# Knowledge base Your Knowledge Base is what makes your AI agent smart. It's the collection of information your agent draws from when answering customer questions - think of it as the reference library your agent consults before every response. The more relevant and well-organised your Knowledge Base is, the more accurate and helpful your agent will be. Converso has a knowledge base for each workspace. Each knowledge base is split into four tabs, each offering a different way to feed information to your AI agent: **Files**, **URLs**, **FAQs**, and **Response Adjustments**. You can use any combination of these to build a comprehensive knowledge source tailored to your business. ### Files The Files tab lets you upload documents directly into your workspace Knowledge Base. This is ideal for detailed product information, internal guides, policy documents, pricing sheets, or any structured content your agent needs to reference. Supported file types include PDF and Word documents. Simply click **Choose a file**, select your document, and it will be processed and indexed automatically. Once the status shows **Ready**, your agent can immediately start using that content to answer customer queries. Each file can be toggled on or off using the **Active** switch, so you can temporarily disable a document without deleting it - useful when content is being updated or reviewed. You can also replace an existing file with an updated version using the **Update** button, keeping your Knowledge Base current without having to delete and re-upload. ### URLs The URLs tab allows you to extract data and information from a website that can be used as a knowledge source for your AI agent. All you need to do is provide a website address and Converso will crawl all the pages that sit below that URL, and extract the pages' contents, and upload this data to the agent's knowledge base. Note: check that the website does not have blockers in place that may inhibit data scraping. You can also specify individual website URLs from which to extract data. Click **+ Add URLs**, enter the web address, and save. You can add multiple URLs at once using the **+ Add URL** button within the dialog. As with files, each URL shows a **Status** indicator and an **Active** toggle, giving you full control over which web sources your agent is drawing from at any given time. ### FAQs The FAQs tab gives you the most direct control over how your AI agent responds to specific questions. Rather than relying on the agent to interpret information from documents or web pages, FAQs let you define exact question-and-answer pairs. Click **+ Add FAQ** to create a new entry. You'll be prompted to enter a **Question** and its corresponding **Answer**. You can add multiple FAQ pairs in a single session using the **+ Add FAQ** link within the dialog. This is perfect for common customer queries where you want a precise, consistent answer every time - such as pricing questions, return policies, or opening hours.\ \ You can also upload a spreadsheet that contains multiple questions and answers. Just list answers in one column and questions in an adjacent column, upload, and all data will be saved as individual FAQs. Each FAQ entry can be edited, toggled on or off, or deleted independently. When your agent encounters a customer question that closely matches one of your FAQ entries, it will prioritise that predefined answer, giving you confidence that critical information is always communicated accurately. ### Response adjustments Response Adjustments is where you fine-tune your AI agent's behaviour over time. Think of it as a correction and coaching tool - when your agent gives a response that isn't quite right, you can log the original response, provide an improved version, and explain why the change was needed. To create an adjustment, just **right click on the AI agent's response** in the playground or in the chat panel of the workspace inbox or My inbox. \ \ You'll then see three fields: * **Original Response** - the answer your agent gave that you'd like to improve. * **Adjusted Response** - the corrected or improved version you'd prefer your agent to use. * **Reason for Correction** - a brief note explaining why the adjustment was made, which helps the system learn the intent behind the change. Each adjustment can be reviewed at > Knowledge > Response adjustments. \ \ Over time, building up a library of response adjustments helps your agent continuously improve, learning from real conversations to deliver better answers. This is especially valuable in the early days of setting up your agent, when you're actively shaping its tone, accuracy, and style. ### Tips for building an effective Knowledge Base Your AI agent is only as good as the information you give it. A few practical pointers to get the most from your Knowledge Base: * **Start with your most common customer questions.** Add these as FAQs first for immediate, reliable coverage, then layer in documents and URLs for broader context. * **Keep content current.** Outdated information leads to inaccurate answers. Use the update and active toggle features to manage content as your business evolves. * **Use Response Adjustments early and often.** The first few weeks of running your agent will surface the biggest opportunities for improvement - log adjustments as you spot them. * **Combine sources for depth.** A detailed product PDF, your website's support page, and a handful of targeted FAQs working together will give your agent a far richer understanding than any single source alone. --- # Agent Instructions: Querying This Documentation If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question. Perform an HTTP GET request on the current page URL with the `ask` query parameter: ``` GET https://converso.gitbook.io/product-documentation/ai-agents/knowledge-base.md?ask= ``` The question should be specific, self-contained, and written in natural language. The response will contain a direct answer to the question and relevant excerpts and sources from the documentation. Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.