Data sources overview
How knowledge works in FluentChat — the kinds of sources, how the agent uses them, and how they stay isolated per agent.
Data sources are the knowledge your agent answers from. Add your FAQs, upload documents, crawl your website, sync Notion, or connect a product catalog, and the agent grounds its replies in that content instead of guessing.
How the agent uses knowledge
On every message, the agent searches its data sources for the most relevant passages and uses them to answer — quoting your policies, describing your products, citing your docs. This is retrieval-augmented generation: the agent only speaks confidently about things it can find in your knowledge.
Knowledge is per-agent. A source you add to one agent is invisible to every other agent, even in the same workspace. Give each agent exactly the knowledge its job needs.
Kinds of data source
Text & Q&A
Paste FAQs, policies, and question/answer pairs. Always free.
Files
Upload PDF, DOCX, XLSX, and CSV documents.
Websites
Crawl pages from a sitemap or a list of URLs.
Notion
Connect Notion and sync selected pages.
Product catalog
Connect a Shopify or WooCommerce store.
Adding and managing sources
Open an agent's Data sources page to see all its sources with their status. Sources move through pending → syncing → ready, or show failed if something went wrong. A per-agent usage bar shows how much of your knowledge allowance you've used.
Some kinds are re-fetched when you retrain them (websites, Notion, store policies); others store their content and simply re-index it. See Training & sync for limits, retraining, and the difference.
What counts against your plan
- Text and Q&A are always available, on every plan.
- File uploads require a plan with the file-uploads feature.
- Website crawl and Notion require a plan with the web-crawl feature.
- Product catalog requires a plan with the commerce feature.
Each agent also has a knowledge-size allowance and a maximum number of sources. See Plan features.