Add a Knowledgebase

What is a Knowledgebase?

A knowledgebase is a collection of information that your AI voice agent can reference during conversations. This allows your agent to provide accurate, up-to-date answers about your products, services, policies, and more.

Creating a Knowledgebase

To create a new knowledgebase:

1. Navigate to the Knowledgebase page in your dashboard.
2. Click the "Create Knowledgebase" button at the top right of the page.
3. Enter a name for your knowledgebase.
4. Choose one of the three content options below.

Content Options

You can add content to your knowledgebase in three ways:

Upload File

Upload a document and urvo will automatically extract all the text content from it. Supported file formats: PDF, DOCX, EPUB, TXT, HTML, HTM, MD (Markdown) — up to 20 MB per file.

Add URL

Provide a website URL and urvo will scrape and extract the content from that page. Note that the website must allow web scrapers in its robots.txt file. If a website blocks scrapers, the content cannot be extracted.

Create Text

Manually type or paste text content directly. This is useful for creating custom FAQs, policies, or any information you want your agent to know.

Managing Your Knowledgebases

After creating a knowledgebase, it will appear on the Knowledgebase page. You can click on any knowledgebase to view its details.

Knowledgebase Details

When viewing a knowledgebase, you can see:

• Name: The name you gave to the knowledgebase
• Type: How the content was added (file, URL, or text)
• Created By: The team member who created it
• Created At: When the knowledgebase was created
• Last Updated: When the knowledgebase was last modified
• Dependant Agents: Which AI agents are using this knowledgebase
• Content: The extracted text content

Large Knowledgebases

A single knowledgebase larger than 275 KB is too big to hand to the agent in full. Above that size it can only be used through RAG, described below. Below it, the whole document is available to the agent directly.

RAG (Retrieval-Augmented Generation)

Without RAG, every knowledgebase attached to an agent is loaded in full for each conversation. That is fine for short documents and stops working for long ones. RAG indexes a knowledgebase once, then retrieves only the passages relevant to what the caller actually asked, so a large manual costs about the same per call as a small one.

Enabling RAG

RAG is a per-agent switch, not a per-knowledgebase one. Open the agent, go to Configuration, and turn on RAG in the Knowledgebase section. Every large knowledgebase attached to that agent is then served through retrieval.

Indexing

Turning RAG on starts an indexing job. The dashboard polls and shows progress while it runs. Indexing is not instant, and it has to finish before retrieval works — if you test the agent mid-index, it will behave as though the content is not there yet. Re-uploading or editing a knowledgebase re-indexes it.

Size limits and quota

RAG availability itself is a plan feature. If your plan does not include it, the toggle is unavailable and large knowledgebases cannot be used — see Manage Subscription.

Connecting Knowledgebase to Agents

Once you've created a knowledgebase, you can attach it to your AI voice agents:

1. Go to the Agents page and select the agent you want to configure.
2. In the agent settings, find the Knowledgebase section.
3. Select the knowledgebase(s) you want the agent to use.
4. Save your changes.

Your agent will now be able to reference the information in the attached knowledgebase when answering questions.

Best Practices

• Keep content focused: Create separate knowledgebases for different topics (e.g., products, policies, FAQs)
• Update regularly: Keep your knowledgebases current with the latest information
• Use clear formatting: Well-structured content helps the AI understand and retrieve information better
• Test your agents: After adding a knowledgebase, test your agent to ensure it responds correctly

Next Steps

With your knowledgebase created, you can:

1. Create or update your AI agent to use the knowledgebase
2. Analyze calls to see how your agent uses the knowledge
3. Review best practices for optimal agent performance