Unstructured
HTML
<div class="p">
<p>...</p></div>
Scraping & NLP Parser
High-Overhead Extraction
DOM
Parsing
HTML
Cleaning
Entity
NLP
Structured
JSON-LD
{"@type":
"Article"}
JSON Parser
Direct Low-Overhead Ingestion
JSON.parse() → Direct Tokens
LLM Vector
DB Index
Semantic Embeddings
[0.012, -0.438, ...]
[0.891, 0.104, ...]
Knowledge Graph
Entity → Relation
Subject → Predicate
Unified Index
RAG Ready
Query Search