Knowledge and training
Choose the right Fetchply source type, understand indexing status, and keep an agent's answers current.
Open an agent and select Training Data. Knowledge is isolated per agent and is available to the widget, API, and connected channels that use that agent.
Websites
Crawl a site or add individual page URLs.
Files
Extract text from PDF, DOCX, PPTX, spreadsheets, HTML, JSON, and more.
Notion & Google Docs
Import shared Notion pages and Google Docs by link.
Text
Add a titled block of approved text.
Instant Answers (Q&A Pairs)
Give common questions a specific answer.
Review everything in one place
The All tab shows a single table of everything the agent is trained on: website pages, files, text, Instant Answers, media, and synced store products and collections. Search by name, filter by type, and page through large lists. From the same table you can preview a source's extracted text, open the original page, or delete a source you no longer want the agent to use.
Each source-type tab offers the same table with actions specific to that type, such as editing a text snippet or retrying a failed page.
Choose a source type
- Use a website for broad public content that has useful page text.
- Use files for policies, manuals, or documents that are not public web pages.
- Use text for a short fact, announcement, or policy you can maintain directly.
- Use Instant Answers (Q&A Pairs) for a precise response to a recurring question.
Only add content you are authorized to use. Remove secrets, private customer data, and obsolete versions before upload.
Learn how training works, then use retraining when source content changes.