Knowledge base

To use AI to generate answers, you need to first build a knowledge base with the information you want the answer from.

A knowledge base is nothing but a collection of data sources. Think of it as chapters in a book, the more chapters the more information it has.

A data source can be one of three types:

• URL

• File

• Text

URL

You can add content using a domain or a set of web pages.

When using a domain, you simply enter the domain address or sitemap address, and our system will automatically import all URL(s) of that domain.

Supported types:

• Domain i.e. https://wotnot.io

• Sitemap i.e. https://wotnot.io/sitemap.xml

As good practice, it is always better to delete the URLs that are not needed to avoid wrong or incorrect answers from the bot. Because, the more junk you add, the more confused the AI gets.

When using webpages, you explicitly list down the exact URL(s) you want to import content from.

Advanced crawling criterias

Crawling of a website happens real quick, if you find no content being crawled try adding the following configurations:

• wait time - delay in crawling until your page loads.

• wait CSS - delay in crawling until a CSS selector is loaded.

• Execute Javascripts - to prevent any unwanted popups to be crawled instead of the actual webpage content.

Refreshing domain

If you have trained your knowledge base with a domain, you can choose or change the refresh frequency by clicking on the 'Domain(s)' button.

Changing the frequency will apply to all child URL(s) which are fetched from this domain. However, changing child frequency manually, will not impact the parent domain.

File

You can add content by uploading a set of documents to your knowledge base.

Upload one or files and click 'Train' to start training your knowledge base with the contents of those files.

Supported file types:

• csv

• xlsx

• txt

• pdf

• doc

• docx

Text

You can also directly add the information as content to the knowledge base.

By choosing this data source method, you can copy-paste the content you want to add.

Chunks

Clicking on a data source displays all the details of that data source i.e. when was it added, its refresh rate etc.

“Chunks” are a way to handle large pieces of information by breaking them down into smaller, manageable parts.

When are chunks used?

Chunks are used when the system needs to obtain specific information from the knowledge base to optimize the response generated.

When a user asks a question, here's what happens:

1. The system retrieves relevant chunks from the knowledge base that may contain the answer.

2. These chunks, along with prompts, are sent to OpenAI.

3. OpenAI generates a response using the provided chunks.

4. The response is then displayed to the user.