Overview
Knowledge sources
What knowledge sources are in FlowHunt AI Studio and how agents use them to answer from your own content.
Knowledge sources are the content your agents can look things up in: files you upload, pages FlowHunt crawls from your website, and question and answer pairs you write yourself. The pages live in the Knowledge sources section of the AI Studio sidebar.
An agent does not read all of this on every message. It reaches it through a retriever step. The Document Retriever searches your indexed documents, crawled pages and Q&A entries for the passages closest to the user’s question and hands them to the agent as context.
How content gets to an agent#
- You add content: upload a document, create a crawl schedule, or write a Q&A entry.
- FlowHunt processes it and indexes it. Documents and Q&A entries show an Indexed status when this is done.
- A retriever step in your agent or workflow searches the index at run time.
- The retriever can be limited to certain categories or to certain crawl schedules, so one agent answers from support articles and another from product files.
Indexing uses credits
Crawling a website, processing a document and indexing a Q&A entry all consume credits. Each plan also caps how many documents, Q&A entries, schedules and chatbots a workspace can hold. See Plans and limits and Credits.
Pick the right source#
Using a source in an agent#
Add a retriever step to your workflow, or give an agent the retriever as a tool, then pick the categories or schedules it should search. Other retrievers, such as URL Content Retriever, read a page at run time instead of using the index. The full list is in the step reference under Retrievers.