Skip to main content
An AI agent with no training data has nothing to answer from beyond its System Prompt. Knowledge Sources are how you give it real business content — FAQs, product information, internal docs — so its replies stay accurate to your business instead of sounding generic. This is configured directly on the agent, as part of creating or editing it.
Knowledge Sources panel on the Create AI Agent form, with a Knowledge Set dropdown (Select Existing or Create New) and six source type buttons — Content, URL, File, Media, Google Sheet, and API — plus an Attached Knowledge list showing No sources yet
1

Pick or create a Knowledge Set

Under Knowledge Sources on the agent’s form, open Knowledge Set and either select an existing one or create a new one. A Knowledge Set is the container multiple pieces of content get added into.
2

Add content from one or more types

Click a type button to add content into the Knowledge Set: Content, URL, File, Media, Google Sheet, or API. Unlike an agent’s System Prompt, a Knowledge Set can hold multiple sources together — the agent uses all of them to answer.
3

Check Attached Knowledge

Everything you’ve added shows up in the Attached Knowledge list on the same panel, with a count next to the section title.

​
Content

Content panel with a single text box: Write the content your agent should learn from...
The simplest source — just type or paste text the agent should learn from, then click Save. Example:
With this saved, a customer asking “Are you open on Sunday?” or “What’s your return policy?” gets an answer pulled straight from this text — no need to write those replies yourself.

​
URL

URL panel with a Campaign URL field, Fetch Content Configuration selectors, and Remove Content Configuration selectors, plus Generate Raw Response and Generate FAQ buttons
Paste in a page’s URL — a pricing page, an FAQ page, an about page — and ChatSyncs reads its content for the agent to learn from. Example: paste https://mystore.com/faq and the agent can now answer questions straight from that page, without you copying anything into ChatSyncs by hand. The two optional sections below the URL field let you fine-tune exactly what gets read:
  • Fetch Content Configuration — only pull content from one specific part of the page, instead of the whole thing. Pick a Selector Type (ID or Class) and enter the Selector Name — the exact ID or class name from the page’s HTML for the section you want.
    Selector Type dropdown open showing ID and Class options
  • Remove Content Configuration — the opposite: exclude specific sections (like a header, footer, or navigation menu) that would otherwise clutter the training data. Click + Add Remove Selector to exclude more than one section.
    Remove Content Configuration with two Selector Type / Selector Name rows and an Add Remove Selector button
Once the page is fetched, choose how it’s stored:
  • Generate Raw Response — keeps it as one detailed block of content. More thorough, but uses more tokens per reply.
  • Generate FAQ — automatically breaks the page into short Question/Answer pairs instead. More efficient, since the agent only pulls in the exact pair it needs.

​
File

File panel with an Upload button, and accepted formats listed: PDF, Word (.doc, .docx), Excel (.xls, .xlsx), CSV, OpenDocument (.ods, .odt), TXT
Upload training manuals, product guides, or FAQs as PDF, Word (.doc/.docx), Excel (.xls/.xlsx), CSV, OpenDocument (.ods/.odt), or TXT. The same Generate Raw Response / Generate FAQ choice from URL applies here too, once the file is uploaded.

​
Media

Media panel with an Image URL field, an Upload button, a Description field with placeholder text: e.g. Red cotton kurti, size M, price 1200 BDT, and an Add more image button
Add product photos or gallery images the agent can send when a customer asks for them. For each image, paste an Image URL (or click Upload) and write a short Description — this is what tells the agent when to send that photo, so be specific: name, and anything that distinguishes it (color, size, price). Example:
With this saved, a customer asking “Do you have this in red?” or “Show me the kurti” gets that exact photo sent back automatically. Click + Add more image to attach more than one.
Don’t put image URLs directly in the System Prompt. Add them here, under Media, instead — the agent only sends a photo from Knowledge Sources when a customer’s question matches its description.

​
Google Sheet

Select a Sheet dropdown open, showing a connected Google account and its available sheets
Pick a sheet already connected — see How to Connect Google to ChatSyncs — the dropdown lists every sheet available to that account.
Google Sheet panel with a sheet selected under Import From Google Sheet, and a Save button
Example: picking a Customer Details sheet lets the agent answer questions using whatever that sheet contains — e.g. order history or account info — without you re-typing it anywhere.

​
API

Add HTTP API panel with an HTTP API List dropdown
Pick one of your configured HTTP APIs — content is pulled dynamically from that endpoint’s response rather than stored statically.

​
Google account permissions this needs

If you plan to use Google Sheet as a training source, or want subscriber contacts synced alongside it, make sure the Google account connected via How to Connect Google to ChatSyncs granted both the Sheets and Contacts permissions during sign-in — granting only Sheets access leaves contact-dependent features without data.

​
Use cases

  • Customer support — answer common questions the way a trained support agent would, in the business’s own voice.
  • Product information — ground answers in the business’s actual catalog or specs instead of generic AI guesses, with real product photos via Media.
  • Internal documentation — let staff-facing agents answer from internal wikis or SOPs.
  • Business FAQs — turn an existing FAQ page or document directly into training content.

​
Frequently asked

Open the agent under Chatbot Manager → AI → Agents, go to its Knowledge Sources, pick or create a Knowledge Set, and add content — Content, URL, File, Media, Google Sheet, or API.
Yes — unlike the old single-source-per-campaign model, a Knowledge Set can hold multiple sources together, and the agent uses all of them to answer.
Add it under Media in Knowledge Sources — not as a URL pasted into the System Prompt. The agent can then send it when a customer asks for photos.
Check that its Knowledge Set actually has content added — an empty Knowledge Set (or an agent with none selected) has nothing to answer from beyond its System Prompt.
  • An agent with no Knowledge Sources attached has nothing to answer from beyond its System Prompt.
  • Put images under Media, never as a URL in the System Prompt.

​
Learn From a Use Case

  • Travel Assistant — stores the business’s company info, destinations, pricing, and FAQs the AI Agent answers from.