Knowledge Base
GPTWeb Two-Phase Semantic RAG Engine
GPTWeb's Two-Phase Semantic RAG Engine is a breakthrough architecture that solves the fundamental challenge of AI knowledge retrieval at scale. Here's how it works:
Two-Phase Architecture
Phase 1: Collection Routing
When a visitor asks a question, the system first performs semantic matching against your collection descriptions to identify which knowledge domains are relevant. This acts as intelligent routing, dramatically reducing the search space. Phase 2: Document Search
Only after relevant collections are identified does the engine perform deep semantic search within those collections, finding the most relevant document chunks to include in the AI response.
When a visitor asks a question, the system first performs semantic matching against your collection descriptions to identify which knowledge domains are relevant. This acts as intelligent routing, dramatically reducing the search space. Phase 2: Document Search
Only after relevant collections are identified does the engine perform deep semantic search within those collections, finding the most relevant document chunks to include in the AI response.
Two-Phase Semantic RAG Flow
Multi-Media Content Support
| Content Type | Processing Method | AI Capability |
|---|---|---|
| PDF Documents | Text extraction and chunking | Full semantic search |
| Word/DOCX | Native text parsing | Full semantic search |
| Spreadsheets (XLSX/CSV) | Data extraction for charts | AI-generated visualizations |
| Images | Stored with metadata | Display in responses |
| Videos | Transcription optional | Playback in chat |
| Flipbooks | Slide-by-slide indexing | Interactive presentations |
RAG Crawler: Accelerate Time to Value
The Web Crawler dramatically speeds up knowledge base population by automatically extracting content from existing websites: 1. Access: Admin Panel → App Configuration → RAG Search Tuning → Crawl Content
Configure: Set URL, depth, page limits, and include/exclude patterns
Extract: Automatically pulls HTML, PDFs, images, and metadata
Curate: Visual review interface to approve/reject/edit content
Categorize: AI-assisted collection assignment
Upload: Batch import to your knowledge base
Executive Business Value
95% fail without good data
AI Pilot Success
Hours vs. Weeks
Time to Value
2-phase precision
Search Accuracy
For Executives: Traditional AI chatbots fail because they lack accurate, current business knowledge. GPTWeb's two-phase architecture ensures: - Accuracy: Responses grounded in YOUR content, not hallucinations
- Scalability: Collection routing handles thousands of documents efficiently
- Speed: Crawler imports existing website content in hours, not weeks
- Control: BYOK model means you control AI costs directly
- Rich Experiences: Videos, charts, and presentations display directly in conversations The result? An AI-ready data foundation that transforms static websites into intelligent, conversational experiences that qualify leads and drive revenue. GPTWeb is the future of engagement - websites and marketing automation combined, built for the AI era, built for now.