Skip to main content
Version: v0.1.0

Knowledge Management

D.Hub's Knowledge module is a platform that transforms unstructured data into searchable knowledge.

Core Features

The Knowledge module provides the following features:

  • Document Collection: Collect data from various sources including web crawling, file uploads, and manual input
  • AI Chat: Have conversations with AI about collected knowledge using RAG (Retrieval-Augmented Generation)
  • Search Test: Validate knowledge quality using Vector, Full-Text, and Hybrid search modes
  • Knowledge Settings: Manage storage targets, embedding models, and metadata

Data Sources

Knowledge can collect data through three methods:

SourceDescriptionSupported Formats
Web CrawlingAutomatically collect web pages based on URLsHTML, dynamic pages (JS rendering)
File UploadUpload document files directly for processingPDF, DOCX, PPTX, XLSX, HTML, TXT, MD
Manual InputWrite and register text chunks directlyFree-form text

Multi-Store Architecture

Collected data is simultaneously stored in up to three stores depending on the purpose:

StoreEnginePurpose
Vector DBVector search engineSemantic similarity search (embedding vectors)
Text DBText search engineKeyword-based full-text search (BM25)
Graph DBGraph databaseEntity/relationship-based graph exploration
tip

You can select storage targets when creating a Knowledge. In most cases, the Vector + Text combination is most effective.

Knowledge Detail Screen

When you select a Knowledge, a detail screen with the following 4 tabs is displayed:

TabDescription
DocumentsManage collected document list and add new documents
ChatAsk questions about knowledge via RAG-based AI chat
SearchTest search queries and verify result quality
SettingsManage metadata, storage options, and embedding models

Next Steps