemini free tier (~1,500 RPD for embeddings)
  • Global safety net: GLOBAL_MAX_PER_IP (default 200) per GLOBAL_WINDOW_MIN (default 15 min) across all routes
  • Socket event throttling (per-user, in-memory, configurable via env):
    • dm:send: SOCKET_DM_INTERVAL_MS (default 200ms)
    • companion:sendRequest: SOCKET_COMPANION_REQ_INTERVAL_MS (default 5000ms)
    • sendInvite: SOCKET_INVITE_INTERVAL_MS (default 3000ms)
  • Throttled socket events emit error responses (dm:error, companion:error, inviteError)
  • HTTP 429 responses include RateLimit-* standard headers
  • trust proxy enabled for correct client IP detection behind reverse proxies

UI/UX

  • Animated components with Framer Motion throughout
  • Responsive design with Tailwind CSS
  • Reusable shadcn/ui components (Button, Card, Input, Label, Select)
  • Overlay notifications for invites and companion requests
  • Toast feedback for invite sent/error states

Coming Soon

  • Screen sharing in video calls
  • Session analytics dashboard
  • Study streak tracking
  • Whiteboard image export to AI (Gemini Vision) for richer spatial analysis

Getting Started

Prerequisites

  • Node.js (v18+)
  • MongoDB instance (local or Atlas)
  • Agora account (for video call App ID)
  • Google Cloud Console OAuth Client ID (for Google sign-in — create one)
  • Google AI Studio API key (for Gemini AI — get one free) or xAI API key (for Grok — get one free)
  • Cloudflare account with R2 bucket + API token (for summary storage — free tier)

Quick Start (Monorepo)

Install all dependencies from the root:

npm run install:all

Start the frontend or backend individually from the root:

npm run dev:frontend   # Starts Vite dev server
npm run dev:backend    # Starts Express + Socket.IO

Backend Setup

cd backend
npm install

Create a .env file in backend/ with:

MONGODB_URI=your_mongodb_connection_string
PORT=7002
JWT_SECRET=your_jwt_secret
AI_PROVIDER=gemini                # "gemini" (default) or "grok"
GEMINI_API_KEY=your_gemini_key    # from https://aistudio.google.com
GEMINI_EMBEDDING_MODEL=text-embedding-004  # Embedding model for RAG Q&A (default: text-embedding-004)
GROK_API_KEY=your_xai_api_key     # from https://console.x.ai (optional, for grok provider)
R2_ACCOUNT_ID=your_cloudflare_id  # Cloudflare account ID (from dashboard URL)
R2_ACCESS_KEY_ID=your_r2_key      # R2 API token access key
R2_SECRET_ACCESS_KEY=your_r2_secret # R2 API token secret key
R2_BUCKET_NAME=study-summaries    # R2 bucket name for saved summaries
R2_MAX_UPLOADS_PER_MONTH=10      # Max summary uploads per user per month (default: 10)
AGORA_APP_ID=your_agora_app_id   # Agora RTC App ID
LISTEN_NOTES_API_KEY=your_key   # Listen Notes API key (300 free calls/month) — https://www.listennotes.com/api/
SMTP_HOST=smtp.gmail.com          # SMTP server for OTP emails
SMTP_PORT=587                     # SMTP port (587 for TLS)
SMTP_USER=your-email@gmail.com    # SMTP sender email
SMTP_PASS=your-app-password       # SMTP password (Gmail: use App Password)
OTP_SECRET=your-random-hex-string # HMAC signing key for stateless OTP
OTP_EXPIRY_MINUTES=5              # OTP validity duration (default: 5 minutes)

# Rate Limiting (all optional — defaults shown, only override what you need)
# AUTH_WINDOW_MIN=15              # Auth window in minutes
# AUTH_MAX_PER_EMAIL=7            # Max login/register per email per window
# AUTH_MAX_PER_IP=15              # Max login/register per IP per window
# OTP_WINDOW_MIN=15              # OTP send window in minutes
# OTP_MAX_PER_EMAIL=5            # Max OTP sends per email per window
# OTP_MAX_PER_IP=7               # Max OTP sends per IP per window
# AI_WINDOW_MIN=15               # AI endpoint window in minutes
# AI_MAX_PER_USER=20             # Max AI requests per user per window
# SEARCH_WINDOW_MIN=1            # User search window in minutes
# SEARCH_MAX_PER_USER=30         # Max searches per user per window
# GLOBAL_WINDOW_MIN=15           # Global safety net window in minutes
# GLOBAL_MAX_PER_IP=200          # Max requests per IP per window (all routes)
# SUMMARY_QA_WINDOW_MIN=15       # Summary Q&A window in minutes
# SUMMARY_QA_MAX_PER_USER=10     # Max Q&A questions per user per window
# EMBEDDING_DAILY_MAX=400        # Max embedding API calls per day (global, all users)
# SOCKET_DM_INTERVAL_MS=200      # Min ms between DM sends per user
# SOCKET_COMPANION_REQ_INTERVAL_MS=5000  # Min ms between companion requests
# SOCKET_INVITE_INTERVAL_MS=3000 # Min ms between room invites

Start the server:

npm start

Frontend Setup

cd frontend
npm install
npm run dev

Create a .env file in frontend/ with:

VITE_API_URL=http://localhost:7002
VITE_AGORA_APP_ID=your_agora_app_id   # Agora RTC App ID
VITE_GOOGLE_CLIENT_ID=your_google_client_id # Google OAuth Client ID (from https://console.cloud.google.com)

Routes

Path Page Auth Required
/ Landing Page No
/login Login No
/register Register No
/home Room Dashboard Yes
/profile User Profile Yes
/settings Settings Yes
/chats Recent Chats Yes
/summaries Summaries Yes
/room/call Video Call Room Yes
/join/:roomId Join Room via Invite Link No (redirects to login)
/whiteboard/:roomId Collaborative Whiteboard Yes
/radio Study Radio Yes
/podcasts Podcasts No
/ask Ask AI (Voice) Yes

API Endpoints

Method Endpoint Description
POST /auth/send-otp Send 6-digit OTP to email (rate limited: per-email + per-IP)
POST /auth/register Verify OTP and register new user (rate limited: per-email + per-IP)
POST /auth/login Verify OTP and login (rate limited: per-email + per-IP)
POST /auth/google Google OAuth sign-in/register (rate limited: per-IP)
POST /companion/request Send companion request
POST /companion/accept Accept companion request
POST /companion/decline Decline companion request
GET /companion/list Get accepted companions
GET /companion/pending Get pending requests
GET /user/profile Get authenticated user's profile (name, email, bio, avatar, education, projects, workExperience, companion count)
PUT /user/profile Update profile (name, bio, avatar, education, projects, workExperience) — re-issues JWT
GET /user/search?q= Search users by name/email (rate limited: per-user)
GET /news Get news feed articles
GET /podcasts/:topic Get podcasts for topic (trending|ai|tech|business|productivity) — JSON file cache, Tue/Sat refresh
POST /ai/ask Ask AI a study question (rate limited: per-user)
POST /ai/summary Generate AI chat session summary (rate limited: per-user)
POST /ai/whiteboard-explain AI analysis of whiteboard drawing (rate limited: per-user)
POST /ai/whiteboard-summary Generate AI whiteboard summary (rate limited: per-user)
POST /ai/save-summary Save summary to Cloudflare R2 + MongoDB, broadcast link to room chat
POST /ai/dm-summary Generate AI summary of DM conversation with a companion
POST /ai/summary-qa RAG Q&A: ask questions across saved summaries (rate limited: per-user + daily embedding cap)
GET /ai/summaries List saved summaries (filter by ?type=room|dm|whiteboard)
DELETE /ai/summaries/:id Delete a saved summary (ownership check)
POST /chat/bulk-save Bulk-save room chat messages to MongoDB (auth required, max 500)
GET /dm/recent Get recent chats (last message per companion, sorted by time)
GET /dm/:companionId Get DM history (includes _id, read state)
GET /dm/unread-counts Get unread message count per companion
PATCH /dm/:companionId/read Mark all messages from companion as read
GET /notifications Get all notifications (last 50, newest first)
PATCH /notifications/:id/read Mark a notification as read
PATCH /notifications/read-all Mark all notifications as read
DELETE /notifications/:id Delete a notification

License

This project is licensed under the GNU General Public License v3.0.

About

Join with all your friends in this awesome virtual group ( Live News feed + Study Companions + Chat Persistence +Video call + Study Music + ASK AI + AI Summaries + Semantic Q/A) 🔥🧯

Topics

redux nodejs redis typescript mongodb reactjs socket-io agora-sdk

Resources

Readme

License

GPL-3.0 license
Activity

Stars

1 star

Watchers

1 watching

Forks

1 fork
Report repository

Releases

No releases published

Packages

 
 
 

Contributors

Languages