Skip to main content
Version: v0.1.0

AI Features Overview

D.Hub provides AI features throughout the entire data management and analysis workflow. Without installing separate tools, you can immediately leverage natural language conversations, code generation, intelligent editing, and knowledge-based queries right within the platform.

Available Features​

D.Hub's built-in AI features are broadly divided into five categories.

1. Chat Assistant​

An AI side panel chat accessible from every page in D.Hub.

  • Click the AI icon on the right side of the screen to open the side panel.
  • Automatically recognizes the context of the current page (ontology, collection, pipeline, etc.) to provide relevant answers.
  • Use @mentions to reference Datasets, Collections, Codes, and Pipelines in your questions.
  • AI responses can include chart (Bar/Line/Pie), map, and graph node reference visualizations.
  • Powered by the dhub2-chatway service.
Chat Assistant vs Knowledge Chat

The Chat Assistant provides AI conversations about the platform in general, while Knowledge Chat provides RAG-based conversations grounded in documents collected in a specific Knowledge. The two features can be used independently.

2. RAG AI Chat​

You can have conversations with AI based on documents and data collected in the Knowledge module.

  • AI references knowledge accumulated through web crawling, file uploads, and manual writing to provide answers.
  • Available in the Chat tab of the Knowledge detail screen (/knowledge/:id?tab=chat).
  • Uses RAG (Retrieval-Augmented Generation) to reduce hallucination and provide evidence-based answers.
  • Source documents referenced in answers are displayed alongside.
  • Powered by the dhub2-knowledge service.
RAG AI Chat Guide

For detailed information about RAG AI Chat, see the Knowledge Management > AI Chat document.

3. AI Code Generation​

Enter natural language prompts in the pipeline code editor to automatically generate Python or SQL code.

  • Activate with the shortcut Cmd+I (Mac) / Ctrl+I (Windows) or the AI button in the toolbar.
  • Generates code that automatically conforms to D.Hub pipeline input/output specifications (inputs, output).
  • Automatically references current context such as dataset schemas and pipeline structures.

4. Inline AI Editing​

You can modify already written code with AI within the code editor.

  • Select the code region to modify, press Cmd+I / Ctrl+I, and enter your modification instructions.
  • Review AI-suggested changes in Diff View, then Accept or Reject.
  • Useful for various refactoring tasks such as variable renaming, error handling additions, and logic optimization.

5. Metadata Auto-Generation​

AI automatically generates metadata for documents collected in Knowledge Settings.

  • Alias: Automatically suggests an alias that summarizes the document's core content.
  • Description: Generates a description based on document content.
  • Tags: Extracts key keywords from documents and automatically assigns tags.
  • Quickly build knowledge asset metadata without manual input.

Feature Access Locations​

AI FeatureAccess LocationShortcut
Chat AssistantAll pages → Right AI side panel—
RAG AI ChatKnowledge Detail → Chat tab—
AI Code GenerationPipeline code editor → AI buttonCmd+I / Ctrl+I
Inline AI EditingCode editor → After selecting codeCmd+I / Ctrl+I
Metadata Auto-GenerationKnowledge Settings—

How AI Features Work​

D.Hub's AI features leverage Large Language Models (LLMs).

  • Code generation and editing: The LLM analyzes the current editor context (dataset schemas, node structures, existing code) to generate or modify appropriate code.
  • RAG AI Chat: Related document chunks are found via vector search for the user's question, then passed to the LLM together to generate evidence-based answers.
  • Metadata generation: The LLM analyzes document content to perform summarization and keyword extraction.

Tips for Effective Use​

  • Specific prompts: Instead of "write some code," be specific like "filter only rows where 'status' is 'active' from the input data and count by date."
  • Step-by-step requests: For complex logic, break it into multiple steps rather than requesting everything at once for more accurate results.
  • Leverage context: When generating code in pipeline nodes with connected datasets, the AI automatically references schema information.
  • Review Diff View: After inline editing, always review changes in Diff View before applying.
  • Knowledge document quality: RAG AI Chat answer quality is proportional to the quality of collected documents. Maintain accurate and up-to-date documents.

Supported Models and Requirements​

  • AI features utilize external LLM services (OpenAI, Claude, etc.).
  • Available models may vary depending on the LLM connection settings configured by the administrator.
  • An internet connection is required to use AI features.

Next Steps​