DocumentationKnowledge Base

Sources & Knowledge Base

Learn how to build and manage your AI knowledge base with sources and documents.

Overview

The knowledge base is the foundation of Cevvo. It stores all the content your AI assistant uses to answer questions. You populate it by creating sources — connections to your existing documentation, files, or websites that Cevvo crawls, indexes, and keeps in sync.

Source Types

Website

Crawl and index entire websites or specific pages. Supports sitemap detection and automatic link following.

File Upload

Upload PDF, DOCX, TXT, and Markdown files directly. Content is extracted and indexed automatically.

Text

Paste or type content directly. Great for FAQs, internal notes, or custom knowledge entries.

How It Works

1

Create a Source

Add a website URL, upload files, or paste text content to create a new source.

2

Ingestion & Indexing

Cevvo crawls the source, extracts text, splits it into chunks, and builds a vector index for semantic search.

3

AI Answers

When a user asks a question, Cevvo retrieves the most relevant chunks and generates an answer with source citations.

4

Keep It Fresh

Enable auto sync or manually re-sync sources to ensure your knowledge base stays current.

Auto Sync

Enable auto sync on any source to keep your knowledge base up to date automatically. Cevvo will periodically re-crawl your sources and update the index with new or changed content. You can configure the sync frequency per source.

Tip: Auto sync is especially useful for website sources where documentation changes frequently. For static file uploads, you can trigger a manual re-sync whenever needed.

Best Practices

  • Use descriptive names for sources so you can easily identify them in analytics.
  • Prefer website sources over file uploads for content that changes frequently.
  • Break large knowledge bases into multiple focused sources for better organization.
  • Review coverage gaps in analytics to identify missing content areas.
  • Test your knowledge base in the playground before deploying to production.