# Knowledge

## **1 Introduction**

Knowledge is the foundation of intelligent AI systems, representing the structured and unstructured information that powers your agents and applications. In our platform, knowledge encompasses various types of data sources that can be processed, indexed, and made searchable for AI agents to reference during conversations and task execution.

### What is Knowledge?

Knowledge in our system refers to textual information that can be stored, organized, and retrieved to enhance AI capabilities. This includes:

* **Documents**: Files in various formats (PDF, DOC, DOCX, PPT, PPTX, XLSX, Markdown, images, etc.)
* **Web Content**: Information scraped from URLs and webpages
* **Manual Inputs**: Custom text content entered directly

### Enhanced Domain Mode

For knowledge RAG system, we support Enhanced Domain Mode.

Unlike traditional RAG approaches that rely on basic chunking and keyword matching, our Enhanced Agentic RAG goes much further. It builds a systemic, graph-based connection across multiple layers of knowledge—from the entire document down to individual chunks—enabling your AI agent to truly understand relationships, context, and hierarchy within your data. Even better, these knowledge graphs dynamically evolve based on user interactions and feedback, continuously refining the agent's retrieval strategies and responses. The result? More accurate, context-aware, and ever-improving answers—tailored to your unique domain.

### Why Use Knowledge?

Knowledge serves several critical purposes:

* **Enhanced AI Accuracy**: Provide agents with domain-specific information to generate more accurate and relevant responses
* **Contextual Understanding**: Enable AI to understand your business context, products, services, and procedures
* **Consistent Information**: Ensure all agents have access to the same up-to-date information
* **Customization**: Tailor AI behavior based on your specific organizational knowledge
* **Efficiency**: Reduce the need for manual input by pre-loading relevant information

## **2 Create a Knowledge**

### **2.1 Knowledge Tab**

Click the "**Knowledge**" tab on the left side of the page to access the **Knowledge Center**.

* Here, you will see all your existing knowledges.
* Now you can start creating a new knowledge database by clicking "**+**"!

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FK4dCNmhQDkIOrgFW92Ue%2F1280X1280.PNG?alt=media&#x26;token=5fdad173-3651-48a3-83b1-40ce48c1b86b" alt=""><figcaption></figcaption></figure>

### **2.2 Create and Configure the Settings for the Knowledge**

Start by adding a **Knowledge Database**, selecting its **type**, and giving it a name and **description**.

#### **Select Knowledge Type**

At the top, you can choose between this type of knowledge databases:

* **Knowledge**\
  Unstructured data like files and webpages.

**Enter the Name**

In the **"Name"** field, you need to give your knowledge database a name. This helps identify it later when used by agents or apps.

#### **Provide a Description**

In the **"Description"** field, briefly describe the purpose or content of this knowledge database. This helps others understand its role or scope.

#### **Save or Cancel**

* Click **"Save"** to create and store the database.
* Click **"Cancel"** to discard the process and return to the previous view.

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FGGExqtT7G4iqFYz5wRDq%2F553ea763-2444-431c-8cd4-3c30a1d7246c.png?alt=media&#x26;token=3801c774-3878-44aa-832f-b542dcc886b2" alt=""><figcaption></figcaption></figure>

### **2.3 Prepare Your Data**

Click "Add Files" **and** "now you can add "documents" to the Knowledge Database you just created here!

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2F2tEkyBlJ1z9oSM3p5Pvw%2F4291042f-14e8-4389-9ab8-284a782c20f5.png?alt=media&#x26;token=2f10cb03-030b-4179-8d55-1d0628405931" alt=""><figcaption></figcaption></figure>

We categorize supported documents into **four types**, and you can choose one of the three from the top of the page:

* **Document**\
  You can upload your **local file** here. Currently, we support the following formats: PDF, DOC, DOCX, PPT, PPTX, XLSX, Markdown, PNG, JPG, TXT, MSG, JSON.

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FZpdHr4hJQcVqXF9i7b15%2Fedited_add_files_redlines.png?alt=media&#x26;token=8413111d-efc7-48ec-b798-5c6678631995" alt=""><figcaption></figcaption></figure>

* **Manual Input**\
  You can manually enter the **text content** that will be used as a document.

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2Fwaz1IsIyEQnhzOITMwLG%2F9e53fb11-215a-458c-9726-ea9bdba9d74e.png?alt=media&#x26;token=cd816d64-313b-496e-8f37-023d322afeec" alt=""><figcaption></figcaption></figure>

* **Webpage**\
  You can enter **URL** to fetch data from the webpage.

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FA43KHSbRTrxvpfjE5aP6%2F3e217a2d-4f99-4ba0-81fd-a074a1eeeb79.png?alt=media&#x26;token=64aa648d-b7d7-46fc-9abb-7d3f8c622b17" alt=""><figcaption></figcaption></figure>

* **Datasources**

You can select documents from external sources.

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FUYIa27gTaeXyLAyeNvqz%2Fimage.png?alt=media&#x26;token=4ae915ca-3540-4362-9e33-f46566ac868a" alt=""><figcaption></figcaption></figure>

### **2.4 Check Your Document**

<div data-full-width="false"><figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FekAY5ac8LRkR69PjPvWy%2Fcf91c3ec16b635f746a5ed26ea05e168.png?alt=media&#x26;token=bce23084-958a-4174-92a6-a44681869221" alt=""><figcaption></figcaption></figure></div>

After successfully creating the document (Doc), you can view it on the corresponding Knowledge page along with several key attributes, including:

* **Name**: The title or identifier of the document
* **File Size**: The size of the uploaded file
* **Word Count**: The total number of words in the document
* **Chunks**: The number of segments the document has been split into
* **Status**: Indicates the current processing state (e.g., “Pending”, “Trained”)
* **Enabled**: Whether the document is currently active and available for retrieval

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FZ5QNj87c0DALcGf1nLmx%2Fb8d05cb5cd9ef7a4af6d76d709c45531.png?alt=media&#x26;token=a7ca05a5-9604-4780-8dc2-d70269ad0e59" alt=""><figcaption></figcaption></figure>

By clicking on the corresponding document (Doc), you can view detailed information about its chunk segmentation as well as associated metadata.

You can also manually **enable** or **disable** individual chunks to flexibly control their participation in the knowledge retrieval process.

## **3 Enhanced Domain Mode**

When creating a new knowledge, you have the option to enable Enhanced Domain Mode for advanced retrieval capabilities.

### What is Enhanced Domain Mode?

Enhanced Domain Mode represents our most advanced knowledge processing system. Unlike traditional RAG approaches that rely on basic chunking and keyword matching, our Enhanced Agentic RAG goes much further. It builds a systemic, graph-based connection across multiple layers of knowledge—from the entire document down to individual chunks—enabling your AI agent to truly understand relationships, context, and hierarchy within your data.

### Key Features

* **Graph-Based Connections**: Creates intelligent links between related content across your knowledge base
* **Multi-Layer Understanding**: Analyzes relationships from document level down to individual chunks
* **Dynamic Evolution**: Knowledge graphs continuously evolve based on user interactions and feedback
* **Contextual Hierarchy**: Maintains understanding of information structure and relationships
* **Adaptive Retrieval**: Continuously refines retrieval strategies based on usage patterns

### Benefits

* **More Accurate Responses**: Deeper understanding leads to more precise answers
* **Context-Aware Results**: Maintains awareness of information relationships and context
* **Continuous Improvement**: System learns and improves over time through user interactions
* **Domain Specialization**: Tailored specifically to your unique domain and use cases

### How to Enable Enhanced Domain Mode

When creating a new knowledge:

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2F5ChX1iupLxScdQhOwPkU%2Fenable_enhancing.drawio.png?alt=media&#x26;token=f556465c-f771-40cc-a47d-303898dee4be" alt=""><figcaption></figcaption></figure>

1. During the knowledge creation process, look for the Enhanced Domain Mode option
2. Toggle the Enhanced Domain Mode setting to enable advanced processing
3. Note that Enhanced Domain Mode may require additional processing time but provides superior retrieval capabilities
4. Save your knowledge configuration with Enhanced Domain Mode enabled

<figure><img src="https://4046886348-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FYf4L4Z4IuV1CQn3whQYb%2Fuploads%2FIlPrHyaMzb3OSJHz32ac%2Fkeyword_pre.drawio.png?alt=media&#x26;token=ffb58e45-b4f0-4265-9575-4004d530ccde" alt=""><figcaption></figcaption></figure>

After enabling enhancing mode, you can see the extracted **keywords** for each chunk in the document interface.

The result is a more intelligent, context-aware, and ever-improving knowledge system that delivers answers tailored to your unique domain.