Knowledge Base & RAG

Features

Knowledge Base & RAG

What is RAG?

Retrieval-Augmented Generation (RAG) means the AI does not rely on its training data alone. Before generating a response, it searches your knowledge base for the most relevant content, then uses that content to craft an accurate, grounded answer.

How It Works (Step by Step)

  1. Import -- You add content via text, URL, sitemap, PDF, or a Markdown/TXT file.
  2. Chunking -- The system splits content into ~500-token chunks with overlap for context continuity.
  3. Embedding -- Each chunk is converted to a vector using OpenAI text-embedding-3-small.
  4. Indexing -- Vectors are stored and indexed for fast similarity search.
  5. Query -- When a visitor asks a question, the query is embedded and a semantic search retrieves the top-K most relevant chunks.
  6. Generation -- The AI model receives the retrieved chunks as context and generates a factual response.

Import Methods

MethodHowBest For
TextPaste directly into the editorFAQs, policies, quick snippets
URLEnter a web page URL; Conviro crawls itBlog posts, product pages, docs sites
PDFUpload a PDF file (max 10 MB)Manuals, whitepapers, catalogs
Markdown / TXTUpload a .md or .txt fileInternal SOPs, FAQs, training notes

Chatbot-Scoped Knowledge Bases

Each chatbot can have its own dedicated knowledge base. This means:

  • A "Sales Bot" only sees sales-related content.
  • A "Support Bot" only sees support-related content.
  • No cross-contamination of answers.

Configure this in Dashboard -> Assistants -> Your Bot -> Knowledge Base tab.

Tips for Best Results

  • Use FAQ format -- "Q: How do I reset my password? A: Go to Settings -> Password..." gives the clearest signal to the AI.
  • Keep chunks focused -- One topic per source. Avoid uploading a 200-page PDF without splitting it.
  • Update regularly -- Outdated information leads to wrong answers. Review sources quarterly.
  • Add negative examples -- "We do NOT offer phone support" prevents the AI from hallucinating features you do not have.
  • Test after every import -- Open the widget and ask questions related to the new content.
  • Monitor AI answers -- Check Sessions history for low-confidence or incorrect responses and refine your KB.
ragknowledge-baseembeddingimportfaqpdfurl

Was this article helpful?