What is RAG and why does it matter?
RAG (Retrieval Augmented Generation) is the latest generation of AI chatbots: before answering, the system retrieves relevant documents and the LLM generates from them.
Result: more accurate answers, cited sources, minimised hallucination. The Nortinia engine uses two-stage retrieval (semantic + keyword) and re-ranking for maximum precision.
What sets Nortinia RAG apart?
Two-stage retrieval
Semantic + keyword search + re-rank. Higher precision on top-k documents.
Citation
Every answer references the source document — full traceability.
Multi-doc reasoning
Connects multiple documents in the answer — doesn't just copy one snippet.
Eval harness
Built-in eval framework: ground truth, retrieval@k, faithfulness, answer relevancy.
RAG rollout in 4 steps
Document import
URLs, PDFs, Markdown, Notion, Confluence, SharePoint — bulk upload.
Chunking + indexing
Adaptive chunking, vector + keyword index, metadata tagging.
Pilot question set
Eval set to measure retrieval@k and faithfulness.
Go live
A/B-tested prompt fine-tuning, model pinning, monitoring.
What it builds on and connects to
Nortinia AI Assistant — product home
Embeddable AI assistant for every surface you own.
Nortinia Engine — LLM routing and decision layer
The Nortinia AI engine — model routing, prompt management, eval harness.
NIP Platform — self-hosted infrastructure
Private cloud and deploy platform for on-premise AI rollouts.
Nortinia.com — AI and software engineering background
The engineering studio behind the Nortinia AI Assistant.
Nortinia Sales AI — outbound side
Assistant talks to visitors on your site. Sales AI calls, writes and advertises prospects on your behalf. Inbound + outbound, one stack.
Nortinia AI Chat — text-only, cheaper
Same Nortinia AI Engine, just text-only — no voice calls. Faster to deploy, lower monthly price. If your focus is text channels, this is the one.
Nortinia AI Call Center — 24/7 voice call center
Full voice-based call center: inbound + outbound, parallel call handling, recording and transcripts. Assistant is task-focused; Call Center is call-focused.
RAG chatbot — questions
What is the difference between a "plain" chatbot and RAG?
A plain chatbot answers from the model's trained knowledge (or hallucinates). RAG always retrieves from your documents and generates only from them.
Which formats are accepted?
PDF, DOCX, Markdown, HTML, URL, JSON, CSV, Notion, Confluence, SharePoint, Google Drive. Uploadable via API too.
How does it minimise hallucination?
Mandatory citation, re-rank on top-k, faithfulness scoring in eval, escalation on low confidence.
Does the knowledge base update automatically?
Yes — automatic crawl + diff-based re-indexing. Configurable interval (hourly to daily).
We'll show the RAG chatbot on your own knowledge base.
30-minute demo, eval set, accuracy estimate at the end.