> For the complete documentation index, see [llms.txt](https://shoppad.gitbook.io/yedric/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://shoppad.gitbook.io/yedric/getting-started/knowledge-base.md).
# Knowledge Base
The Knowledge Base is where you load the content Yedric uses to answer your users' questions. Think of it as what Yedric studies before talking to anyone. The more relevant content you add, the more accurate and specific his answers will be.
Without anything in the knowledge base, Yedric works from his system prompt alone. That is fine for general guidance, but if you want him to answer questions about your specific product, your documentation, or your processes, this is where that content lives.
***
## What to add
Good candidates for the knowledge base include:
* Help articles and FAQs
* Product documentation
* Setup guides
* Common troubleshooting steps
* Anything your support team answers on a regular basis
If your team answers the same questions over and over, those answers belong here.
***
## How to add content
There are a few ways to add content to the knowledge base.
Add links, docs, or write your own content to complete your assistant's knowledge base.
**Add a website** Enter a domain and Yedric will crawl it automatically, pulling in pages up to the depth and URL limits you set. Useful for importing an entire help center or documentation site in one step. To pull in only part of a site, use the optional **Only add pages under this path** field to restrict the crawl to a URL prefix (for example, `/docs/`) so pages outside that section are skipped. Website documents are re-scraped automatically once a week to keep the content current.
The crawl copes with most real sites without any setup on your part. Sites that build their pages with JavaScript work: if the plain crawl finds little text, Yedric retries with the pages rendered in a browser and keeps whichever pass extracted more. Addresses that redirect (an `http://` address that forwards to `https://`, for example) are followed. Sites without a usable sitemap are crawled by following links instead. If an import still finishes with no text, the document shows an **Error** status with a message saying what is likely wrong, such as a site that blocks automated crawlers or requires a login.
**Upload files** Upload files directly from your computer. Supported formats include PDF, Word documents (.docx), and plain text files (.txt). You can select several files at once. Each file is uploaded on its own, so one file that cannot be read does not stop the others; its name and the reason it failed stay on screen next to the drop zone while the rest are added.
**Write in yourself** Write or paste content directly into a text editor. Useful for FAQs, internal notes, or anything that does not have a URL or a file.
You can see all of your content filtered by type using the tabs above the content list: All, Web, File, and Text.
***
## Document details
Click on any document in the list to open its detail view. This shows the source URL or file name, when it was last indexed, how many sections it was split into for searching, and its current status. For website and URL documents, you can trigger a manual rescrape from here if you need to pull in updates right away.
Documents go through a short processing period after you add them. The status shows **Adding...** while they are being indexed and **Ready** once they are available for the assistant to use. If something goes wrong during indexing, the document will show an **Error** status with more detail. If this happens, try removing the document and adding it again.
***
## Prompt versus knowledge base
These two things work together but serve different purposes. The system prompt is for instructions and tells Yedric how to behave and communicate. The knowledge base is for reference content and gives Yedric the information he needs to answer specific questions accurately. If you find yourself writing long explanations about your product inside the system prompt, that content probably belongs in the knowledge base instead.
***
## Keep your content current
Yedric answers as well as the content you give him. When your product changes, your pricing updates, or you add new features, come back and update the knowledge base to match. Keeping your content current means Yedric keeps giving accurate, helpful answers, and your users get what they need without friction.
Website documents are re-scraped weekly automatically. For everything else, you will need to update or re-upload the content manually.
***
## Storage
Your knowledge base has a storage limit shown in the top right corner of the page. Keep an eye on this as you add content. If you are approaching the limit, review your documents and remove anything outdated or no longer relevant before adding new content.
Track your knowledge base storage capacity.
***
## How it works
Once content is loaded, Yedric searches the knowledge base automatically every time a user asks a question. You do not need to configure anything for this to happen. It is built in.
If users are asking questions Yedric cannot answer well, that is usually a sign the knowledge base needs more content on that topic. Use the Conversations tab to see what users are asking and come back here to fill the gaps.
