MBio AI Chat Docs
Documentation / AI & RAG

Knowledge base & RAG

Ground AI answers in approved business information using chunking, embeddings and hybrid semantic retrieval.

What belongs in Knowledge

Add product documentation, pricing explanations, FAQs, policies, onboarding guides, feature descriptions and other facts the assistant is allowed to use. Keep content factual and current.

How RAG works

MBio splits content into searchable chunks, creates embeddings and retrieves relevant candidates for each question. Retrieved knowledge is treated as reference material rather than as instructions, helping reduce prompt-injection risk.

Knowledge gaps

When retrieval confidence is weak, MBio can record the question as a Knowledge Gap. Review these gaps to discover missing documentation and improve future answers.

Quality recommendations

  • Prefer focused sources over very large mixed documents.
  • Remove outdated pricing or policies.
  • Write headings that describe the question each section answers.
  • Test important queries after every major update.