How DĀNIS Works
An inside look at our retrieval-augmented generation engine. We turn raw website documents into a secure, hyper-accurate conversation agent.
Data & Inference Pipeline
1. Input Sources
2. RAG Processing
Clean & Chunk
Strips HTML headers/footers, cuts text into 500-token blocks.
Vectorization
Embeds text meanings using our secure model.
Isolated Namespace DB
Saves vectors into a client-scoped, secure database.
3. End User Chat
Customer Asks Question
Grounded AI Answer
Step 1 — Crawl or Upload
Getting content into DĀNIS is completely hands-free. You provide a seed website link (for example, your main support page or admissions hub). DĀNIS scans all nested links, downloads the HTML content, and strips out unnecessary components like navigation headers, footers, and advertising banners.
Have brochures, course books, or student handbooks trapped inside static document formats? Simply drag and drop your PDFs or docx files into the dashboard. DĀNIS extracts the text from them directly, merging multiple distinct formats into a unified repository.
Step 2 — Index & Understand
Once content is collected, it cannot simply be fed to the AI as is. Long documents would overload the context limit. DĀNIS uses semantic chunking to break documents down into readable, self-contained paragraphs.
These chunks are processed by an embedding model. This is the magic step: it translates the textual meaning into mathematical vectors. Instead of searching by exact keyword matches, our vector search searches by semantic intent. If a user asks "how much is it to attend?", the system immediately understands that "attend" relates closely to "tuition fees" and "costs," retrieving the correct source chunks instantly.
Step 3 — Answer in Real Time
When a customer asks a question in the embedded chat widget, the RAG cycle triggers. We query the isolated vector space for the top 3-4 chunks that are semantically nearest to their question.
These retrieved facts are compiled together with the user's question and a strict bounding instruction: "You are a helpful assistant. Rely ONLY on the facts below. Do not make anything up." The AI processes this prompt and generates a natural, conversational response within seconds, complete with direct links to the relevant website source pages.
See the RAG engine in action
Test how DĀNIS answers questions about pricing, setup, and college handbooks using our interactive sandbox page.