Your chatbot knows your business
— not the internet
While general AI models are based on broad knowledge, Smartwebchat uses the content you provide for your assistant – for example opening hours, service lists, price lists and FAQs. This knowledge base provides the context for its answers.
What is behind it?
Explained simply — no technical background needed.
RAG stands for retrieval-augmented generation — a method in which an AI model does not answer from memory, but actively looks things up in a stored document library before it replies.
Think of your chatbot as a new employee on their first day. You give them a folder with the information they need for customer conversations. Smartwebchat looks for relevant content in that knowledge base and uses it as context for the answer. If that context is not sufficient, the request can be handed over to your team.
Retrieved content is used as context for the answer.
How it works
Three steps to your own knowledge base.
Upload documents
Load your content into your bot's knowledge base: PDFs, text files, or have Smartwebchat read your website automatically. Each bot has its own isolated knowledge base — what you upload for bot A is invisible to bot B.
Automatic processing in the background
Smartwebchat splits your documents automatically into semantic units and stores them in a searchable vector database, entirely on servers within the EU. This runs in the background and only has to be repeated when something changes.
Precise answers in real time
When a user asks a question, Smartwebchat searches your knowledge base for the most relevant passages and formulates a natural-language answer from them. No answers from the open internet.
Supported formats
- PDF documents (price lists, treatment information, terms, information sheets)
- Text files (.txt, .md)
- Website content (automatic scan by entering a URL)
- FAQ documents
- More formats in preparation
Every knowledge base is an island.
Particularly important for agencies: client A's knowledge base is completely separate from client B's — no exchange of information, no mixing, not even while an answer is being generated.
Technically, each knowledge base runs in its own data namespace (pgvector). Even when two bots run on the same server, the data is logically fully isolated.
All documents are stored on servers within the EU; the embedding runs through Mistral in the EU. With the standard model, no content from your knowledge base leaves the EU. If you enable a premium model (OpenAI), the passages used for an answer are transferred to servers outside the EU — that is your decision, bot by bot. GDPR-compliant from the ground up.
Use cases by industry
What your knowledge base can contain.
| Industry | The knowledge base contains |
|---|---|
| Medical practices | Consultation hours, covered services, emergency information, repeat prescriptions |
| Pharmacies | Out-of-hours service, e-prescriptions, delivery service, medication FAQs |
| Lawyers | Areas of law, how an initial consultation works, fee information |
| Driving schools | Licence categories, course times, costs, how to sign up |
| Veterinary practices | Range of treatments, what to bring along |
| Trade & craft businesses | Services, area covered, how a quote is prepared |
| Physiotherapy | Covered services, first appointment, waiting list, referral process |
Build your own knowledge base — 14 days free.
No credit card · No developer needed · GDPR-compliant