Rüko GPT

An enterprise RAG chatbot that lets 50+ employees of a construction-machinery company query internal knowledge in plain language.

Year: 2025 —
Role: AI Developer
Client: RÜKO GmbH Baumaschinen
Stack: Next.js · TypeScript · Prisma · PostgreSQL · OpenAI API · NextAuth.js · RAG

Rüko GPT — enterprise RAG chatbot interface

01 — Problem

RÜKO's institutional knowledge — machine specs, internal processes, project documents — lived scattered across folders and in people's heads. Finding an answer meant knowing whom to ask, and asking cost both sides time.

Off-the-shelf chatbots couldn't be used: the data is confidential, the questions are domain-specific, and answers need to be grounded in the company's own documents rather than a model's imagination.

02 — Approach

I design and build Rüko GPT end to end: a Next.js + TypeScript application with Prisma and PostgreSQL underneath, NextAuth.js for company-account authentication, and the OpenAI API for generation.

Documents are processed into a retrieval pipeline so every answer is grounded in real company sources. A separate REST API server extends the system into the company's wider knowledge-management landscape.

Rollout is iterative — shipping to real users early, then tuning retrieval quality and UX based on how employees actually ask questions.

System architecture of Rüko GPT — Next.js client, retrieval pipeline core, PostgreSQL data tier and REST API extension

50+Employees served

2025In production since

E2EDesigned and built solo

03 — Result

Rüko GPT is in production as the company's internal answer machine, serving 50+ employees. Instead of digging through folders or interrupting a colleague, people ask the bot — and get answers grounded in RÜKO's own documents.

Rollout loop and adoption panel — ship, observe, tune; 50+ employees served — Fig. 03 — Iterative rollout: ship early, tune retrieval on real questions.

Next projectTeams-BOT02