🎙️ PulseMeet — Low-Network Resilient Video Meetings & AI Meeting Intelligence

"Where Networks Fail, PulseMeet Lives"

PulseMeet is a production-ready, low-network resilient video conferencing and AI meeting intelligence platform designed, architected, engineered, and deployed by A. Sri Sai Charan. Built with Next.js 16 (App Router), React 19, LiveKit WebRTC, Supabase Auth & Database, and Google Gemini 2.5 Flash AI, PulseMeet is engineered specifically to ensure continuous meeting survival during poor internet conditions — dynamically adapting from video streams to low-bandwidth adaptive video and automatic audio-priority mode, capturing real-time closed captions, buffering network outage gaps, and generating automated post-meeting AI reports (summaries, action items, meeting insights, and PDF export).

🏷️ Badges & Metrics

🌐 Live Demo

🔗 Production Deployment: https://pulsemeet.vercel.app

Log in with Google OAuth to create meeting rooms, join active video calls, view live closed captions, and generate AI meeting reports.

⭐ Highlights

🎥 Real-time WebRTC video conferencing powered by LiveKit
🤖 AI-powered meeting summaries, action items, and insights using Google Gemini
💬 Live subtitles with browser-based speech recognition
🌐 Designed for resilient meeting experiences under unstable network conditions
🔐 Secure Google OAuth authentication with Supabase
📄 Automatic transcript generation and PDF meeting reports
📱 Fully responsive modern UI built with Next.js and Tailwind CSS

🎯 Project Motivation

Most modern video conferencing platforms perform well on fast broadband connections, but quickly degrade or disconnect when internet stability fluctuates. In low-bandwidth or spotty mobile network environments, users frequently experience lost audio/video context, missed commitments, and total call dropouts.

PulseMeet was built to address these core challenges:

Uninterrupted Communication: Prioritizes audio continuity over heavy video streams when bandwidth deteriorates, guaranteeing meetings stay alive on 2G/3G connections.
Context Recovery: Automatically tracks network disconnect intervals and generates an AI Recovery Summary detailing missed discussion points as soon as a participant reconnects.
Automated Post-Meeting Intelligence: Synthesizes raw speech transcripts into executive summaries, owner-assigned action items, and structured insights using Google Gemini AI, backed by zero-downtime local fallback processing engines.

📌 Repository Overview

Dimension	Details
Languages	TypeScript, React, Next.js, SQL
Architecture	Next.js App Router, LiveKit WebRTC, Supabase, Google Gemini AI
Development	Production Ready, Responsive Design, Secure Authentication, AI Meeting Intelligence
Repository	https://github.com/cherry-git999/PulseMeet

📑 Table of Contents

🌐 Live Demo
⭐ Highlights
🎯 Project Motivation
📌 Repository Overview
✨ Core Features
🏗 System Architecture
⚡ Performance & Optimization
🌐 Low-Network Experience & Resilience
🔒 Security & Data Protection
🗄 Database Schema & Entity Relationship
🔌 API Route Reference
🧩 Component Architecture
🛠 Technology Stack
🚀 Installation & Setup
🔑 Environment Variables
📁 Folder Structure
🔄 End-to-End User Flow
🔧 Troubleshooting
🗺 Product Roadmap
🤝 Contributing
📜 License
🙏 Acknowledgements
👨‍💻 Author

✨ Core Features

PulseMeet combines real-time WebRTC communications with client-side speech recognition and server-side generative AI to deliver a complete meeting lifecycle management solution. Designed, built, and tested as a portfolio-grade full-stack engineering project by A. Sri Sai Charan, PulseMeet demonstrates modern WebRTC architecture and AI integration.

📶 Low-Network First Engineering Objective

Unlike conventional conferencing apps that freeze or drop calls during network instability, PulseMeet's primary engineering objective is low-network survival:

Adaptive Quality Control: Dynamically scales down WebRTC bitrates based on connection telemetry.
Audio Priority Fallback: When connection reaches critical thresholds (2G / slow 3G), video publication is automatically paused to guarantee uninterrupted audio continuity.
AI Outage Recovery: Outages during meetings are automatically tracked and summarized upon reconnection via the RecoverySummaryModal so participants never lose context.

🔐 Authentication & Profile Management

Google OAuth 2.0 via Supabase Auth: Secure single-click authentication using standard OAuth redirection flows.
Automatic Profile Provisioning: Extracts Google user metadata (full_name, email, avatar_url) and stores it in a dedicated profiles table.
Profile Completion & Editing: Page (/complete-profile) enabling users to customize display names and avatar fallbacks (via ui-avatars.com).

📹 Real-Time WebRTC Video Meetings

LiveKit WebRTC Infrastructure: Ultra-low-latency video/audio streaming utilizing LiveKit server tokens and SDK client connection management.
Dynamic Token Server: Secure backend endpoint (/api/livekit-token) issuing short-lived JWTs scoped to specific room channels.
Dynamic Video Grid Layout: Responsive grid renderer adapting seamlessly to 1 to multi-participant streams with speaker highlighting.
Media Controls: Independent microphone toggle, camera toggle, screen sharing, raise hand status indicator, and live closed-caption overlay.
Local Media Preview: Instant pre-join camera and mic hardware verification.
Adaptive WebRTC Streams: Dynamic stream adaptation (dynacast, adaptiveStream) balancing quality based on subscriber viewport.

🗣 Real-Time Speech Recognition & Subtitles

Browser-Native Web Speech API: Continuous zero-latency speech recognition (webkitSpeechRecognition) integrated directly into the meeting room loop.
Live Captions Banner: Subtitle overlay bar displaying real-time transcribed dialogue on screen during active meetings.
Transcript Buffer & Deduplication: Real-time entry management (TranscriptManager) tracking timestamps, epoch times, speaker labels, speech fragments, and system state changes.

🧠 Dual-Engine AI Intelligence (Cloud AI + Zero-Downtime Local Fallback)

PulseMeet features a dual-engine AI pipeline. Primary summarization uses Google Gemini 2.5 Flash via a secure Next.js API route. If Gemini quota limits or network outages occur, PulseMeet transparently shifts to local heuristic processing engines.

AI Output	Primary Engine (Google Gemini AI)	Local Fallback Engine (Zero-Downtime)
Executive Summary	`generateSummary()` via Gemini 2.5 Flash API	`SummaryService`: Speaker-line extraction & dialogue filtering
Action Items	`generateActionItems()` structured prompt extraction	`ActionItemsService`: Regex sentence splitting & keyword analysis (`assigned to`, `will`, `needs to`, `should`)
Meeting Insights	`generateInsights()` structured metrics prompt	`InsightsService`: Word-frequency noun parsing, timestamp duration calculation, decision phrase tracking
Missed Meeting Recovery	`RecoveryService` bullet point generation	Local phrase summarizer prefixing missed content with `"You missed:"`

📊 Meeting Reports & Analytics Dashboard

Comprehensive Reports Hub (/reports): Filter, search, sort, and review past AI meeting reports.
Multi-Format Data Exporting:
- 📄 Plain Text Transcript (.txt): Full timestamped dialogue log.
- 📝 Summary & Action Items (.txt): Executive summary and formatted bulleted action items.
- 📕 Formatted PDF Report (.pdf): Styled multi-page PDF document generated client-side via jsPDF, containing Executive Summary, Action Items, Topics, Key Themes, and Full Transcript.
Batch Report Deletion: Multi-select report deletion with real-time UI synchronization and Supabase RLS validation.
Interactive Dashboard (/dashboard): Visual metric cards displaying total rooms, total reports, action item counts, monthly meeting minutes, extracted topics, AI summaries, and PDF downloads.

🏗 System Architecture

PulseMeet separates real-time media streams, authentication, speech processing, and AI generation into decoupled layers.

1. High-Level Application Architecture

flowchart TD
    subgraph Client ["Client Browser (Next.js 16 Client Bundle)"]
        UI["React 19 UI Components"]
        Speech["Web Speech API (SpeechRecognition)"]
        TM["TranscriptManager (In-Memory Buffer)"]
        LKClient["LiveKit WebRTC Client"]
    end

    subgraph Auth ["Authentication & Storage (Supabase)"]
        SupaAuth["Supabase Auth (Google OAuth)"]
        SupaDB[("Supabase PostgreSQL DB")]
    end

    subgraph RTC ["Real-Time Communication"]
        LKServer["LiveKit Cloud / WebRTC Server"]
    end

    subgraph AI ["AI Intelligence Services"]
        APIRoute["Next.js Server API (/api/meeting/ai)"]
        Gemini["Google Gemini 2.5 Flash AI"]
        LocalAI["Local Heuristic AI Engines (Fallback)"]
    end

    UI <-->|OAuth Flow| SupaAuth
    UI <-->|CRUD Rooms / Reports| SupaDB
    LKClient <-->|WebRTC Media Tracks| LKServer
    Speech -->|Speech Events| TM
    TM -->|Meeting End Transcript| APIRoute
    APIRoute -->|Primary Prompt| Gemini
    APIRoute -.->|Fallback Execution| LocalAI
    APIRoute -->|Save Results| SupaDB

2. Authentication & Profile Callback Flow

sequenceDiagram
    autonumber
    actor User
    participant Login as Login Page
    participant Supabase as Supabase Auth
    participant Google as Google OAuth Provider
    participant Callback as Auth Callback
    participant Profile as Complete Profile
    participant Dashboard as Dashboard

    User->>Login: Click Continue with Google
    Login->>Supabase: signInWithOAuth
    Supabase->>Google: Redirect to Google Login
    Google-->>User: Authenticate & Authorize
    Google->>Callback: Redirect with Access Tokens
    Callback->>Supabase: getSession
    Callback->>Dashboard: Redirect to Dashboard
    Dashboard->>Supabase: Query profiles table
    alt Profile missing or incomplete
        Dashboard->>Profile: Redirect to complete profile
        User->>Profile: Enter display name and avatar
        Profile->>Supabase: Upsert profiles record
        Profile->>Dashboard: Redirect back to Dashboard
    end

3. LiveKit WebRTC Session Lifecycle & Token Flow

sequenceDiagram
    autonumber
    actor Attendee
    participant RoomPage as Meeting Room
    participant TokenAPI as API Route
    participant LKSDK as LiveKit Server SDK
    participant LKServer as LiveKit WebRTC Server

    Attendee->>RoomPage: Navigate to Room
    RoomPage->>TokenAPI: GET livekit-token endpoint
    TokenAPI->>LKSDK: createToken roomName and username
    LKSDK-->>TokenAPI: Signed AccessToken JWT
    TokenAPI-->>RoomPage: Return token object
    RoomPage->>LKServer: room.connect url and token
    LKServer-->>RoomPage: Connected Event
    RoomPage->>LKServer: Enable Mic & Camera Tracks
    LKServer-->>Attendee: Media Stream Established

4. Speech Recognition & AI Intelligence Pipeline

flowchart LR
    subgraph Capture ["Audio Capture"]
        Mic["Microphone Input"] --> SpeechRec["Web Speech API"]
    end

    subgraph Buffer ["Local Memory"]
        SpeechRec -->|onresult| TM["TranscriptManager"]
        TM -->|addSpeech| Array["Transcript Entries Array"]
    end

    subgraph PostMeeting ["Meeting Termination Cleanup"]
        Array -->|generateTranscript| FullText["Raw Transcript Text"]
        FullText --> PostAPI["POST /api/meeting/ai"]
    end

    subgraph Generation ["Dual-Engine Processing"]
        PostAPI -->|Primary| Gemini["Google Gemini 2.5 Flash"]
        PostAPI -.->|API Failure / Catch| LocalFallback["Local Service Engines"]
        
        LocalFallback --> SummarySvc["SummaryService"]
        LocalFallback --> ActionSvc["ActionItemsService"]
        LocalFallback --> InsightsSvc["InsightsService"]
    end

    subgraph Storage ["Database Persistence"]
        Gemini --> Save[("Supabase: meeting_ai_results")]
        SummarySvc --> Save
        ActionSvc --> Save
        InsightsSvc --> Save
    end

5. Network Quality Monitoring & AI Recovery Flow

flowchart TD
    Start["Active Meeting Session"] --> Monitor["NetworkStats Monitor"]
    Monitor --> Check{"Connection Effective Type?"}

    Check -->|4G / High Speed| Excellent["Status: Excellent (720p Video)"]
    Check -->|4G| Good["Status: Good (480p Adaptive)"]
    Check -->|3G| Poor["Status: Poor (360p Low Data)"]
    Check -->|2G / Slow-2G| Critical["Status: Critical (Audio Priority)"]

    Critical --> AutoAudio["Auto-Disable Local Camera"]
    Critical --> PauseRec["Pause AI Recording Signal"]
    Critical --> RecordStart["Record Network Drop Start Timestamp"]

    Excellent --> ReconnectCheck{"Was in Critical State?"}
    Good --> ReconnectCheck

    ReconnectCheck -->|Yes| RecoveryTrigger["Trigger Network Recovery Process"]
    ReconnectCheck -->|No| NormalRun["Continue Normal Streams"]

    RecoveryTrigger --> EnableCam["Re-enable Camera Track"]
    RecoveryTrigger --> ResumeRec["Resume AI Recording Signal"]
    RecoveryTrigger --> ExtractMissed["Extract Missed Entries"]
    ExtractMissed --> RecoverySvc["RecoveryService.generate"]
    RecoverySvc --> Modal["Display RecoverySummaryModal"]

⚡ Performance & Optimization

PulseMeet is optimized for modern web performance standards:

Next.js 16 App Router: Leverages React Server Components for static page shell delivery and fast first contentful paint (FCP).

Client Component Code-Splitting: Heavier client-side modules (WebRTC UI, PDF generators, filter modals) are dynamically imported using next/dynamic with ssr: false:

const MeetingRoom = dynamic(() => import("@/components/meeting/MeetingRoom"), { ssr: false });
const RecoverySummaryModal = dynamic(() => import("@/components/meeting/RecoverySummaryModal"), { ssr: false });
const SummaryViewer = dynamic(() => import("@/components/reports/SummaryViewer"), { ssr: false });

WebRTC Adaptive Bitrate (dynacast & adaptiveStream): Automatically scales down video track resolutions for hidden or background grid tiles to conserve bandwidth and CPU usage.
Memoized State Computations: Critical components wrap array filtering, statistics calculations, and participant mapping in useMemo and useCallback hooks to eliminate unnecessary re-renders.
Asynchronous Audio Track Detachment: LiveKit audio elements are managed with DOM element cleanup routines upon track unsubscribing to prevent memory leaks.

🌐 Low-Network Experience & Resilience

PulseMeet provides robust network degradation support for users on unstable internet connections:

Active Connection Telemetry (NetworkStats): Reads navigator.connection properties (effectiveType, downlink, rtt) to monitor real-time network conditions.
Automatic Audio Priority: When network quality degrades to Critical (2g or slow-2g), PulseMeet automatically disables camera publication, prioritizing clear audio continuity.
AI Transcript Outage Buffering: Connection drops trigger transcriptManager.pauseRecording(). When reconnected, transcriptManager.resumeRecording() logs the outage interval.
Automated Missed Content Summarization: PulseMeet calculates the exact time interval of the disconnect, queries getEntriesBetween(startMs, endMs), parses speaker statements with RecoveryService, and prompts the user with an AI Recovery Summary Modal (RecoverySummaryModal) detailing missed discussion points.

🔒 Security & Data Protection

Security controls protect user privacy and system access:

Supabase Authentication: Passwords and credentials are managed directly by Supabase Auth with Google OAuth PKCE exchange flow.
Row Level Security (RLS): PostgreSQL tables enforce database-level access policies ensuring users can only read, edit, or delete rooms and reports created by their own authenticated user ID.
Short-Lived LiveKit JWT Tokens: WebRTC access tokens are signed on the server (livekit-server-sdk) with strictly scoped room permissions (roomJoin, canPublish, canSubscribe) and never expose the master secret.
Encapsulated Gemini API Credentials: All generative AI requests execute server-side within the /api/meeting/ai route. The GEMINI_API_KEY is never exposed in client bundle code.
XSS & Injection Defense: Dynamic URLs use site URL helpers (lib/config/site.ts) to validate redirect domains and protect against open redirect vulnerabilities.

🗄 Database Schema & Entity Relationship

PulseMeet uses a relational PostgreSQL database schema managed via Supabase.

PostgreSQL Tables

1. `profiles`

Stores user profile information synced from Supabase Auth.

Column	Type	Constraints	Description
`id`	`uuid`	Primary Key, `references auth.users(id)`	User's unique Auth ID
`email`	`text`	`NOT NULL`	User email address
`full_name`	`text`	`NULLABLE`	Display name
`avatar_url`	`text`	`NULLABLE`	Profile picture URL
`updated_at`	`timestamptz`	`DEFAULT now()`	Last update timestamp

2. `rooms`

Stores meeting room configurations and AI toggle preferences.

Column	Type	Constraints	Description
`id`	`uuid`	Primary Key, `DEFAULT gen_random_uuid()`	Unique room UUID
`name`	`text`	`NOT NULL`	Room title/name
`created_by`	`uuid`	`NOT NULL`, `references profiles(id)`	Room creator ID
`screen_share_enabled`	`boolean`	`DEFAULT true`	Screen share feature toggle
`ai_transcript`	`boolean`	`DEFAULT false`	AI speech transcription toggle
`ai_summary`	`boolean`	`DEFAULT false`	AI summary generation toggle
`ai_action_items`	`boolean`	`DEFAULT false`	AI action item extraction toggle
`ai_recovery_mode`	`boolean`	`DEFAULT false`	AI network recovery feature toggle
`created_at`	`timestamptz`	`DEFAULT now()`	Room creation timestamp

3. `meeting_ai_results`

Stores post-meeting AI reports, raw transcripts, summaries, and metrics.

Column	Type	Constraints	Description
`id`	`uuid`	Primary Key, `DEFAULT gen_random_uuid()`	Report unique ID
`room_id`	`uuid`	`NOT NULL`, `references rooms(id) ON DELETE CASCADE`	Associated meeting room
`transcript`	`text`	`NULLABLE`	Full timestamped transcript log
`summary`	`text`	`NULLABLE`	Generated executive summary
`action_items`	`text`	`NULLABLE`	Extracted action items string
`insights`	`text`	`NULLABLE`	Duration, topics, decisions & themes
`created_at`	`timestamptz`	`DEFAULT now()`	Report generation timestamp

Database Entity Relationship Diagram (ERD)

erDiagram
    profiles ||--o{ rooms : "creates"
    rooms ||--o{ meeting_ai_results : "generates"

    profiles {
        uuid id PK
        text email
        text full_name
        text avatar_url
        timestamptz updated_at
    }

    rooms {
        uuid id PK
        text name
        uuid created_by FK
        boolean screen_share_enabled
        boolean ai_transcript
        boolean ai_summary
        boolean ai_action_items
        boolean ai_recovery_mode
        timestamptz created_at
    }

    meeting_ai_results {
        uuid id PK
        uuid room_id FK
        text transcript
        text summary
        text action_items
        text insights
        timestamptz created_at
    }

🔌 API Route Reference

PulseMeet includes server-side Next.js route handlers:

1. `GET /api/livekit-token`

Generates a LiveKit JWT token for joining a WebRTC meeting room.

Query Parameters:
- room (string, optional): Room UUID or name. Default: "default-room"
- username (string, optional): Display name for the participant. Default: "guest"

Success Response (200 OK):

{
  "token": "eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9..."
}

2. `POST /api/meeting/ai`

Generates meeting summary, action items, and insights via Google Gemini 2.5 Flash AI.

Request Body:

{
  "transcript": "[10:00 AM] Alex: We need to launch the app by Friday.\n[10:01 AM] Sarah: I will write the documentation.",
  "summaryEnabled": true,
  "actionItemsEnabled": true,
  "insightsEnabled": true
}

Success Response (200 OK):

{
  "summary": "Executive summary of the discussion...",
  "actionItems": "- Sarah: Write the documentation by Friday",
  "insights": "Duration: 2 min\nTopics Discussed:\n- App Launch\nDecisions Count: 1\nAction Items Count: 1\nKey Themes:\n- Collaborative discussion"
}

Error Response (400 Bad Request):
```
{
  "error": "Transcript is required"
}
```

3. `GET /api/test-gemini`

Diagnostics route for testing Google Gemini API key configuration and connectivity.

Success Response (200 OK):

{
  "status": "success",
  "response": "Gemini Connection Successful"
}

🧩 Component Architecture

PulseMeet follows a modular component structure:

Key Pages (`app/`)

app/layout.tsx: Root HTML layout importing Google Geist fonts and styling frameworks.
app/page.tsx: Entry point redirecting users to /login.
app/login/page.tsx: Google OAuth login page with action cards.
app/auth/callback/page.tsx: OAuth redirection callback route processing Supabase auth tokens.
app/complete-profile/page.tsx: User profile completion page.
app/dashboard/page.tsx: User dashboard displaying quick links, metrics, and recent reports.
app/create-room/page.tsx: Interface to construct meeting rooms and feature flags.
app/join-meeting/page.tsx: Direct entry form accepting room IDs or meeting URLs.
app/room/[roomId]/page.tsx: Dynamically rendered meeting room wrapper (ssr: false).
app/rooms/page.tsx: Grid view of all created rooms with copy link and deletion features.
app/rooms/[roomId]/edit/page.tsx: Room configuration editor.
app/reports/page.tsx: Reports hub with search, filters, batch operations, detail views, and PDF export.

Core Components (`components/`)

components/meeting/MeetingRoom.tsx: Core WebRTC meeting orchestrator managing LiveKit rooms, track publishing, network monitoring, and transcript saving.
components/meeting/MeetingControls.tsx: Meeting footer containing hardware toggles, screen share button, captions, and raise hand.
components/meeting/MeetingVideoArea.tsx: Grid container rendering local preview and remote participant video tiles.
components/meeting/VideoTile.tsx: Individual participant tile managing media elements and active speaker highlights.
components/meeting/NetworkStats.tsx: Network quality monitor displaying connection effective type, downlink, and RTT.
components/meeting/LiveCaptions.tsx: Floating closed-captions banner.
components/meeting/RecoverySummaryModal.tsx: Outage recovery summary popup.
components/reports/InsightsViewer.tsx: Rendered metrics cards displaying duration, topics, decisions, and themes.
components/reports/ReportCard.tsx: Individual report card with search highlight and multi-selection checkbox.

🛠 Technology Stack

Core Framework & Runtime

Technology	Version	Purpose
Next.js	`16.2.6`	App Router, Server Components, API Routes
React	`19.2.4`	UI Library & Hooks Engine
TypeScript	`^5.0.0`	Static Type Safety

WebRTC & Real-Time Media

Technology	Version	Purpose
LiveKit Client	`^2.19.1`	WebRTC audio/video client connection handling
LiveKit Components React	`^2.9.21`	React components for LiveKit streams
LiveKit Server SDK	`^2.15.4`	Server-side JWT token generation (`AccessToken`)

Database & Authentication

Technology	Version	Purpose
Supabase Client	`^2.106.2`	Auth, PostgreSQL client queries & RLS enforcement

AI & Speech Recognition

Technology	Version	Purpose
@google/genai	`^2.8.0`	Official SDK for Google Gemini 2.5 Flash AI
Web Speech API	Native Browser	Client-side continuous speech recognition

Styling & Utility Libraries

Technology	Version	Purpose
Tailwind CSS	`^4.0.0`	Utility-first CSS framework
Lucide React	`^1.17.0`	Icon set
jsPDF	`^4.2.1`	Client-side PDF report rendering
clsx & tailwind-merge	`^2.1.1`	Class name manipulation

🚀 Installation & Setup

Follow these instructions to run PulseMeet locally.

1. Prerequisites

Node.js: v18.17.0 or higher
npm or pnpm or yarn
Supabase Account: A running Supabase project with Google Auth enabled.
LiveKit Server Instance: LiveKit Cloud account or self-hosted LiveKit instance.
Google Gemini API Key: API key from Google AI Studio.

2. Clone Repository & Install Dependencies

git clone https://github.com/cherry-git999/PulseMeet.git
cd PulseMeet
npm install

3. Setup Environment Variables

Create a .env.local file in the project root by copying .env.example:

cp .env.example .env.local

Fill in your environment credentials in .env.local:

# Supabase Configuration
NEXT_PUBLIC_SUPABASE_URL=https://your-supabase-project.supabase.co
NEXT_PUBLIC_SUPABASE_ANON_KEY=your-supabase-anon-key

# LiveKit Configuration
NEXT_PUBLIC_LIVEKIT_URL=wss://your-livekit-project.livekit.cloud
LIVEKIT_API_KEY=your-livekit-api-key
LIVEKIT_API_SECRET=your-livekit-api-secret

# Gemini AI API Configuration
GEMINI_API_KEY=your-gemini-api-key

# Centralized Site URL for redirects
NEXT_PUBLIC_SITE_URL=http://localhost:3000

4. Database Setup (Supabase SQL Script)

Run the following SQL migration script in your Supabase SQL Editor to construct the required tables and security policies:

-- 1. Create Profiles Table
CREATE TABLE IF NOT EXISTS public.profiles (
    id UUID PRIMARY KEY REFERENCES auth.users(id) ON DELETE CASCADE,
    email TEXT NOT NULL,
    full_name TEXT,
    avatar_url TEXT,
    updated_at TIMESTAMPTZ DEFAULT now()
);

-- 2. Create Rooms Table
CREATE TABLE IF NOT EXISTS public.rooms (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    name TEXT NOT NULL,
    created_by UUID NOT NULL REFERENCES public.profiles(id) ON DELETE CASCADE,
    screen_share_enabled BOOLEAN DEFAULT true,
    ai_transcript BOOLEAN DEFAULT false,
    ai_summary BOOLEAN DEFAULT false,
    ai_action_items BOOLEAN DEFAULT false,
    ai_recovery_mode BOOLEAN DEFAULT false,
    created_at TIMESTAMPTZ DEFAULT now()
);

-- 3. Create Meeting AI Results Table
CREATE TABLE IF NOT EXISTS public.meeting_ai_results (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    room_id UUID NOT NULL REFERENCES public.rooms(id) ON DELETE CASCADE,
    transcript TEXT,
    summary TEXT,
    action_items TEXT,
    insights TEXT,
    created_at TIMESTAMPTZ DEFAULT now()
);

-- Enable Row Level Security (RLS)
ALTER TABLE public.profiles ENABLE ROW LEVEL SECURITY;
ALTER TABLE public.rooms ENABLE ROW LEVEL SECURITY;
ALTER TABLE public.meeting_ai_results ENABLE ROW LEVEL SECURITY;

-- RLS Policies for profiles
CREATE POLICY "Users can view all profiles" ON public.profiles FOR SELECT USING (true);
CREATE POLICY "Users can update own profile" ON public.profiles FOR UPDATE USING (auth.uid() = id);
CREATE POLICY "Users can insert own profile" ON public.profiles FOR INSERT WITH CHECK (auth.uid() = id);

-- RLS Policies for rooms
CREATE POLICY "Users can view own rooms" ON public.rooms FOR SELECT USING (auth.uid() = created_by);
CREATE POLICY "Users can insert own rooms" ON public.rooms FOR INSERT WITH CHECK (auth.uid() = created_by);
CREATE POLICY "Users can update own rooms" ON public.rooms FOR UPDATE USING (auth.uid() = created_by);
CREATE POLICY "Users can delete own rooms" ON public.rooms FOR DELETE USING (auth.uid() = created_by);

-- RLS Policies for meeting_ai_results
CREATE POLICY "Users can view reports for their rooms" ON public.meeting_ai_results
    FOR SELECT USING (
        EXISTS (SELECT 1 FROM public.rooms WHERE rooms.id = meeting_ai_results.room_id AND rooms.created_by = auth.uid())
    );

CREATE POLICY "Users can insert reports for their rooms" ON public.meeting_ai_results
    FOR INSERT WITH CHECK (
        EXISTS (SELECT 1 FROM public.rooms WHERE rooms.id = meeting_ai_results.room_id AND rooms.created_by = auth.uid())
    );

CREATE POLICY "Users can delete reports for their rooms" ON public.meeting_ai_results
    FOR DELETE USING (
        EXISTS (SELECT 1 FROM public.rooms WHERE rooms.id = meeting_ai_results.room_id AND rooms.created_by = auth.uid())
    );

5. Run Local Development Server

Start the Next.js development server:

npm run dev

Open http://localhost:3000 in your browser.

6. Production Build & Deployment

To generate a production build:

npm run build
npm run start

Deploy seamlessly on Vercel by linking your repository and populating all environment variables in the Vercel dashboard.

🔑 Environment Variables

Variable Name	Required?	Example Value	Description & Security Notes
`NEXT_PUBLIC_SUPABASE_URL`	Yes	`https://xyz.supabase.co`	Public Supabase API project endpoint
`NEXT_PUBLIC_SUPABASE_ANON_KEY`	Yes	`eyJhbGciOi...`	Public client API key for Supabase Auth and database operations
`NEXT_PUBLIC_LIVEKIT_URL`	Yes	`wss://app.livekit.cloud`	Public WebSocket endpoint for LiveKit WebRTC server
`LIVEKIT_API_KEY`	Yes	`APIxxxxxxxxx`	Private LiveKit API Key (server-side token generation)
`LIVEKIT_API_SECRET`	Yes	`secretxxxxxxx`	Private LiveKit API Secret (never expose to client)
`GEMINI_API_KEY`	Yes	`AIzaSyXXXXXX`	Private Google Gemini API Key used in `/api/meeting/ai`
`NEXT_PUBLIC_SITE_URL`	Recommended	`https://pulsemeet.vercel.app`	Base URL used for invitation links and OAuth redirects

📁 Folder Structure

PulseMeet/
├── app/                        # Next.js 16 App Router Pages & API Routes
│   ├── api/                    # Server-side API endpoints
│   │   ├── livekit-token/      # Token generator for WebRTC rooms (route.ts)
│   │   ├── meeting/            # AI generation handler (meeting/ai/route.ts)
│   │   └── test-gemini/        # Gemini diagnostic endpoint (route.ts)
│   ├── auth/                   # Authentication handling
│   │   └── callback/           # Google OAuth callback handler (page.tsx)
│   ├── complete-profile/       # Profile setup and update (page.tsx)
│   ├── create-room/            # New meeting room form (page.tsx)
│   ├── dashboard/              # User dashboard & statistics (page.tsx)
│   ├── join-meeting/           # Join room by ID or URL (page.tsx)
│   ├── login/                  # OAuth login page (page.tsx)
│   ├── reports/                # Meeting reports hub & PDF exporter (page.tsx)
│   ├── room/                   # WebRTC meeting room page ([roomId]/page.tsx)
│   ├── rooms/                  # Room management list & editor ([roomId]/edit)
│   ├── error.tsx               # Global error boundary component
│   ├── globals.css             # Tailwind v4 styles & theme utilities
│   ├── layout.tsx              # Root HTML layout with Geist typography
│   └── page.tsx                # Base route (redirects to /login)
├── components/                 # Modular React Components
│   ├── dashboard/              # Dashboard metrics & statistic cards
│   │   └── DashboardStats.tsx
│   ├── meeting/                # Meeting room WebRTC UI components
│   │   ├── AIStatusChip.tsx         # Active AI feature status chip & dropdown
│   │   ├── InviteModal.tsx          # Copy link & share modal
│   │   ├── LiveCaptions.tsx         # Real-time subtitle overlay
│   │   ├── LocalVideoPreview.tsx    # Camera/mic preview widget
│   │   ├── MeetingHeader.tsx        # Top status bar & network indicator
│   │   ├── MeetingRoom.tsx          # Main WebRTC orchestrator
│   │   ├── MeetingVideoArea.tsx     # Dynamic grid container
│   │   ├── NetworkIndicator.tsx     # Compact quality badge
│   │   ├── NetworkStats.tsx         # Connection metrics inspector
│   │   ├── ParticipantSidebar.tsx   # Participant list & hand raise list
│   │   ├── RecoverySummaryModal.tsx # Missed content recovery modal
│   │   ├── VideoGrid.tsx            # Flex/Grid layout manager
│   │   └── VideoTile.tsx            # Individual participant media tile
│   ├── reports/                # Report viewing & export components
│   │   ├── ActionItemsViewer.tsx    # Action items checklist view
│   │   ├── DeleteReportModal.tsx    # Deletion confirmation modal
│   │   ├── InsightsViewer.tsx       # Metrics summary cards
│   │   ├── ReportCard.tsx           # Report grid item
│   │   ├── ReportFilters.tsx        # Filter controls
│   │   ├── ReportSearch.tsx         # Search bar
│   │   └── SummaryViewer.tsx        # Executive summary viewer
│   ├── reports/                # Full dialogue transcript viewer
│   │   └── TranscriptViewer.tsx
│   ├── rooms/                  # Room CRUD components
│   │   ├── DeleteRoomModal.tsx      # Room deletion confirmation modal
│   │   ├── RoomCard.tsx             # Room grid item
│   │   └── RoomForm.tsx             # Shared room creation & edit form
│   └── ui/                     # Reusable UI primitive components
│       ├── AppCard.tsx
│       ├── CopyToast.tsx
│       ├── MeetingControls.tsx
│       ├── ParticipantCard.tsx
│       ├── PrimaryButton.tsx
│       └── StatusBadge.tsx
├── lib/                        # Core Utilities & AI Logic Services
│   ├── ai/                     # Dual-Engine AI services
│   │   ├── prompts/             # Gemini prompt templates (summary, action items, insights)
│   │   ├── action-items-service.ts  # Fallback action items regex parser
│   │   ├── gemini.ts            # GoogleGenAI SDK client initialization
│   │   ├── insights-service.ts  # Fallback insights metrics generator
│   │   ├── networkRecovery.ts   # Network outage interval tracker
│   │   ├── provider.ts          # Server-side Gemini execution methods
│   │   ├── recovery-service.ts  # Outage recovery summary generator
│   │   ├── summary-service.ts   # Fallback executive summary generator
│   │   └── transcription-service.ts # Client Web Speech API wrapper
│   ├── config/                 # Centralized site configurations (site.ts)
│   ├── reports/                # Exporters & parsers
│   │   ├── duration-parser.ts   # Duration string to minute converter
│   │   ├── export-utils.ts      # PDF (jsPDF) & text export helper functions
│   │   └── insights-parser.ts   # Markdown insights text parser
│   ├── livekit.ts              # Server-side LiveKit access token generator
│   ├── supabase.ts             # Supabase client instantiation
│   ├── transcript-manager.ts   # Real-time transcript state buffer
│   └── utils.ts                # Tailwind class merger
├── types/                      # TypeScript Global Declarations
│   └── speech-recognition.d.ts # SpeechRecognition browser type definitions
├── .env.example                # Template for environment variables
├── next.config.ts              # Next.js configuration
├── package.json                # Dependencies & script definitions
└── tsconfig.json               # TypeScript compiler config

🔄 End-to-End User Flow

┌─────────────────┐       ┌─────────────────┐       ┌─────────────────┐
│ 1. Google OAuth │ ────> │ 2. Profile Sync │ ────> │  3. Dashboard   │
│     Sign In     │       │ & Metadata Check│       │ Overview & Stats│
└─────────────────┘       └─────────────────┘       └────────┬────────┘
                                                             │
                                                             ▼
┌─────────────────┐       ┌─────────────────┐       ┌─────────────────┐
│ 6. Meeting End  │ <──── │ 5. Speech Rec   │ <──── │ 4. Join WebRTC  │
│  & Dual-AI Run  │       │ & Subtitles Run │       │   Meeting Room  │
└────────┬────────┘       └─────────────────┘       └─────────────────┘
         │
         ▼
┌─────────────────┐       ┌─────────────────┐
│ 7. Save Report  │ ────> │ 8. Browse, Filter│
│   to Supabase   │       │  & Download PDF │
└─────────────────┘       └─────────────────┘

Authentication: User signs in with Google OAuth on /login. Token exchange completes via /auth/callback.
Dashboard: User lands on /dashboard, views meeting statistics, recent reports, and quick actions.
Room Creation: User navigates to /create-room, specifies room name and toggles AI features (Transcript, Summary, Action Items, Recovery Mode).
Meeting Session: User joins /room/[roomId]. The app fetches a LiveKit token from /api/livekit-token and initializes media tracks.
Real-time AI Features:
- Web Speech API listens and displays live captions.
- TranscriptManager buffers dialogue entries with timestamps.
- NetworkStats tracks connection telemetry. If quality drops, audio priority triggers and an outage interval is recorded.
Meeting Termination: Leaving the room triggers meeting completion logic:
- Final system message "Meeting ended" is appended.
- The transcript is sent to /api/meeting/ai.
- Primary AI (Gemini) or local fallback engines process the transcript into Summary, Action Items, and Insights.
- Results are saved to meeting_ai_results in Supabase.
Report Management & PDF Export: User opens /reports, searches or filters reports, views breakdown tabs, and downloads .txt or styled .pdf meeting reports.

🔧 Troubleshooting

Symptom / Issue	Possible Cause	Solution
LiveKit Connection Fails	Missing or incorrect `NEXT_PUBLIC_LIVEKIT_URL`, `LIVEKIT_API_KEY`, or `LIVEKIT_API_SECRET`.	Verify LiveKit environment variables in `.env.local` and ensure WebSocket URL begins with `wss://`.
Microphone / Camera Permission Error	Browser permissions blocked or hardware in use by another app.	Grant media permissions in browser site settings and close conflicting applications using camera/mic.
Speech Recognition Not Working	Browser does not support Web Speech API (`webkitSpeechRecognition`).	Use Google Chrome, Microsoft Edge, or a WebKit-compatible browser. Note that Firefox lacks native `SpeechRecognition` support.
Gemini AI API Error (500 or Empty Response)	Missing `GEMINI_API_KEY` or quota exceeded on Google AI Studio.	Check `/api/test-gemini` endpoint. If Gemini fails, PulseMeet automatically falls back to local processing engines without breaking meeting execution.
Supabase RLS Error when Deleting Reports	Missing `DELETE` policy on `meeting_ai_results` table in Supabase.	Re-run the SQL migration script provided in Step 4 of setup to grant correct RLS DELETE permissions for room creators.
PDF Download Failures	Popup blocker or canvas rendering error in jsPDF.	Ensure browser allows popups/downloads for the site domain.

🗺 Product Roadmap

🤝 Contributing

Contributions are welcome! If you would like to contribute improvements, feel free to fork the repository and submit a pull request.

Fork the Repository
Create your Feature Branch:
```
git checkout -b feature/AmazingFeature
```
Commit your Changes:
```
git commit -m 'Add some AmazingFeature'
```
Push to the Branch:
```
git push origin feature/AmazingFeature
```
Open a Pull Request

📜 License

Distributed under the MIT License. See LICENSE for details.

MIT License

Copyright (c) 2026 A. Sri Sai Charan

Permission is hereby granted, free of charge, to any person obtaining a copy
of this software and associated documentation files (the "Software"), to deal
in the Software without restriction, including without limitation the rights
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
copies of the Software, and to permit persons to whom the Software is
furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all
copies or substantial portions of the Software.

🙏 Acknowledgements

Next.js — React Framework for the Web
React — Library for web and native user interfaces
LiveKit — Open-source WebRTC Infrastructure
Supabase — Open-source Firebase Alternative
Google Gemini AI — Next-generation Multimodal AI
Tailwind CSS — Utility-first CSS framework
TypeScript — Strongly typed programming language
Vercel — Cloud platform for static sites and Serverless Functions
Lucide Icons — Beautiful & consistent icons
jsPDF — Client-side HTML5 PDF solution

👨‍💻 Author

A. Sri Sai Charan
Computer Science Engineer | Full Stack Developer | AI Enthusiast

GitHub: https://github.com/cherry-git999
LinkedIn: https://www.linkedin.com/in/sri-sai-charan-arasada-1a1b12320/

PulseMeet — Designed, engineered, and built with ❤️ by A. Sri Sai Charan.

Name		Name	Last commit message	Last commit date
Latest commit History 49 Commits
app		app
components		components
lib		lib
public		public
scratch		scratch
scripts		scripts
types		types
.env.example		.env.example
.gitignore		.gitignore
LICENSE		LICENSE
README.md		README.md
THIRD_PARTY_LICENSES.md		THIRD_PARTY_LICENSES.md
components.json		components.json
eslint.config.mjs		eslint.config.mjs
next.config.ts		next.config.ts
package-lock.json		package-lock.json
package.json		package.json
postcss.config.mjs		postcss.config.mjs
tsconfig.json		tsconfig.json

Folders and files

Latest commit

History

Repository files navigation

🎙️ PulseMeet — Low-Network Resilient Video Meetings & AI Meeting Intelligence

🏷️ Badges & Metrics

🌐 Live Demo

⭐ Highlights

🎯 Project Motivation

📌 Repository Overview

📑 Table of Contents

✨ Core Features

📶 Low-Network First Engineering Objective

🔐 Authentication & Profile Management

📹 Real-Time WebRTC Video Meetings

🗣 Real-Time Speech Recognition & Subtitles

🧠 Dual-Engine AI Intelligence (Cloud AI + Zero-Downtime Local Fallback)

📊 Meeting Reports & Analytics Dashboard

🏗 System Architecture

1. High-Level Application Architecture

2. Authentication & Profile Callback Flow

3. LiveKit WebRTC Session Lifecycle & Token Flow

4. Speech Recognition & AI Intelligence Pipeline

5. Network Quality Monitoring & AI Recovery Flow

⚡ Performance & Optimization

🌐 Low-Network Experience & Resilience

🔒 Security & Data Protection

🗄 Database Schema & Entity Relationship

PostgreSQL Tables

1. profiles

2. rooms

3. meeting_ai_results

Database Entity Relationship Diagram (ERD)

🔌 API Route Reference

1. GET /api/livekit-token

2. POST /api/meeting/ai

3. GET /api/test-gemini

🧩 Component Architecture

Key Pages (app/)

Core Components (components/)

🛠 Technology Stack

Core Framework & Runtime

WebRTC & Real-Time Media

Database & Authentication

AI & Speech Recognition

Styling & Utility Libraries

🚀 Installation & Setup

1. Prerequisites

2. Clone Repository & Install Dependencies

3. Setup Environment Variables

4. Database Setup (Supabase SQL Script)

5. Run Local Development Server

6. Production Build & Deployment

🔑 Environment Variables

📁 Folder Structure

🔄 End-to-End User Flow

🔧 Troubleshooting

🗺 Product Roadmap

🤝 Contributing

📜 License

🙏 Acknowledgements

👨‍💻 Author

About

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages

1. `profiles`

2. `rooms`

3. `meeting_ai_results`

1. `GET /api/livekit-token`

2. `POST /api/meeting/ai`

3. `GET /api/test-gemini`

Key Pages (`app/`)

Core Components (`components/`)