Knowledge Base

Tahsim Ahmed
Tahsim Ahmed
  • Updated

Introduction to the Knowledge Base

The Knowledge Base serves as the repository and management system for the content your AI Agent uses to inform its AI-interactions. This critical component allows for the integration of a diverse array of data, enabling the Agent to draw upon a wealth of information to provide accurate, informative, and contextually relevant responses.

Knowledge Base Overview

  • Data Source Status: This section provides real-time updates on the status of your data sources as they go through the stages of uploading, processing, and readiness. It ensures you are aware of when your data becomes active for AI referencing.

  • Preview KB: You can simulate potential user interactions with the 'Preview' button. This functionality allows you to vet the AI's responses in a controlled environment, ensuring they meet your standards and are free of errors.

  • KB Settings: The settings section is the control hub for fine-tuning how your AI interprets and utilizes the data. Here, you can adjust the AI's performance, ensuring that the responses are not only correct but also delivered in a manner that aligns with your desired user experience.

  • Source Chunks: A read-only feature that allows you to view the individual chunks of content generated by Voiceflow which your Agent uses to generate responses. This helps you understand how your document data is segmented and ensures the relevancy and accuracy of the information it draws from.

Adding a Data Source

Click on the demo above to see how to add a source

  1. Access Knowledge Base. Begin by going to the Knowledge Base tab, located within the Content section of your dashboard.

  2. Add Data Source. Click the 'Add Data Source' button to start the upload process. This button is prominently displayed in the header for easy access.

  3. Choose the type of data source you wish to upload. Each type is suited for different scenarios:

    • URL(s): Use this for web pages where the AI can read static text. Ideal for FAQ pages or information repositories. Any dynamically-loaded text, table data, or image content will not be used within your response. Any content behind an authentication wall will not be accessible. 
      • Optional - Select preferred Refresh rate.
    • Plain text: Best for when you have unformatted, straightforward text data you want to include.
    • Sitemap: Useful for giving the AI an understanding of the hierarchy and layout of your content-rich website.
    • Text: For uploading plain .txt files containing information the AI needs to access. File size is limited to 10MB.
    • PDF: When you have documentation in PDF form that needs to be referenced by the AI. File size is limited to 10MB.
    • Docx: For Word documents that contain relevant data for the AI. File size is limited to 10MB.
    • Integration: Connect to a third-party software (e.g. Zendesk Help Center) to import help articles.
  4. Confirmation: After the data source is selected and the upload process is initiated, you will see an indicator that shows the upload progress. Once complete, a confirmation message will indicate that the data source is now integrated and active.

Once you add a data source, it will show you the loading status, and then confirm once the data source has been loaded and is ready to use. 

Managing Data Sources

In the Knowledge Base, managing data sources is streamlined to ensure your AI agent is working with the most current and relevant information. Two key features facilitate this: Re-sync and Refresh.


The Re-sync feature is designed to manually update the selected URL data sources instantly. Here's how it works:

Initiating a Re-sync

  1. Locate the Data Source: In your list of data sources, find the one you want to update.
  2. Accessing the Option: For individual data sources, click the three-dot menu icon (...) on the right side of the data source entry.
  3. Select Re-sync: From the dropdown menu, choose the 'Re-sync' option to start the update process immediately.

Re-sync in Bulk

  1. Select Multiple Data Sources: If you need to re-sync several data sources, check the boxes on the left side of each entry you want to update.
  2. Bulk Re-sync: With multiple data sources selected, click the 'Re-sync' button on the toolbar above the list to initiate a bulk re-sync.
  3. Confirmation: The system will then confirm the initiation of the re-sync process for all selected data sources.


The Refresh feature (only available on paid plans), on the other hand, allows you to set an automatic update frequency for each URL data source, ensuring that the Agent always pulls from the most updated content without manual intervention.

Setting the Refresh rate

Screen Shot 2024-01-08 at 2.08.34 PM.png

  1. Select the data source. Select the refresh link in the column of the data source, or by right-clicking the data source.
  2. Change frequency. In the dropdown menu, you have the option to select how often you want the data source to be refreshed: 'Never', 'Daily', 'Weekly', or 'Monthly'. Select your preferred frequency based on how often the information on the source website is updated.
  3. Update. Once you’ve selected the frequency, the system will automatically update and save your preference.

Set Refresh in Bulk

  1. Select Multiple Data Sources: If you need to set the refresh rate of several data sources, check the boxes on the left side of each entry you want to update.
  2. Bulk update: With multiple data sources selected, click the 'Refresh rate' button on the toolbar above the list to display the dropdown of options.
  3. Confirmation: After selection, the system will then notify you of the successful update of the refresh rate settings.

Delete a Data Source

Single Data Source

  1. To delete a single data source, select the data source you wish to remove, and in the editor view click the '...' (more options) button, and select 'Delete' from the dropdown menu.
  2. Confirm the deletion when prompted to remove the data source from your CMS.

Bulk Deleting Data Sources

  1. For bulk actions, select multiple sources by selecting the checkboxes next to the data source names.
  2. Once selected, on the bulk action toolbar above the table, click the 'Delete' button to initiate the bulk deletion process.
  3. Confirm the bulk deletion to remove all selected intents simultaneously.

Configuring your Knowledge Base Settings

You can modify the performance of your Knowledge Base by using the KB Settings modal, access from the 'Settings' icon in the header to the left of 'Preview' and 'Add Data Source' buttons.Screen Shot 2024-01-08 at 11.34.26 AM.png

To modify how answers are generated from the data found in the sources you provided, you can use the following settings:

  • Model - This is the model that will be used to created your prompt completion. Each model has it's own strengths, weaknesses, and token multiplier, so select the one that is best for your task. 
    • GPT-3.5-Turbo - fast results, average performance on reasoning and conciseness
    • GPT-4 - most advanced reasoning and conciseness, slower results (only available on Pro and Enterprise Plans)
    • Claude 1 - consistently fast results, moderate reasoning performance
    • Claude Instant 1.2 - fastest results, best suited for simpler functions
    • Claude 2 - advanced context handling (strong summarization capabilities)
    • Claude 
    • Claude 3 - Haiku - designed for brevity and impact, delivering concise and powerful responses. Ideal for applications requiring short, impactful messages such as notifications, alerts, and concise answers to user queries.
      Claude 3 - Sonnet: Balancing between brevity and detail, Sonnet is optimized for medium-length interactions. This model is excellent for customer support, FAQ interactions, and more in-depth dialogue scenarios.
    • Claude 3 - Opus: Opus excels in generating long, detailed responses. This model is perfect for content creation, and detailed explanations.
  • Temperature - This will allow you to influence how much variation your responses will have from the prompt. Higher temperature will result in more variability in your responses. Lower temperature will result in responses that directly address the prompt, providing more exact answers.
  • Max Tokens - This sets the total number of tokens you want to use when completing your prompt. The max number of tokens available per response is 512, (we include your prompt and settings on top of that). Greater max tokens means more risk of longer response times.
  • Instructions (only available in Preview mode or on AI Steps): This field is optional. You can use it to add custom instructions to your prompt.
  • System - This is the instruction you can give to the LLM model to frame how it should behave. Giving the model a 'job' will help it provide a more contextual answer. Here you can also define  response length, structure, personality, tone, and/or response language.
  • Chunk Limit - controls the amount of chunks used to synthesize the response. 

Add documents via Integration

Zendesk Help Center

Screen Shot 2024-02-26 at 12.18.04 PM.png

  1. Integration Option: Click on the 'Add Data Source' button, then select 'Integration'.
  2. Select Zendesk Help Center: From the list of integrations, choose 'Zendesk Help Center’
  3. Awaiting Authorization: After selecting Zendesk, you will be prompted to authorize the connection to your Zendesk account.
  4. Importing Articles: Post-authorization, the import modal allows you to filter and import only articles that are applicable for your knowledge base. All applicable filter values will be available in the dropdown for selection. NOTE: only Published articles are available for import. Filters include:
    • Brand (required): Before importing any articles, you must select the brand which to fetch the articles from. NOTE: you can run the import operation as many times as necessary from across various brands if necessary.
    • User Segment: This field allows you to filter by user segment.
    • Locale: This field allows you to only import articles for a certain language. You can get the locale from the Zendesk Help Center URL after you've selected a specific language.
    • Categories: If you have multiple categories then you can optionally filter articles for a specific category.
    • Labels: This field allows you to filter by labels applied to your articles.
    • Refresh Rate: Decide how often you want the document to sync with Zendesk, ensuring your AI has the latest information.
  5. Import Confirmation: Once you finalize the import filters, you will be shown the number of articles that will be imported in the Import button.

Integration management

Screen Shot 2024-02-26 at 9.07.56 AM.png

  1. Open Integrations panel: In the Knowledge Base, you can manage your integrations by selecting the Integrations icon in the header of table.
  2. Update or remove connection: This will show you a list of all your connected integrations (more coming soon), including Zendesk, with the option to reconnect with a different account or to remove them. NOTE: if you delete the connection, any refresh rate assigned to the Zendesk articles will fail unless you reconnect with an account that has access to those same articles.

Preview Knowledge Base Response

Previewing your Knowledge Base is an essential step in ensuring the accuracy and relevance of your Agent's responses before they reach your users. This feature simulates real-world interactions, allowing you to assess and refine the Agent’s responses.

  1. Access the Preview: Locate the 'Preview' button positioned in the header of your Knowledge Base CMS interface.

  2. Input Your Question: In the 'Question' field of the Knowledge Base preview pane, enter a question or query that a user might ask. This should be representative of the actual queries you expect once your Agent is live.

  3. Send the Query: After typing your question, click the 'Send' button. The system will process the query using the data sources and AI model you’ve configured.

  4. Review the Response: The AI's response will be displayed in the 'Response' section. It's crucial to examine not only the direct answer but also the sources from which the Agent retrieved the information, which are listed under the 'Sources' section. All sources can be selected to link to the relevant document they are coming from. 

  5. Evaluate the Sources: Review the listed sources to verify the Agent is pulling data from the correct segments of your Knowledge Base.

  6. Refine and Iterate: Based on the preview results, you may find it necessary to adjust your data sources, tweak the AI model settings, or refine the information within your Knowledge Base to improve the accuracy and quality of the responses. You can select the 'Re-use last question' button to resend the previous question again with the updated settings. 

Tips for Effective Previewing:

  • Test with Variety: Use a wide range of queries, including those phrased differently, to ensure robustness in the AI’s responses.
  • Check for Consistency: Ensure that the responses are consistent with previous answers and with the data provided.
  • Test with different settings: View the previewing process as iterative. The goal is to continuously enhance the AI’s ability to respond accurately. 
    • Preview-Specific Settings: Changes made in Preview are temporary and will not impact the KB's overall settings.
    • Overriding Defaults: To test different behaviours, you can adjust the AI model, temperature, max tokens, instructions and system prompt which will override the global KB settings. You can see how many settings have been changed, or reset back to the KB default by selecting the link at the top of the settings modal:

Screen Shot 2024-01-08 at 3.23.16 PM.png

    • Permanent Changes: If satisfied with the preview results, you must manually update these preferences in the KB Settings to apply them globally.


The Knowledge Base is used as a component of the below steps and events in Voiceflow:


Was this article helpful?

10 out of 12 found this helpful

Have more questions? Submit a request



Please sign in to leave a comment.