Vector Hybrid RAG + Knowledge Graph Visualization

Talk to Your Documents,
Visualize the Knowledge Network

The next-generation document platform. Upload scanned PDFs, extract interactive knowledge networks, chat across multiple resources, and trace exact sources with real-time citations.

Engineered for Technical Excellence

Powered by a production-ready AI stack including hybrid dense-sparse search, cross-encoder rerankers, and automated RAG evaluation.

Knowledge Graph Network

Automatically extracts core entities and relationships from PDFs to draw interactive SVG connection graphs. Visualize concepts rather than scanning tables of contents.

Hybrid Retrieval + Reranking

Combines semantic vector search (`pgvector`) with exact keyword match (`Postgres FTS`) via Reciprocal Rank Fusion, followed by a BGE Cross-Encoder reranker. High relevance guaranteed.

RAG Evaluation (LLM-as-a-Judge)

Monitors response quality in real-time. Calculates and plots Faithfulness, Answer Relevancy, and Context Precision metrics to guarantee system trust and avoid hallucinations.

Multi-Format Document Parsing

Support for PDFs, DOCX, and TXT. Features intelligent scanned PDF detection using text density analysis, with fallback to PyTesseract OCR if standard parsing fails.

Source Citations & Highlighting

Every assistant claim contains clean reference tags. Clicking a citation highlights the exact paragraph source inside the preview pane. No more guessing.

Resume-Grade Design System

Glassmorphism workspace, collapsible panels, interactive Recharts dashboard, skeleton loaders, and responsive layouts built with Next.js 15, Framer Motion, and Tailwind CSS.

Interactive Interface

Experience the dual-pane workspace where you chat on the right, and review source citations or knowledge nodes on the left.

app.smartdocai.com/dashboard/chat/annual_report
Knowledge Graph View14 Entities
User NodeAI AgentPostgreSQLRAG Pipeline
Assistant ChatRAG metrics: Grounded 0.98
User: What database does the project utilize?
Assistant: The platform utilizes PostgreSQL with the pgvector extension [1] for high performance vector storage and dense semantic queries. It also includes local FAISS support [2].