Knowledge Base Getting Started

A knowledge base organizes files, notes, directories, or web pages into a reusable, searchable collection of materials. First use recall tests to confirm the system can find the correct passages, then hand the knowledge base over for use in chat or agents.

If you only need to handle a small piece of text temporarily, pasting it directly into the chat is faster. When the materials will be reused repeatedly and the answer must be based on internal sources, then create a knowledge base.

When is it suitable to use

Requirement	Recommended approach	Reason
Look up employee policies, product manuals, project materials	Create a knowledge base	The materials will be reused, and you need to reliably quote the original text
Temporarily analyze an attachment	Upload it directly in the chat	No need for long-term maintenance or indexing
The materials are still being organized	First sort it out with [Notes]	Avoid unverified content being treated as a formal answer
Need to automatically process materials for the long term	Bind the agent after creating the knowledge base	The agent can continuously use the same material scope during tasks

What happens during one answer

The materials are first parsed and split into chunks, then candidate passages are found using keywords or semantic matching; the chat model is only responsible for organizing the answer based on the recalled content.

First, learn these terms

Name	Meaning in this task
Knowledge base	A set of materials and retrieval settings organized around the same topic
Material item	An imported file, a file in a note or directory, or a webpage snapshot
Chunk	A small passage used for retrieval after materials are split
Recall	The process of finding relevant passages based on the question
Embedding model	Converts text into vectors to match content with different wording but similar meaning; optional
Reranking model	Scores and sorts candidate passages again; also optional

You can still use a knowledge base without an embedding model, in which case it mainly relies on BM25 keyword retrieval. For your first try, you can choose [Do not use] first and get the creation, import, and recall flow working.

Complete your first use in 5 minutes

1

1. Create a knowledge base with a clear scope

Open [Knowledge Base] in the left navigation bar → click the add button above the knowledge base list. Use “object + purpose” for the name, for example [Employee Travel Policy].

2

2. Choose a retrieval method

For your first try, you can set [Embedding Model] to [Do not use]. When you need to match conversational questions or synonyms, then configure an embedding model.

3

3. Add materials

After entering the knowledge base, click the Add Materials button and choose [File], [Note], [Directory], or [Link].

Choose the entry point according to the source of the materials; do not import unrelated directories just to reduce the number of steps.

4

4. Wait for the materials to become ready

After processing is complete, the materials will appear in the list. Randomly check the text and chunks to confirm there is no garbled text, missing pages, or obvious ordering errors.

Even after the materials are ready, you should still spot-check the content; import completion does not mean the retrieval quality is already qualified.

5

5. Complete recall testing

Open [Recall Test], enter a real question whose answer you already know, and check whether the correct source appears in the first few results.

6

6. Enter chat or bind an agent

After recall is stable, select the knowledge base in the chat input area; when you need a long-term workflow, bind the knowledge base on the agent editing page.

Recommended starting point

Configuration item	Product default	Recommended starting point	Function	Applicable scenario	Notes
Knowledge base scope	—	A clear topic	Controls which materials participate in retrieval together	Policies, product materials, project materials	Content with different permissions or lifecycles should be separated
Embedding model	Do not use	Do not use for now	Determines whether vector retrieval is included	There is a large difference between conversational wording and the original text	Cloud model billing and data handling depend on the service provider
Test questions	—	3–5 real questions	Establish a long-term regression baseline	After each update to the materials or settings	Do not test only with the material title and original sentences
Enter formal use	—	Materials, chunks, and recall all pass	Avoid bringing parsing or retrieval problems into the chat	All knowledge bases	The chat model cannot recover key information that was not recalled

How to tell it is ready to use

• All required materials are shown as available, and none remain stuck in processing or error status for a long time.
• Randomly open one or two materials; the text and chunks have no garbled text, missing pages, or obvious ordering errors.
• Use fixed questions for recall testing; the correct source should consistently appear in the first few results.

User case

Xiaolin wants colleagues to look up travel policies. He creates an [Employee Travel Policy] knowledge base, imports three materials on lodging, transportation, and approval, and does not configure an embedding model at first. After the three materials are ready, he tests with questions such as “What is the Beijing lodging limit?” and “Who approves amounts over 5,000 yuan?”

The completion standard is: for each question, the correct policy source can be found, and the passage contains both the applicable conditions and the conclusion. Only after meeting this standard does he bind the knowledge base to the agent responsible for employee Q&A.

FAQ

What is the difference between a knowledge base and a chat model?

The knowledge base is responsible for finding passages from your materials, while the chat model is responsible for understanding the question and organizing the answer. If the recall stage does not find key information, simply switching the chat model usually will not solve the problem.

Can it be used directly after materials are imported successfully?

You still need to spot-check the text and chunks and complete recall testing. Import success only means the processing flow has finished; it does not mean the results are complete or correctly ranked.

After changing the chunking settings, will old materials change automatically?

No. To make existing materials use the new chunking settings, you need to perform [Reindex] and then retest with the same set of questions.

Continue reading

Create a knowledge base

Define the name boundary and retrieval method.

Add and organize materials

Import files, notes, directories, and web pages.

Check materials and recall

Verify retrieval quality with fixed questions.

Model and retrieval settings

Continue adjusting embedding, reranking, and chunking.