> For the complete documentation index, see [llms.txt](https://docs.clickai.vn/clickai-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.clickai.vn/clickai-docs/clickai-docs-en/database/manage-knowledge-content-documents-and-chunks.md).

# Manage Knowledge Content — Documents & Chunks

## Table of Contents

·       \[Manage Documents]\(#manage-documents)

·       \[Manage Chunks]\(#manage-chunks)

·       \[Best Practices]\(#best-practices)

&#x20;

## Manage Documents

On the Knowledge Base detail page, you can manage all uploaded documents:

### Document Actions

<table data-header-hidden><thead><tr><th valign="top"></th><th valign="top"></th></tr></thead><tbody><tr><td valign="top">Action</td><td valign="top">Description</td></tr><tr><td valign="top">View</td><td valign="top">View all documents, processing status, and details</td></tr><tr><td valign="top">Add</td><td valign="top">Upload new documents to the Knowledge Base</td></tr><tr><td valign="top">Enable / Disable</td><td valign="top">Toggle documents. Disabled documents won't be retrieved but are still stored</td></tr><tr><td valign="top">Archive</td><td valign="top">Archive documents no longer frequently used</td></tr><tr><td valign="top">Delete</td><td valign="top">Permanently delete documents from the Knowledge Base</td></tr><tr><td valign="top">Re-index</td><td valign="top">Re-index documents when settings are changed</td></tr></tbody></table>

&#x20;

### Document Status

<table data-header-hidden><thead><tr><th valign="top"></th><th valign="top"></th></tr></thead><tbody><tr><td valign="top">Status</td><td valign="top">Description</td></tr><tr><td valign="top">Queuing</td><td valign="top">Waiting to be processed</td></tr><tr><td valign="top">Processing</td><td valign="top">Being chunked and indexed</td></tr><tr><td valign="top">Available</td><td valign="top">Ready for retrieval</td></tr><tr><td valign="top">Error</td><td valign="top">Processing failed — click for details</td></tr><tr><td valign="top">Disabled</td><td valign="top">Turned off — excluded from retrieval</td></tr><tr><td valign="top">Archived</td><td valign="top">Archived</td></tr></tbody></table>

&#x20;

💡 TIP: Use Enable/Disable instead of Delete when you want to temporarily remove documents from retrieval. You can re-enable them anytime without re-uploading.

&#x20;

## Manage Chunks

Click on a document to view and manage its chunks:

### Chunk Actions

<table data-header-hidden><thead><tr><th valign="top"></th><th valign="top"></th></tr></thead><tbody><tr><td valign="top">Action</td><td valign="top">Description</td></tr><tr><td valign="top">View</td><td valign="top">View detailed chunk content</td></tr><tr><td valign="top">Edit</td><td valign="top">Modify chunk content</td></tr><tr><td valign="top">Add</td><td valign="top">Manually add a new chunk</td></tr><tr><td valign="top">Enable / Disable</td><td valign="top">Toggle individual chunks</td></tr><tr><td valign="top">Delete</td><td valign="top">Delete a chunk</td></tr><tr><td valign="top">Add Image</td><td valign="top">Attach images to a chunk</td></tr></tbody></table>

&#x20;

### Notes on Editing Chunks

With Parent-Child Mode:

·       When editing a parent chunk, you can choose to regenerate its child chunks or keep them unchanged

·       Editing a child chunk does not update its parent chunk

⚠️ IMPORTANT: When editing chunks, new content is re-embedded automatically. However, test retrieval after editing to ensure results remain accurate.

&#x20;

## Best Practices

### Check Chunk Quality

After creating a Knowledge Base, review chunks to identify issues:

·       Chunks too short — May lack sufficient context, leading to semantic loss and inaccurate answers

·       Chunks too long — May include irrelevant information, introducing semantic noise and lowering retrieval precision

·       Semantically incomplete chunks — Caused by forced chunking that cuts through sentences or paragraphs

### Use Child Chunks as Retrieval Hooks for Parent Chunks

In Parent-Child mode, child chunks serve as "hooks" — retrieval entry points, after which the parent chunk provides fuller context.

Effective child chunk types:

<table data-header-hidden><thead><tr><th valign="top"></th><th valign="top"></th></tr></thead><tbody><tr><td valign="top">Type</td><td valign="top">Example</td></tr><tr><td valign="top">Keywords</td><td valign="top">blinking light, won't turn on, red light, connection error, frozen</td></tr><tr><td valign="top">Summaries</td><td valign="top">Guide to interpreting LED colors and troubleshooting hardware issues</td></tr><tr><td valign="top">Queries</td><td valign="top">What does a solid red light mean?</td></tr></tbody></table>

&#x20;

### Use Summaries to Bridge Query-Content Gaps

Summaries are especially useful when:

·       User queries differ from document language: Add summaries in the way users actually ask questions

·       Concepts are implicit or buried in details: Add high-level summaries that surface core concepts

·       Raw text is non-textual: When a chunk is primarily code, tables, or logs — add descriptive summaries

·       Related chunks should be retrieved together: Apply identical summaries for grouped retrieval

&#x20;

*📖 Previous: \[External Knowledge API]\(./07-external-knowledge-api.md) · Next: \[Metadata]\(./09-metadata.md)*
