How to create a knowledge base: basic setup

Basic Settings

General Information

You can define the knowledge base name and add a description on the following page:

Retrieval Fragments

The number of retrieval fragments indicates the maximum number of data fragments the AI will reference when answering. The system default is "12", meaning the AI will retrieve 12 of the most relevant fragments per answer.

Therefore, you can increase or decrease the number of retrieved fragments to adjust how much information the AI assistant references when answering.

What is a Parser?

A parser allows the system to "understand" the content of PDF documents so they can be searched, edited, or converted into other formats.

If the uploaded file format is PDF or Word, the parser provides the following three options:

MaiAgent Parser (default)
MaiAgent Parser (Online)
MaiAgent Parser (OCR beta)

The system currently defaults to MaiAgent Parser, which is especially suitable for extracting structured data from complex professional documents for AI analysis.

To understand the differences among the three parsers, please refer to:Technician Manual - Parser Tools

If you encounter issues parsing data, you can click the [Reparse] icon to have the parser refresh the data.

Retrieval Model Settings

In the knowledge base settings, you can choose the Embedding model and Reranker model you want to use.

Embedding Model

Embedding is like translating human language into a "numeric language" the AI can understand, allowing the computer to grasp the true meaning of text; this process is called "vectorization." Different embedding models have different characteristics, such as languages they handle well and supported deployment environments. Different model settings in the knowledge base can be used to adjust how files are vectorized when uploaded, so you can choose the most suitable embedding model for different scenarios.

You can freely choose from multiple embedding models:

To learn about the differences between embedding models, please refer to:Technician Manual — Embedding Models

Reranker Model

A reranker is like a professional judge who re-evaluates which data best answers the customer's question among the initial search results. What difference does using a reranker make?

When a customer asks: "What tent is suitable for beginners? Budget under 8000"

No Reranker:

The AI might answer:
"We have tents at various price points; products around 8000 include..."
(May mention advanced models and not target beginner needs)

With Reranker:

The AI answers:
"For beginners, we especially recommend these tents under 8000..."
(Precisely targets beginner + budget + product recommendations)

When search result reranking is enabled, the AI assistant will reorder retrieved knowledge base content fragments and respond based on the most relevant documents.

To learn about the Reranker model, please refer to:Technician Manual — Reranker Models

In summary, using embedding combined with a reranker allows the AI assistant to understand the knowledge you provide, then reassess the importance of content after retrieval and respond using the knowledge most relevant to the question.

Associate AI Assistants

Associating AI assistants means specifying which AI assistants are authorized to use this knowledge base. If you have two AI assistants:

Product Customer Service AI
Order Customer Service AI

When both need to answer questions related to returns, you can in the "Return Policy" knowledge base settings associate both AI assistants at once:

Select AI assistants to associate

Tap Add AI Assistant

After adding, it will appear in the selected AI assistants area. Tap the bottom-right "Save" to complete the association.

Once associated, both AI assistants can share the "Return Policy" knowledge base and respond based on the same content. Future maintenance only requiresupdating one knowledge baseto ensure the AI assistants use the latest information.