Training Maven

Building knowledge

Maven answers are powered by the knowledge you give it. An accurate, complete and robust set of knowledge for Maven enhances its performance and ensures it provides accurate, reliable, and contextually relevant answers.

Adding Knowledge

Compile all the resources your support agents utilize to answer customer queries and upload content to Maven. The supported formats are:

  • Knowledge bases from customer service platforms like Salesforce, Zendesk, and Freshdesk
  • PDF, DOCs and CSVs of training materials (See the FAQ section below.)
  • Public websites and support pages can be added by pointing Maven to the URL and Maven will retrieve the website content
  • Individual knowledge items can be added through the manual option

Knowledge best practices

  • Ensure Data Quality and Diversity: Make sure the data is high-quality, representative, and unbiased to prevent skewed information.
  • Maintain Transparency and Documentation: Document how information is stored, updated, and retrieved within the knowledge base. This helps in understanding the AI’s reasoning and simplifies troubleshooting.
  • Prioritize Privacy and Security: Data used in the knowledge base will be shared by Maven so protect sensitive information to comply with data privacy regulations.

Advanced Options

When creating a new Knowledge Base, there are multiple advanced options you may configure:

  • Refresh Frequency: Set your URL and External knowledge base types to automatically update on a daily, weekly, or monthly basis.

  • URL Exclusion Pattern: Add URL patterns to be excluded from crawling using a glob.

    • Note that this affects only links found on pages, but not Start URLs, which are always crawled.
    • Example: https://{store,docs}.example.com/** excludes all URLs starting with https://store.example.com/ or https://docs.example.com/, and https://example.com/**/*\?*foo=* excludes all URLs that contain foo query parameter with any value.
    • You can learn more about globs and test them here.

  • CSS Elements to Remove: Define specific CSS elements to be ignored when pages are crawled by Maven

  • Max Crawl Depth: Define the depth from your URL’s starting point to be crawled.

  • Max Crawl Pages: Define the total number of pages to be crawled

FAQs

Q: Among the options to add knowledge, which one(s) are preferable?

A: Maven prefers knowledge bases from customer service platforms like Salesforce, Zendesk, and Freshdesk, since they tend to be well-formatted and easily refreshable. File uploads and manual entries are also great options.

Q: What format does the CSV files need to be in to upload into Maven’s Knowledge Base?

A: The CSV files should have a header and at least two columns “title” and “content”. Each row represents a document entry. Optionally, you can also include “url” and “metadata/languagecode” (ISO-639-1 language codes, e.g., “en”, “es”, “de”).