Skip to main content
Version: v0.1.0

Settings

In the Settings tab of the Knowledge detail screen, you manage metadata (name, description, tags) and search-related options (search methods, embedding model). AI auto-generation features can be used to efficiently populate metadata, and you can understand the impact of search option changes on existing data to safely adjust settings.


General Settings

The area for managing basic Knowledge metadata. Each field can be entered manually or generated using AI.

Alias

The display name of the Knowledge. This name is used in the list screen, search results, etc.

ItemDescription
Manual InputEnter the desired name directly in the text field
AI GenerationClick the Generate button to have AI automatically suggest an appropriate name based on collected document content

Description

The description text of the Knowledge. Describe the purpose, included content, target domain, etc.

ItemDescription
Manual InputWrite a description freely in the text area
AI GenerationClick the Generate button to have AI analyze collected documents and automatically generate a summary description

Tags

A tag list for categorizing the Knowledge. Used for filtering and searching on the list screen.

ItemDescription
Manual InputManually add or remove tags
AI GenerationClick the Generate button to have AI analyze document content and suggest relevant tags. Tags are merged with existing tags, and duplicates are automatically removed
tip

Systematically managing tags enables quick discovery even as the number of Knowledges grows. It's recommended to establish tag naming conventions (e.g., domain prefixes) within your team in advance.

Collections

Link Knowledge to collections to systematically group resources.

ActionDescription
Add CollectionSelect an existing collection from the dropdown to link
Remove CollectionClick the remove button next to a linked collection
Create New CollectionUse the creation option at the bottom of the dropdown to create a new collection and link it immediately

Generate All

A button that auto-generates Alias, Description, and Tags all at once using AI. It comprehensively analyzes collected document content to fill all three fields simultaneously. This produces more consistent results than generating each field individually.

info

The AI generation feature requires collected documents in the Knowledge to provide accurate results. Results may be of lower quality when there are no or few documents.


Search Options

The area for configuring Knowledge search methods and embedding models. These settings determine how data is stored during document indexing, so changes are restricted when existing data is present.

Search Methods

Select search methods for the Knowledge using checkboxes. At least one must be selected.

Search MethodEngineDescription
VECTORVector DBSemantic similarity search based on embedding vectors
TEXTText search engineBM25-based keyword full-text search
GRAPHGraph DBEntity/relationship extraction-based graph exploration
Search Method Selection Guide

In most cases, the VECTOR + TEXT combination is most effective. Vector search ensures semantic similarity, while Text search supplements keyword accuracy to maximize Hybrid search quality. Add GRAPH only for special cases where entity relationship exploration is needed.

Embedding Model

When the VECTOR search method is selected, specify the embedding model. Selecting a model from the dropdown displays its dimensions (vector dimensions) and provider information.

Display InfoDescription
Model NameUnique name of the embedding model
DimensionsNumber of dimensions of the generated vector (higher is more precise but increases storage)
ProviderModel provider (e.g., local, openai)

Search quality and performance vary depending on the embedding model. For detailed model comparisons, see the Chunking and Options document.

Search Options Lock

For Knowledges that already have indexed documents, Search Methods and Embedding Model become locked. This is because existing indexed data may not be compatible with new settings. To change search options, you must first perform a Settings Reset as described below.


Saving Setting Changes

After modifying General Settings items (Alias, Description, Tags, Collections), click the Save button to save changes. Changes will be lost if you navigate to another tab without saving.


Settings Reset

When you need to change Search Options (Search Methods, Embedding Model), all existing indexed data must be reset.

Reset Procedure

  1. Click the Reset button at the bottom of the Settings tab
  2. Confirm the reset scope in the confirmation modal
  3. Click Confirm to execute the reset

Reset Scope

When Reset is executed, data is deleted in the following order:

OrderDeletion TargetDescription
1All ChunksIndexed chunk data (Vector DB, Text DB, Graph DB)
2All DocumentsCollected document metadata and original references

Once reset is complete, the Search Methods and Embedding Model fields are unlocked so you can set new values.

Settings reset cannot be undone

Settings reset permanently deletes all indexed data for that Knowledge. All chunks and document metadata stored in Vector DB, Text DB, and Graph DB are removed. This action cannot be undone, so only execute it when reset is truly necessary.

Re-collection After Reset

After reset, you must reconfigure search options and re-collect documents for indexing. You can add documents again using the same methods as before — web crawling, file upload, and manual input.


Precautions When Changing Settings

ItemCan Be ChangedNotes
Alias✅ AnytimeNo impact on existing data
Description✅ AnytimeNo impact on existing data
Tags✅ AnytimeNo impact on existing data
Collections✅ AnytimeNo impact on existing data
Search Methods⚠️ ConditionalLocked when documents exist → Reset required
Embedding Model⚠️ ConditionalLocked when documents exist → Reset required

Next Steps

  • Chunking and Options — Detailed reference for chunking strategies, embedding models, and indexing options
  • Search Test — Verify the impact of changed settings on search quality
  • AI Chat — Utilize knowledge through RAG-based conversation