# Knowledge Base

> Build and manage your AI Agent's Knowledge Base. Learn how to add content, structure information, and optimize for accurate AI responses.

The **Knowledge Base** is the core component of your Quickchat AI Agent. It serves as the foundation for your AI Agent’s understanding of your company, products, and services. A well-structured and comprehensive Knowledge Base ensures your AI Agent can effectively handle user queries and deliver accurate responses.

This guide provides an overview of the Knowledge Base, strategies for building it effectively, and tips for optimizing its performance.

## Data sources

The toolbar at the top of the Knowledge Base page lets you add knowledge from any of these sources:

### Add Article

Manually-written articles. Title + rich-text body. Use this when you want full control over the exact wording — FAQs, internal policies, product specs, or anything you'd otherwise paste from a doc.

:::tip
Keep each article focused on a single topic. Long, multi-topic articles confuse retrieval — splitting them gives the AI cleaner answers.
:::

### Add Paragraph

Short, atomic facts that don't need a full article. Paragraphs are grouped together (e.g., under a "Single Paragraphs" folder) and treated as bite-sized knowledge units. Good for one-liners like _"Shipping takes 3–5 business days"_ or _"Returns are accepted within 30 days"_.

### Add Website

Import content from any URL. Four import modes:

- **Single Page** — imports one webpage. Use **Advanced options → HTML selector** to restrict the import to a specific part of the page (e.g., `main`, `#article-body`, `.docs-content`) when the rest of the page is navigation, ads, or footer.
- **Multiple Pages** — paste a list of URLs; each becomes its own article.
- **Sitemap** — point at a `sitemap.xml`; every page in the sitemap is imported as a separate article. Best for documentation sites and structured blogs.
- **All Pages** — site-wide crawl of every reachable URL on the domain. Use for full documentation imports or company blogs.

URLs stay linked to their source. You can re-pull the latest content for one article from its row menu, or use **Refresh all URLs** in the top right to re-pull every URL-backed article at once.

### Upload File

Upload PDFs or Word documents. Text is extracted automatically and indexed. Max file size: **32 MB**.

### Shopify MCP

Connect your Shopify store so the AI can answer questions grounded in your live product catalog, orders, and inventory. Paste your store URL to set up the connection. Behind the scenes this configures a **Shopify MCP** Custom Action — you can also manage it from **Actions & MCPs**. See [Custom Actions](/ai-agent/actions/#custom-actions).

### Intercom Articles

Bulk-import every article from your connected Intercom Help Center. Requires Intercom to be connected under **External Apps → Intercom** first. Re-running the import syncs new and updated articles.

### YouTube

Paste a YouTube URL to import the video's transcript as an article. Useful for tutorial channels and product walkthroughs you don't want to re-write by hand.

## **Tips for building an effective Knowledge Base**

Each uploaded text article, URL, or PDF should be self-contained and comprehensive, covering the topic thoroughly. A well-structured Knowledge Base ensures the AI Agent has the necessary context to respond effectively.

Read the below tips to set your AI Agent for success:

## Steps to build and refine your Knowledge Base

1. **Get clear on the purpose**  
   Define the role of your AI Agent and the types of questions it will answer. Organize your knowledge into categories based on user needs and flows.

   **E-Commerce AI Agent Example:**

   - **Browsing & Shopping**: Product details, comparisons, and FAQs.
   - **Buying**: Payment methods, discount codes, and taxes.
   - **Delivery**: Shipping, order modifications, and tracking.

   **Student Loan AI Agent Example:**

   - **Before Applying**: Eligibility and loan conditions.
   - **Applying**: Guides, terms, and walkthroughs.
   - **After Graduation**: Repayment schedules and deferrals.

2. **Ensure data quality**  
   The quality of AI responses depends on the data you upload. Focus on:

   - **Readability**: Remove irrelevant information, tables, or redundant headers from uploaded content.
   - **Relevance**: Ensure the content answers common questions and aligns with user expectations.
   - **Consistency**: Avoid contradictory or outdated information, and ensure consistent formatting (e.g., dates, prices).
   - **Text-only content**: Since images are not supported, ensure vital information is conveyed through text.

3. **Prioritize clarity**  
   Ensure each article is clear, concise, and easy to follow. Use logical paragraph structures and focus on relevant details.

4. **Be exhaustive**  
   Provide as much detail as possible to equip your AI Agent with the knowledge needed to handle complex queries.

5. **Use answers instead of Q&A**  
   Focus on providing comprehensive information rather than specific question-and-answer pairs. The AI Agent can extract relevant details as needed.

6. **Embed URLs within text**  
   Always include URLs within the text for reference. Even when uploading via the URL uploader, linking to products or forms within the text improves usability.

   **Example**:  
   For an e-commerce AI Agent, include links to forms like order cancellation:  
   _"To cancel an order, submit a request here: https://example.com/cancel-order."_

7. **Test thoroughly**  
   Testing is essential to refine your Knowledge Base. Simulate user interactions to identify gaps or inconsistencies and adjust the content accordingly.

## **Common issues**

- **The AI Agent skips links or doesn’t include URLs in responses**
  Ensure relevant links are embedded within the text of each article.
- **Repeated or redundant information**
  Avoid uploading duplicate content across multiple Articles.
- **Missing key details in responses**
  Consider creating separate Articles for specific details, like criteria or special cases.
- **Incorrect dates or prices**
  Verify and update all Articles to maintain accuracy.

Building an effective Knowledge Base is an iterative process. Start with clear objectives, organize content logically, and test extensively to refine your AI Agent’s performance.

:::note
If you need assistance setting up your Knowledge Base or require custom integrations, feel free to reach out to us at **contact@quickchat.ai**.
:::

## **Search**

The **Search** feature allows you to quickly find specific articles in your Knowledge Base by searching through **both their titles and content**.

**How to use search**

1. Enter keywords or phrases into the **Search** bar at the top of the Knowledge Management interface.
2. The results will instantly update to display articles containing the entered text in their title or content.

## Multi-select and bulk delete

The multi-select and bulk delete functionality makes it easy to manage multiple articles in your Knowledge Base.

**How to use multi-select**

1. Navigate to the Knowledge Base list
2. Click the “…” sign next to the Article you want to select and click “Select”
3. Check the boxes on the left next to the Articles you want to select. You can select one or multiple articles simultaneously
4. Once selected, the Articles will remain highlighted for further actions

**How to perform bulk delete**

1. After selecting the articles using the multi-select feature, click the **Delete** option at the bottom to remove the selected articles
2. Confirm your action to delete all selected articles permanently

## Tags

Tags help you organize the Knowledge Base and let the AI Agent narrow down which articles to consult for a given question. You'll see tags on every article row and as a filter at the top of the list.

There are two ways to tag an article:

- **Manually** — open an article and use **+ Add Tag** in the right-hand metadata panel.
- **Auto-tag with AI** — click **Auto-tag with AI** in the same panel. Quickchat reads the article and suggests tags based on its content. Useful for bulk-imported KB (Sitemap, All Pages, Intercom) where manually tagging hundreds of articles isn't realistic.

Use the **Tags** filter in the toolbar to narrow the list to articles matching one or more tags.

## Retraining your AI Agent

Adding, editing, or deleting articles doesn't update the live AI Agent immediately. After meaningful changes you'll see an orange banner at the top of the app:

> **AI Knowledge is outdated. Retrain required.**

Click **Retrain AI** in the banner to push the latest Knowledge Base to your live Agent. Retraining typically takes under a minute.

:::caution
Retraining affects the **live** AI Agent — including any deployed widget, channel integration, or API consumer. Always test changes via **AI Preview** or a [simulation dataset](/conversations/testing/) before retraining if your Agent is in production.
:::

You only need to retrain after **Knowledge Base** or **Identity** changes. Inbox replies, Insights, and Conversation Rating settings take effect immediately.