How the assistant reads documents, links and images
A visitor can ask “What does error E17 mean?” and share a link to a manual. Depending on the question, the assistant can find the error code, read the instructions around it, or inspect a diagram on the relevant page. You can see the tool catalog in Agent Tools → Built-in.
New chat agents have the reading tools in the table below enabled automatically. Older agents may still have an earlier selection. If you need a tool on an older agent, email support@kanbu.ai with the agent’s name and an example question. Support can check its settings and activate the tool on request.
Which tool helps with your question?
| What the visitor wants to know | How the tool helps |
|---|---|
| “How long is the warranty on this model?” | Document Search finds the relevant passage in knowledge sources the agent can access. It can match a customer’s wording to the terms used in your warranty policy. |
| “Here’s the manual. How do I clean the unit?” | Fetch URL opens a specific page or file and reads the instructions directly from the source. A known link is enough. Web search can stay off for this task. |
| “Which page of this PDF covers error E17?” | Search PDF looks for a code or phrase in the embedded text of a PDF at a web address. It returns the surrounding passage and page number so the visitor can check the instructions. For a PDF already in the knowledge base, the assistant uses Document Search. |
| “Which part is marked B in the diagram on page 12?” | View PDF Page inspects a whole page or a detail within a figure. It helps with diagrams, tables and scans where a text search can miss the answer. It works with an accessible knowledge-base PDF or a permitted PDF link. |
| “Which indicator light is on in this photo?” | See Image inspects a linked photo, chart or diagram. It can pick up details the caption leaves out. |
Link reading respects the agent’s domain restrictions. Searching the embedded text of a linked PDF is available in text chat. Inspecting PDF pages and images also requires a model that supports images. To let the agent discover new links, enable web search separately.
When you have many similar documents
If your knowledge base contains dozens of similarly named manuals, two optional tools can help the assistant choose where to look. You can ask support to enable them even on a new agent.
- Find Documents helps with “Which manuals do we have for the X series?” The assistant uses titles and summaries to choose a document, then searches inside it for a precise answer.
- Document Overview helps with “Which chapter of this manual covers maintenance?” The assistant uses the outline and summary to locate the section. It can then check the passage for a specific detail, such as the filter replacement interval.