Source Labeling overview

By default, when a user asks your agent a question, the agent searches your entire knowledge base for an answer. Source labeling lets you change that. You can tag your knowledge base pages with labels and then control - at the moment a user asks a question - which labeled pages the agent is allowed to search.

The agent only sees what you point it at. Everything else is ignored for that question.


Why this matters

Most businesses do not have a single, uniform knowledge base. They have content meant for different audiences, regions, languages, or products - all sitting in the same agent. Without source labeling, the agent pulls from all of it, which can produce responses that mix up products, show the wrong language, or surface content that was not meant for that user.

Source labeling gives you precise control over which content answers which questions.


Common scenarios

Multiple languages

You have support documentation in English, Spanish, and French. Rather than building a separate agent for each language, you label pages by language and route each user to the right content automatically.

Multiple products

Your knowledge base covers Product A and Product B. A customer asking about Product A should never get an answer that pulls from Product B documentation. Labels keep them separate.

Multiple departments or roles

HR policies, finance procedures, and IT guides all live in one agent. Labels let you serve each team only the content relevant to them.

Regional content

You operate in different countries with different policies, pricing, or regulations. Labels let you serve region-specific content without building separate agents.


How filtering works

Each time a user asks your agent a question, labels determine which pages the agent is allowed to search. You can apply one or more labels.

When you apply multiple labels, you can choose how they combine:

  • Any match - the agent searches pages that carry at least one of the labels. Useful when content overlaps across categories.
  • All conditions - the agent searches only pages that satisfy every label group. Useful when responses need to meet multiple conditions at once. For example, the right department AND the right language.

You can also choose whether pages with no label are included alongside labeled ones. This is useful when some content applies to everyone and some is audience-specific. Click here to learn how to use labels when querying your agent.


What source labeling is not

  • It does not delete or hide content from your knowledge base. Labels are applied at query time only. Your pages are always there.
  • It does not replace your agent's persona or guardrails. Those still apply regardless of which sources are used.
  • It does not require building multiple agents. One agent, one knowledge base, precise control over what each user sees.

Related articles