PDFs & Knowledge

Understanding how Sideconvo works is the first step to answering this question.

Sideconvo indexes all your website URLs and any additional knowledge you upload and stores it in a vector database. Additionally, your site description and special instructions are used as foundational instructions. When you have a conversation with Sideconvo, the answer is constructed exclusively from this knowledge, your site description and any special instructions you provide.

Experiment in the Playground with conversations and observe the related links that are presented with the Sideconvo response. If you do not find what you expect:

• Go to Sitemap & URLs and confirm your content is in Active state. You can use the search filter to find the page by URL or page title.
• If it is found, click on it and review the content preview. Did it capture what you published on the page? Does the page have what you were looking for? If there was an error crawling the page or if you made an update to your website page, you can always manually recrawl the page to ingest the information again.
• If it is not found, add the URL manually by tapping on the New URL button on Sitemap & URLs and Sideconvo will visit the page and ingest its content.
• Responses will immediately be reflected after the page is crawled. Confirm the result is to your liking by trying the conversation again in the Playground in a new session.

If the above does not work for you, the problem could be that other content is ranking higher in the results. If you think this is the case, consider the content on that page:

• Is the content secondary or not important to the experience you want to provide? If yes, you can exclude it from Sitemap & URLs.
• If your site is a store, consider excluding category and tag pages if your product pages are individually indexed.
• Is the content on that page ‘richer’ or better structured than the content on the page you expect? Consider improving your website pages by better structuring your content across the two pages.
• Responses will immediately be reflected after the page is crawled. Confirm the result is to your liking by trying the conversation again in the Playground in a new session.

Finally, the process for Knowledge & PDFs is mostly the same.

You can delete a file by hovering over it in the list and clicking the Delete icon (trash can) or by opening its Details View and clicking the Delete icon in the header.

Excluding a file removes it from the assistant’s active knowledge base, so its content won’t be used in answers. The file is not deleted and can be “Re-included” later.

• Rename: This changes the “friendly title” of the file to make it easier to find in your list. It doesn’t change the file’s content.
• Edit: This allows you to manually change the extracted markdown content that the assistant uses. This is for fine-tuning or correcting extraction errors.
• Update: This replaces the entire file with a new version (e.g., replacing v1 of a document with v2).

Simply click the file name in the list. A panel will open on the right showing a preview of the file’s content as markdown. This extracted markdown is the exact information the assistant uses.

• Error: The system could not process your file. The error message (e.g., “file is password protected”) will tell you what to fix before you re-upload it.
• Warning: The system found a potential issue (e.g., “might contain harmful content”), but the file may still be usable.

• Show All Active: Displays all files that are currently being used by your assistant as a knowledge source.
• Show All Files: Displays every file you have ever uploaded, including those that are “Active” and those you have “Excluded.”

You can either click the + Add File button to select files from your computer, or you can drag and drop files directly onto the file list area at any time.

The system accepts common file types, including PDF, PPTX, DOCX, CSV, and TXT.

This screen is your central repository for non-website content. You can upload files like PDFs, presentations, and text files to give your assistant a deeper knowledge base beyond your public website pages.