-
Notifications
You must be signed in to change notification settings - Fork 0
Features
Comprehensive feature catalog covering Indexa's complete capabilities.
Indexa combines multiple retrieval and ranking approaches to deliver high-quality search results:
Parallel Retrieval System
- Simultaneous execution of full-text search and semantic vector search
- Full-text search contributes 60% to final ranking using BM25 algorithm
- Semantic search contributes 20% using vector embeddings for conceptual matching
- PageRank authority signals contribute 10% based on link analysis
- Domain authority contributes 10% based on site credibility scoring
Query Intelligence
- Automatic query expansion for concise searches (three words or fewer)
- Expands queries with relevant synonyms and acronyms to improve result coverage
- Spell correction system suggests alternatives for misspelled queries
- Levenshtein distance algorithm powers "Did you mean?" functionality
Search Assistance
- Real-time autocomplete with query suggestions
- Keyboard navigation support for accessibility
- 300ms debounce optimization for smooth user experience
Intelligent Content Discovery
- Automatic sitemap.xml detection and processing
- Priority-based crawling respects site-defined importance
- Comprehensive robots.txt compliance with crawl delay respect
- Blocked path detection and adherence to site preferences
Modern Web Support
- Two-phase rendering strategy for optimal performance
- Static HTML processing for traditional websites
- JavaScript rendering for single-page applications (React, Vue, Angular)
- Browser automation for dynamic content capture
Content Quality
- Near-duplicate detection using SimHash fingerprinting
- 64-bit fingerprints for efficient similarity comparison
- Structured data extraction from schema.org markup
- Support for Recipe, Product, Event, and Article types
An integrated conversational AI system designed to enhance search with natural language interaction:
Intelligent Query Processing
- Multi-mode chat interface for versatile interactions
- Intent classification distinguishes factual, conversational, and computational queries
- Context aggregation ranks and synthesizes relevant search results
- Multi-query detection splits complex questions into manageable sub-queries
Response Generation
- Real-time token streaming for natural conversation flow
- Source citations with clickable links to original content
- Markdown rendering with code block syntax highlighting
- Conversation history with ability to start fresh chats
Specialized Modes
- Chat mode for general questions and discussions
- Image mode for visual search queries
- Code mode (activated with /code) for programming assistance
- Quick reply system for greetings and common interactions
Knowledge Integration
- Wikipedia fallback for comprehensive knowledge coverage
- Multiple data source aggregation for thorough answers
- Featured answer extraction for direct responses
| Mode | Description |
|---|---|
| Chat | General conversational search assistance |
| Image | Visual content discovery and questions |
| Video | Coming soon |
Interactive Immersive Search Engine (Coming Soon)
Next-generation spatial search experience currently in development:
- Immersive three-dimensional navigation of search results
- Interactive knowledge exploration with visual relationships
- Topic relationship mapping for connected learning
- Spatial organization of information for intuitive browsing
Dedicated shopping search functionality with specialized product indexing:
Product Database
- Full-text indexed product catalog
- Multi-site aggregation from major Indian e-commerce platforms
- Supported retailers: Croma, Vijay Sales, Myntra
- Structured product data with complete metadata
Advanced Filtering
- Price range selection and sorting
- Rating and review filtering
- Brand-specific searches
- Category browsing
- Stock availability status
Sorting Options
- Relevance-based ranking (default)
- Price ascending and descending
- Customer rating sorting
- Featured product highlighting
Product Display
- Rich product cards with images and details
- Price, rating, and availability at a glance
- Direct links to purchase pages
- Featured products for best matches
Visual content discovery across indexed web pages:
Search Capabilities
- Full-text search across indexed images
- Alt text and caption matching
- Surrounding page content analysis for context
- BM25 relevance ranking for results
Result Information
- Image URL and dimensions
- Source page reference
- Image metadata and context
- Direct links to original content
Enhanced result cards for specific content types:
| Card Type | Information Displayed |
|---|---|
| Recipe | Star ratings, cooking time, calories, ingredients |
| Event | Dates, venue, tickets, description |
| Product | Pricing, availability, reviews, specs |
| Article | Author, publication date, reading time |
| FAQ | Collapsible Q&A format |
Comprehensive entity information display:
- Wikipedia integration for authoritative information
- Brand homepage prioritization for official sources
- Domain authority scoring (0-100 scale)
- Featured answer extraction for direct information access
Redis-powered caching for optimal performance:
| Cache Layer | TTL | Purpose |
|---|---|---|
| Search results | 1 hour | Query responses |
| Embeddings | 24 hours | Vector representations |
| Autocomplete | 10 minutes | Suggestions |
Performance Improvements
- 15x speedup for cached queries
- Response times: 10-20ms cached vs 150-300ms uncached
- 80-85% cache hit rate after system warm-up
Horizontal scaling for large-scale web indexing:
- Redis-based FIFO queue architecture
- LPUSH/BRPOP for atomic job distribution
- Distributed locking prevents duplicate crawls
- 10 workers capable of 1000+ pages per second
Database partitioning for performance at scale:
- Consistent hashing by domain
- Independent SQLite databases per shard
- Parallel search execution across shards
- Configurable shard count (4-16 default range)
Optimized approximate nearest neighbor search:
- Hierarchical Navigable Small World graph structure
- O(log N) search complexity
- Int8 quantization for 75% memory reduction
- Custom implementation for semantic search needs
Secure user authentication and profile management:
Security Features
- User registration and login system
- bcrypt password hashing for secure storage
- Session management with secure cookies
- Separate authentication database (owlguard.db)
User Features
- Profile dashboard with customization
- Upload support for profile pictures
- Session persistence across visits
- Secure logout functionality
API Endpoints
| Endpoint | Purpose |
|---|---|
/api/auth/signup |
Registration |
/api/auth/login |
Login |
/api/auth/logout |
Logout |
Advanced authority scoring for result ranking:
PageRank Algorithm
- Iterative link analysis computation
- Incremental updates for efficiency
- Authority score from web graph structure
- 0-10 scale for ranking contribution
Domain Authority
- Per-domain credibility scoring system
- 0-100 scale similar to industry standards
- Multiple signal aggregation
- Historical performance tracking
Brand Detection
- Automatic identification of brand queries
- Official homepage prioritization
- Authority boost for verified sources
- Zero-Result Prevention: Query expansion reduces failed searches by 15-20%
- Result Diversity: Maximum 2 results per domain in standard results
- Intent Optimization: Context-aware ranking adjustments
- Mobile-responsive design
- Lucide icon system for clarity
- Intuitive navigation patterns
- Fast, smooth interactions
- All processing performed locally
- No external API tracking
- No personally identifiable information sharing
- User data remains within system boundaries
| Component | Technology |
|---|---|
| Language | Full TypeScript implementation |
| Database | SQLite with FTS5 |
| Embeddings | ONNX runtime (local) |
| Server | Fastify |
| Model | Xenova/bge-base-en-v1.5 |
Available npm scripts for system administration:
- Cache warming and clearing
- Crawl coordination and worker management
- Index shard migration
- PageRank computation
- Domain authority calculation
Feature Requests: feedback@indexa.site
General Support: support@indexa.site
Feature set subject to ongoing development and improvement.
© 2026 Indexa Inc. • Privacy Policy • Terms of Service • indexa.site