ChatBot REX
Lessons-learned records scattered across spreadsheets and email. A chatbot and a KPI dashboard now centralize them for engineers, inside the internal network.
DOXA
Query internal documents in natural language without sending them to US servers. Verifiable answers with source citations, hosted in Europe.
- Role
- Design and development
- Period
- April 2026
- Setting
- Personal project
- Next.js 16
- Tailwind v4
- shadcn/ui
- Prisma 7
- PostgreSQL + pgvector
- NextAuth.js v5
- Gemini 2.0 Flash
- Vercel AI SDK v6
In action
Context
European SMEs and mid-sized companies want to query their internal documents (contracts, procedures, reports) in natural language, but stay stuck on tools like NotebookLM for GDPR reasons: their documents must not transit through unmanaged US servers.
Mission
Designed a multi-tenant SaaS: Prisma data model with workspaces isolated per company, NextAuth.js v5 authentication (JWT, secure sessions), a document pipeline (PDF/Word/txt upload, parsing, chunking, embeddings), and a RAG engine with pgvector retrieval and streamed generation via Gemini 2.0 Flash, with citations pointing to the exact source page of the document. Hosting planned on a private European Docker VPS. Project actively in development: multi-tenant infrastructure, authentication and document pipeline shipped, RAG engine with citations functional.
Impact
Companies can query their internal documents in natural language with verifiable answers (citations to the exact source), without their data ever leaving a managed European hosting environment.
Let's talk about your project.
Describe what you need in a few lines; we'll see together whether I can help.
Get in touch