Stashly

A personal bookmarking application with AI-powered analysis, summarization, and tagging. Save links and let Gemini automatically analyze content to generate intelligent summaries and tags.

Features

• Smart Bookmarking: Save any URL with an optional personal note
• AI-Powered Analysis: Gemini automatically fetches and analyzes web pages using URL Context and Google Search Grounding
• Intelligent Summaries: Generates both concise and detailed summaries of content
• Auto-Tagging: Automatically categorizes bookmarks with relevant tags
• Full-Text Search: Search bookmarks by title, URL, content, or tags
• Tag Filtering: Browse bookmarks by specific tags
• User Authentication: Powered by Stack Auth (formerly Neon Auth)
• Database: Neon Postgres with Drizzle ORM
• UI: Tailwind CSS + shadcn/ui components

Prerequisites

• Neon account
• Stack Auth account (or use Neon Auth)
• Google AI Studio API key for Gemini
• Node.js 18+ installed locally

Local Development Setup

1. Set up Stack Auth

1. Create or open a Neon project
2. Go to Neon Auth → Setup instructions
3. Click Set up Auth to generate your configuration
4. Copy these environment variables to .env.local:
   • NEXT_PUBLIC_STACK_PROJECT_ID
   • NEXT_PUBLIC_STACK_PUBLISHABLE_CLIENT_KEY
   • STACK_SECRET_SERVER_KEY
   • DATABASE_URL

2. Set up Google AI API

1. Get your API key from Google AI Studio
2. Add to .env.local:
   • GEMINI_API_KEY

3. Install dependencies and run migrations

# Install all dependencies
npm install

# Generate database migrations
npm run drizzle:generate

# Apply migrations to your database
npm run drizzle:migrate

4. Start the development server

npm run dev

Visit http://localhost:3000 to see the application running. Sign in and start adding bookmarks!

How It Works

Bookmark Processing Pipeline

1. User submits a URL with optional note and tags
2. URL normalization - Removes tracking parameters, www prefix, etc.
3. AI Analysis - Google Gemini 2.5 Flash uses URL Context and Grounding to:
   • Fetch and analyze the web page content
   • Generate a human-friendly title
   • Create short summary (1-2 sentences)
   • Create detailed multi-paragraph summary
   • Detect content language
   • Generate 3-5 relevant tags
   • Identify content category
4. Database storage - Saves all data with full-text search indexing
5. Status update - Marks as processed or failed

Technology Details

URL Context + Grounding: Instead of manually extracting HTML content, the app uses Gemini's built-in capabilities:

• URL Context Tool: Directly fetches and processes content from the provided URL
• Google Search Grounding: Provides additional web context when needed
• Supports multiple formats: HTML, PDF, images, JSON, and more
• Uses cached content when available for cost optimization

Database Schema

• users - User accounts
• bookmarks - Core bookmark data (URL, title, description, status)
• bookmark_contents - Extracted content and AI summaries
• tags - User-specific tags (normalized)
• bookmark_tags - Many-to-many relationship

API Routes

• POST /api/bookmarks - Create new bookmark
• GET /api/bookmarks - List with search/filter
• GET /api/bookmarks/[id] - Get single bookmark
• POST /api/bookmarks/[id]/reprocess - Retry failed processing
• GET /api/tags - Get all tags with counts

Project Structure

app/
├── api/
│   ├── bookmarks/
│   │   ├── [id]/
│   │   │   ├── route.ts           # GET single bookmark
│   │   │   └── reprocess/
│   │   │       └── route.ts       # POST retry processing
│   │   └── route.ts               # POST create, GET list bookmarks
│   └── tags/
│       └── route.ts               # GET all tags with counts
├── b/
│   └── [id]/
│       └── page.tsx               # Bookmark detail page
├── handler/
│   └── [...stack]/
│       └── page.tsx               # Stack Auth handler
├── schema/
│   └── schema.ts                  # Drizzle ORM schema (all tables)
├── bookmarks-list.tsx             # Main bookmarks list component
├── bookmarks-page.tsx             # Bookmarks page wrapper
├── add-bookmark.tsx               # Add bookmark form component
├── db.tsx                         # Database helper with auth
├── header.tsx                     # App header component
├── layout.tsx                     # Root layout
├── page.tsx                       # Home page
└── stack.tsx                      # Stack Auth configuration

components/
├── bookmarks/
│   ├── bookmark-card.tsx          # Bookmark card UI component
│   ├── favicon.tsx                # Favicon display component
│   └── search-filter.tsx          # Search and tag filter UI
└── ui/
    ├── button.tsx                 # shadcn/ui button
    ├── card.tsx                   # shadcn/ui card
    └── input.tsx                  # shadcn/ui input

lib/
├── bookmark-utils.ts              # URL normalization, tag management
├── bookmark-processor.ts          # Main bookmark processing pipeline
├── llm-client.ts                  # Google Gemini AI integration (URL Context + Grounding)
└── utils.ts                       # General utilities (cn helper)

drizzle/
├── 0000_lean_dragon_man.sql       # Initial migration
├── 0001_dry_caretaker.sql         # Bookmarks schema migration
├── 0002_add_gin_index.sql         # Full-text search index
└── meta/                          # Drizzle metadata
    ├── _journal.json
    ├── 0000_snapshot.json
    └── 0001_snapshot.json

Production Setup

• Neon DB lives here: https://console.neon.tech/app/projects/shiny-sunset-02500855
• Deployed on Vercel: https://vercel.com/prakhars-projects-76dc8cec/bookmarks

Future Enhancements

• Semantic search with pgvector embeddings
• Export/import functionality