Knowledge Base

The knowledge base is the foundation of Appilot's intelligence. Unlike generic AI assistants, Appilot's answers come from organization-curated content, not free-form model generation.

Knowledge content

Each piece of knowledge is:

• Authored by your organization — not AI-generated
• Scoped to a domain and URL pattern — so it's relevant to specific pages
• Versioned and publishable — draft → published workflow

Example

{
  "title": "How to complete the insurance claim form",
  "content": "Step 1: Enter the policy number (found on your insurance card)...",
  "domain": "claims.insureco.com",
  "url_pattern": "/new-claim",
  "status": "published"
}

When a user is on claims.insureco.com/new-claim and asks a question, the AI receives this knowledge as grounding context.

Scoping: Domains and URL patterns

Knowledge is matched to the user's current page:

Domain	URL Pattern	Matches
app.example.com	/register	Exact page
app.example.com	/settings/*	Any settings subpage
app.example.com	*	Any page on the domain

More specific patterns take priority over general ones.

The demand-driven loop

Appilot creates a feedback loop between users and knowledge teams:

User asks question
    │
    ├─── Knowledge exists → AI answers with grounded content
    │
    └─── Knowledge missing → User creates a request
                                    │
                          Knowledge team sees request
                                    │
                          Team creates/publishes content
                                    │
                          User is notified ← ─ ─ ─ ┘

This means your knowledge base is driven by actual user demand, not guesswork about what to document.

RAG files

In addition to authored knowledge content, organizations can upload documentation files (PDFs, SOPs, manuals) that the AI can reference:

• Files are processed, embedded, and stored
• When a user asks a question, relevant file chunks are retrieved
• The AI uses both knowledge content and file context for answers

See RAG Files for management details.