AI knowledge base for businesses.
RAG system from Germany, accessible on every channel.
Knowledge sits everywhere, in PDFs, wikis, Word documents, FAQ lists, old emails. Employees and customers must search manually. Classic knowledge management tools are search tools, they show search results but don't answer. Generative AI like ChatGPT can answer, but doesn't know your company knowledge. What's missing: a RAG system that connects both, and is accessible across every channel (phone, chat, email).
Get startedIn short
RAG (Retrieval Augmented Generation) is the standard architecture for enterprise AI knowledge bases: documents are translated into vector embeddings and stored in a vector database. On queries, the system searches the most relevant passages, gives them as context to an LLM that generates an answer. Advantage over classic search tools: natural language answer instead of link list. Advantage over generic AI: specific company knowledge instead of general knowledge. But building RAG systems yourself is custom enterprise engineering: license a vector DB (US cloud service or self-hosted), configure LLM integration, build embedding pipeline, choose chunking strategy, deploy re-ranking model, build compliance layer for GDPR/AI Act. Time: 2-6 months custom implementation. Pricing: 5-digit setup + 4-digit monthly cloud costs. rinqo solves this: knowledge management is out-of-the-box in the product, a self-operated vector database on Hetzner in Germany, AI models with contractually guaranteed EU data processing, AI Act compliant disclosure. Activation via wizard.
With rinqo you get an AI knowledge base with multi-channel output. One central RAG-based knowledge base per organization, fed from PDFs, website crawl, FAQ lists, helpdesk articles; systems with a REST API such as Confluence or Notion you connect as a link, the agent looks things up there in the conversation. Accessible on every channel: phone (a call delivers the answer from the knowledge base), chat (widget with function calling), email (agent presents answers to standard requests for approval) and the personal agent (internal Q&A search for the whole team), marketing and future channels draw on the same base. Hetzner DE hosting.
Contents
Why rinqo?
RAG out-of-the-box instead of custom engineering
Classic RAG implementation: 2-6 months custom engineering with vector DB license, LLM setup, embedding pipeline, chunking, re-ranking, compliance. rinqo: wizard setup, self-operated vector database, AI service with EU data processing, AI Act disclosure integrated by default.
Multi-channel output without separate integration
Classic KB tools are web frontends. rinqo makes the same knowledge base accessible via phone, chat widget, email triage and the personal agent, without separate channel integrations.
Hetzner DE instead of US cloud vector DB
Common vector DB services host in US regions. rinqo runs the vector database itself on Hetzner Falkenstein/Nuremberg, your knowledge base doesn't sit with a US cloud vector service, it stays in the EU. DPA automatic on contract close.
Knowledge gap resolution continuous learning
When the AI cannot answer a query, the gap is reported in dashboard. You add the answer in one minute, next time the AI can answer. Knowledge base grows from real customer queries instead of theoretical FAQ maintenance.
AI Act Article 50 disclosure standard
On every AI-generated answer: disclosure notice (voice announcement, chat badge, email footer). Mandatory from 2026-08-02, default with rinqo. Classic RAG custom implementations must build disclosure manually.
Function calling for structured answers
Instead of just text answers the AI can call functions: book an appointment in the calendar, log lead in CRM, trigger webhook, escalate to human. Knowledge base becomes active (acts) instead of passive (only informs).
Before. After.
Classic options for company knowledge management: (1) wiki and document tools, web frontend for search, no AI answer, no multi-channel availability. (2) Custom RAG implementation, 2-6 months engineering, 5-digit setup costs, US cloud vector DB, manual GDPR compliance. (3) Generic AI chats, don't know your company knowledge, give invented answers (hallucinations).
rinqo (DE, Hetzner): RAG system out-of-the-box with a self-operated vector database in Germany, AI models with EU data processing, AI Act standard. Multi-channel output: phone + chat + email + personal agent. Knowledge gap resolution for continuous learning. Setup via wizard.
What is an AI knowledge base? RAG for businesses explained
RAG (Retrieval Augmented Generation) is the leading architecture for AI knowledge bases. Three steps: first, documents are translated into semantic vector embeddings and stored in a vector database. rinqo runs the vector database itself on Hetzner in Germany. Second, on a query the system computes the embedding of the question and searches the most similar document passages, typically top-5 hits. Third, found passages are given as context to an LLM (at rinqo with contractually guaranteed EU data processing) that generates a natural language answer. Advantage: specific company knowledge instead of general knowledge, natural language answer instead of link list. Trade-off: needs quality knowledge base content and clean chunking. rinqo has chunking + re-ranking calibrated for European SMB knowledge bases by default.
How rinqo's knowledge base is built, four input sources with approval
The knowledge base accepts several input types. First, website crawl. You provide your main domain URL, rinqo crawls all indexable pages automatically, extracts text, embeds in the self-operated vector database. Second, file upload. PDFs, Word documents, Markdown files, Excel sheets are uploaded. The text is extracted and embedded automatically. Works for helpdesk article collections, employee handbooks, product documentation, compliance guidelines. Third, third-party sources. Systems with a REST API such as Confluence or Notion you connect as a link, the agent looks things up there in the conversation; on request we set that up. Fourth, from everyday work itself. When someone asks the personal agent something that isn't stored yet, or a human answers an escalated request, rinqo proposes the new knowledge block. Important: a suggestion never goes live automatically. A manager approves it and decides whether it belongs in the shared company knowledge or in that person's personal knowledge. This way the knowledge base grows in a controlled way out of daily work, without tedious upkeep, and nothing lands in the answers unchecked.
Multi-channel output, one knowledge base for every channel
Classic knowledge management tools are web frontends, employees and customers must open the app, search, read. On phone calls, chat queries or emails the knowledge must be manually retrieved by someone on the team. rinqo flips this: the same knowledge base is dispatched across every channel. Phone: AI phone agent takes the call, searches the knowledge base, answers spoken (speech synthesis from a German provider, EU-hosted). Chat: web widget with the same knowledge base. Email: autonomous triage agent connects via IMAP to the inbox, answers queries that exist in the knowledge base, escalates complex emails. Personal agent: direct Q&A search for the whole team, with a source citation on every answer. One knowledge base, every channel, no knowledge silo.
Assign knowledge, each agent only sees what it should
One knowledge base doesn't mean everyone sees everything. In rinqo you assign each agent exactly which knowledge it may answer from. The chat agent on your website gets opening hours, services and prices; the email agent additionally the complaints procedures; internal HR files, purchasing terms or contracts stay entirely out of reach. That way a customer-facing agent only answers what is meant for customers and never accesses internal documents, the separation is structural, not a matter of how the prompt is worded. The Personal Agent for the team, conversely, may access the internal knowledge, and each person additionally their own personal knowledge, which is visible to no one else. You set the assignment in the dashboard and change it at any time; a change takes effect immediately across every channel, without maintaining documents twice.
Knowledge gap resolution, knowledge base learns from real queries
Classic RAG systems have a typical problem: knowledge base is initially incomplete. Employees and customers ask questions not covered in knowledge base. Classic systems then answer 'I cannot answer that' or worse, hallucinate. rinqo has a continuous learning loop. Step 1: on an unanswerable query, AI escalates honestly ('This question I cannot currently answer, connecting you with a colleague') AND captures the gap structured in dashboard. Step 2: you see all open knowledge gaps sorted by frequency. Step 3: you add the answer in 30 seconds. Step 4: next same or similar query is answered automatically. Knowledge base grows from real customer queries instead of theoretical FAQ maintenance.
rinqo vs. classic knowledge management tools, comparison
Three options European SMBs compare in 2026 for knowledge management. First, classic wiki and KB tools. Strength: mature editorial interfaces, versioning, permissions, team collaboration. Weakness: web frontend only, no AI answer, no multi-channel availability. Pricing typically €8-15/user/mo, no AI layer included. Second, custom RAG implementation. Strength: maximum flexibility in architecture and integrations. Weakness: 2-6 months engineering, 5-digit setup costs, cloud DB licenses, manual GDPR/AI Act compliance. Third, generic AI chats. Strength: mature LLM answers. Weakness: don't know company knowledge (unless you paste context manually), hallucinate on specific queries, often US cloud hosting. rinqo combines the three advantages: a maintained knowledge base with editorial basics plus RAG answers plus GDPR/EU hosting (Made-in-Germany).
Frequently asked questions
Sources
Last updated: 2026-09-05
Founder, rinqo
Founder of rinqo. Builds a company memory where an organisation's knowledge comes together and stays available, and the AI agents that work from it: on the phone, in the inbox, in chat, in marketing and in further roles. Writes here about bringing AI into companies, about data protection, and about what holds up in daily operation. Developed in Germany, processed in the EU, no US providers.
Last updated:
Try it live
View agent