RAG-based AI chatbot — answers grounded in your own documents
RAG (Retrieval-Augmented Generation) means the chatbot answers from YOUR own documents, not from its training data. Nortinia AI Chat indexes your PDF, DOCX, HTML and CSV materials, runs hybrid retrieval (vector + keyword), and cites the source under every answer. For business decision-makers this means: the chatbot does not hallucinate — it always tells the organisational truth.
What it does
How people use it
Internal HR knowledge base
HR is overloaded with vacation, bonus and policy questions. Nortinia AI Chat indexes the HR handbook, the collective agreement and the latest internal memos; employees get precise answers with source citations, and HR only handles the genuinely unique cases.
Legal document library
The legal team holds old contracts, litigation files and internal policies. The RAG engine enables natural-language search across these ("What is the notice period in the 2022 executive contracts?") — answers come with source citations referencing the page number.
Technical documentation
Technical support teams work with product catalogs, installation guides and error-code references. The chatbot indexes all of them, and a field engineer asks a natural-language question to get the exact PDF page — service time drops measurably.
Training materials
Education institutions / corporate training units store notes, slides and case studies. Students / trainees ask natural-language questions, and the relevant passage is shown immediately under the answer — a personal tutor for every participant.
Frequently asked
Which LLM does it work with?
It is LLM-agnostic: you can choose between OpenAI (GPT-4o, GPT-4o-mini), Claude (Sonnet 4, Opus 4), Mistral and Llama 3. The Nortinia AI Assistant engine picks the best and most cost-effective model for each task automatically — and you can pin a specific model if you prefer.
What vector database does it use?
The SaaS version runs on our hosted vector store (pgvector + Qdrant hybrid) in the EU region. For on-premise deployments Qdrant, Weaviate or Milvus are options — the Nortinia team recommends one based on document volume and the latent query pattern.
Can it be deployed on-premise on our own server?
Yes. On-premise deployment is Docker / Kubernetes based, running Nortinia AI Chat, the vector database and optionally a self-hosted LLM in a closed network. This is the typical choice for compliance-sensitive environments (banking, healthcare, government).
How often does the document index refresh?
Configurable: daily cron, hourly cron or webhook-driven real-time refresh. In webhook mode your own document system notifies us automatically and the index refreshes within minutes — the chatbot always answers from the freshest documents.
More Nortinia AI Chat solutions
Interested? Let us chat for 15 minutes.
Send a quote request and hear back within 24 hours. No commitment, no pressure.