SMEAI

Subject Matter Expert AI
RAG-powered Subject Matter Expert AI built with Next.js & Gemini

🧠 What is SMEAI?

SMEAI transforms your documents into an AI-powered subject matter expert. Upload your documents, ask questions, and get expert answers grounded in your data.

Key Features:

📄 Document-Based RAG — Upload PDFs, TXT, JSON, MD files
🤖 Gemini-Powered Responses — Streaming AI answers using Google's Gemini 2.5 Flash
🔑 Custom API Keys — Use your own Gemini API key for personal quota
🔍 Semantic Search — Vector embeddings for intelligent context retrieval
🔒 Secure Authentication — Google OAuth via Supabase
⚡ Real-Time Streaming — Modern chat experience with SSE
🎨 Beautiful UI — Built with Next.js 16, Tailwind CSS & shadcn/ui
📤 Export Conversations — Download chat history as JSON

🏗️ Architecture

Documents → Text Extraction → Chunking → Vector Embeddings → 
  → Semantic Search → Context Injection → Gemini → SMEAI Response

Tech Stack:

Frontend: Next.js 16 (App Router), React 19, TypeScript
UI: Tailwind CSS 4, shadcn/ui, Lucide Icons, react-markdown
AI/ML: Google Gemini 2.5 Flash, LangChain, Custom Vector Store
Auth: Supabase (Google OAuth) with SSR support
Notifications: Sonner (toast notifications)
Deployment: Vercel-ready

🚀 Quick Start

Prerequisites

Node.js 18+ and npm/yarn
Google Gemini API Key (Get one here)
Supabase Project (Create one)

Installation

Clone the repository

git clone https://github.com/yourusername/smeai.git
cd smeai

Install dependencies

npm install

Set up environment variables

Create a .env.local file:

# Google Gemini API
GEMINI_API_KEY=your_gemini_api_key_here

# Supabase Authentication
NEXT_PUBLIC_SUPABASE_URL=your_supabase_project_url
NEXT_PUBLIC_SUPABASE_ANON_KEY=your_supabase_anon_key

Configure Supabase Authentication
- Go to Supabase Dashboard → Authentication → Providers
- Enable Google OAuth
- Add your site URL and redirect URL in Site URL settings
- Add authorized redirect URLs:
  - Development: http://localhost:3000/auth/callback
  - Production: https://yourdomain.com/auth/callback
Run the development server

npm run dev

Open http://localhost:3000

📖 Usage

1. Sign In

Click "Sign in with Google" on the landing page
Authenticate via Supabase

2. Upload Documents

Click the upload button (📎 icon)
Select a document (TXT, PDF, JSON, MD)
Maximum 2 documents, 2MB each
Documents are chunked and vectorized automatically

3. Ask Questions

Type your question in the chat input
SMEAI retrieves relevant context from your documents
Receives streaming AI responses grounded in your data
Press Enter to send, Shift+Enter for new line

4. Manage Documents

View uploaded documents with chunk counts
Delete documents with the ❌ icon
Maximum 2 documents to keep the system lightweight

5. Custom API Key (Optional)

Click the "API key" button in the top navigation
Enter your own Gemini API key to use your personal quota
Key is stored locally and encoded for security
Remove it anytime to use the default server API key

6. Export Conversations

Click "Export as JSON" to download your chat history
Includes messages, document count, and metadata

📁 Project Structure

smeai/
├── app/
│   ├── auth/
│   │   └── callback/        # OAuth callback handler
│   ├── chat/                # Chat interface (protected)
│   ├── api/
│   │   ├── chat/            # Streaming chat endpoint
│   │   ├── documents/       # Document management (GET/DELETE)
│   │   └── upload/          # Document upload & ingestion
│   ├── layout.tsx           # Root layout
│   └── page.tsx             # Landing page
├── lib/
│   ├── ai/
│   │   └── gemini.ts        # Gemini AI configuration
│   ├── rag/
│   │   └── vector.ts        # Vector store & embeddings
│   └── supabase/
│       ├── client.ts        # Supabase client (browser)
│       ├── server.ts        # Supabase server client
│       └── proxy.ts         # Auth middleware/proxy
├── components/
│   ├── common/
│   │   └── TopNav.tsx       # Navigation component
│   ├── Prism.tsx            # Animated background component
│   └── ui/                  # shadcn/ui components
├── proxy.ts                 # Next.js middleware entry
├── vector_store.json        # Persistent vector storage (gitignored)
└── README.md

🔐 Authentication Flow

User lands on / (landing page)
Clicks "Sign in with Google"
Redirects to Supabase OAuth
Returns to /auth/callback
Redirects to /chat (protected route)
Authenticated users can access chat and upload documents

🧩 How RAG Works

Document Ingestion

Upload document via /api/upload
Extract text content
Split into chunks (1000 chars, 200 overlap)
Generate embeddings using Gemini
Store in persistent vector store

Query & Retrieval

User asks a question
Question is embedded using Gemini embeddings
Semantic similarity search finds top 3 relevant chunks
Context is injected into Gemini 2.5 Flash prompt
Gemini generates answer based on provided context
Response streams to client in real-time via SSE
Custom API keys are supported (stored in localStorage, encoded)

🎨 UI Features

Current Features

✅ Clean, modern chat interface with animated Prism background
✅ SMEAI branding with Brain icon and glass-morphism navigation
✅ Avatar-based message distinction (User/Bot)
✅ Rich markdown rendering for AI responses (code blocks, lists, tables)
✅ Auto-scrolling chat area with user scroll detection
✅ Fixed input area at bottom
✅ Document upload with progress indicators
✅ Document management (view chunk counts, delete)
✅ Toast notifications for all actions (Sonner)
✅ Error handling and loading states
✅ Empty state with example question prompt
✅ Export conversation as JSON
✅ Custom API key management dialog
✅ Clear chat functionality
✅ Mobile-responsive design

🚀 Deployment

Deploy to Vercel

Connect your repository to Vercel
Add environment variables in Vercel Dashboard:
- GEMINI_API_KEY
- NEXT_PUBLIC_SUPABASE_URL
- NEXT_PUBLIC_SUPABASE_ANON_KEY
Update Supabase redirect URLs:
- Add your Vercel domain to authorized redirect URLs
Deploy:

npm run build

The vercel.json is configured to handle all routing correctly.

⚙️ Configuration

Vector Store

Storage: File-based (vector_store.json)
Persistence: Survives server restarts
Embeddings: Gemini Embedding API (gemini-embedding-001)
Similarity: Cosine similarity calculation
Scalability: For production, migrate to Pinecone, Supabase Vector, or Weaviate

Document Limits

Max documents: 2
Max file size: 2MB per document
Supported formats: TXT, PDF, JSON, MD

Chunking Strategy

Chunk size: 1000 characters
Overlap: 200 characters
Splitter: LangChain Recursive Character Text Splitter
Metadata: Each chunk tagged with document ID and filename

🛠️ Development

Available Scripts

npm run dev          # Start development server
npm run build        # Build for production
npm run start        # Start production server
npm run lint         # Lint code
npm run lint:fix     # Auto-fix linting issues

Adding UI Components

npx shadcn@latest add [component-name]

🧪 Troubleshooting

Issue: "Found relevant docs: 0"

Cause: Vector store not persisting or documents not uploaded
Fix: Check vector_store.json exists and contains data

Issue: 429 Rate Limit Error

Cause: Gemini API quota exceeded
Fix: Wait for quota reset or upgrade API plan

Issue: Authentication not working

Cause: Missing or incorrect Supabase credentials
Fix: Verify .env.local and Supabase dashboard settings

Issue: Route not found on Vercel

Cause: Missing routing configuration
Fix: Ensure vercel.json is present and configured correctly

Issue: Custom API key not working

Cause: API key not properly encoded or invalid
Fix: Check that the API key is valid and try saving it again. The key is base64 encoded in localStorage.

📝 License

MIT License - feel free to use this project for learning and production.

🙏 Acknowledgments

Google Gemini for powerful AI capabilities (Gemini 2.5 Flash & Embeddings)
LangChain for RAG tooling and text splitting
Supabase for seamless authentication with SSR support
shadcn/ui for beautiful, accessible components
Sonner for elegant toast notifications
Vercel for effortless deployment

📬 Contact

Built with ❤️ for creating domain-specific AI experts.

For questions or contributions, open an issue or pull request!

SMEAI — Your AI-powered Subject Matter Expert.

Name		Name	Last commit message	Last commit date
Latest commit History 10 Commits
app		app
components		components
lib		lib
.gitignore		.gitignore
README.md		README.md
components.json		components.json
eslint.config.mjs		eslint.config.mjs
example.env		example.env
next.config.ts		next.config.ts
package-lock.json		package-lock.json
package.json		package.json
postcss.config.mjs		postcss.config.mjs
proxy.ts		proxy.ts
tsconfig.json		tsconfig.json
vercel.json		vercel.json

Folders and files

Latest commit

History

Repository files navigation

SMEAI

🧠 What is SMEAI?

🏗️ Architecture

🚀 Quick Start

Prerequisites

Installation

📖 Usage

1. Sign In

2. Upload Documents

3. Ask Questions

4. Manage Documents

5. Custom API Key (Optional)

6. Export Conversations

📁 Project Structure

🔐 Authentication Flow

🧩 How RAG Works

Document Ingestion

Query & Retrieval

🎨 UI Features

Current Features

🚀 Deployment

Deploy to Vercel

⚙️ Configuration

Vector Store

Document Limits

Chunking Strategy

🛠️ Development

Available Scripts

Adding UI Components

🧪 Troubleshooting

Issue: "Found relevant docs: 0"

Issue: 429 Rate Limit Error

Issue: Authentication not working

Issue: Route not found on Vercel

Issue: Custom API key not working

📝 License

🙏 Acknowledgments

📬 Contact

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Uh oh!

Contributors

Uh oh!

Languages

Packages