// 7 apps · deploy as one
Private AI Plus Workspace
Self-hosted ChatGPT + local LLMs + vector RAG + collaborative docs + voice transcription. Your team's AI infrastructure under your own DNS, no SaaS dependency.
What this stack is
Private AI Plus Workspace is a complete AI-native productivity suite that you (the hosting provider) deploy in one click to a customer's VPS. Your customer's team gets a self-hosted ChatGPT, local LLMs, vector RAG search, collaborative docs, and voice transcription — all on a single VPS, under their own domain, with zero upstream telemetry.
This is the deployment product for hosters who want to sell turnkey AI workspaces to organizations that cannot — or will not — push their data to US LLM clouds. Works in any region; EU residency is just one of the deployment options the hoster picks at provisioning time.
What it's for
- Internal team AI assistant — Open WebUI on top of Ollama gives every employee a private ChatGPT with conversation history, RAG, and team workspaces
- Document intelligence — Paperless OCR-scans PDFs and contracts, AI assistant answers questions across the corpus
- Real-time collaborative editing — Nextcloud + OnlyOffice replaces Microsoft 365 for files, docs, sheets, presentations
- Vector knowledge base — Qdrant powers semantic search across documentation, sales notes, support tickets
- Voice → text — Whisper transcribes meetings, calls, voice notes for indexing and Q&A
- Audit-ready compliance — every prompt, every document access logged on customer infrastructure, no third-party processor
Who it's for
- EU enterprises that can't push data to US AI clouds (GDPR Schrems II, sector regulators)
- Consulting firms wanting an internal ChatGPT trained on client matter docs, with zero leakage risk
- R&D departments running confidential research that must stay on-prem
- Government agencies at municipal/regional scale wanting modern productivity without vendor lock
- Universities giving students AI tooling under FERPA + GDPR
Apps in this stack
| App | Role |
|---|---|
| Open WebUI | ChatGPT-style frontend, multi-user auth, conversation history, RAG document upload |
| Ollama | Local LLM runtime — Llama 3.3, Mistral, Qwen, DeepSeek run on the customer's VPS |
| Qdrant | Vector database for semantic search + RAG retrieval |
| Nextcloud | File sync, sharing, team folders, calendar, contacts |
| Paperless | Document OCR, tagging, full-text search across scanned PDFs |
| OnlyOffice | Word/Excel/PowerPoint compatible real-time collaborative editing |
| Whisper | Voice-to-text transcription for meetings, calls, dictation |
Sizing & deployment
- Recommended VPS: 16 GB RAM, 8 vCPU, 200 GB NVMe storage
- GPU recommended for 7B+ LLMs (RTX 4060 / L4 / equivalent). CPU-only works for smaller models + RAG-only workflows
- EU regions: Hetzner (Falkenstein/Helsinki), Contabo (Nuremberg), OVH (Strasbourg/Gravelines), Scaleway (Paris)
- Deploy time: ~25-35 min one-click, all SSL automatic via Let's Encrypt
- Backup profile: nightly snapshots of Nextcloud data + Paperless archive + Qdrant collections + Ollama model state
Why hosters sell this
Resellers position this stack as a turnkey "Private AI Workspace" subscription for end-customer teams. The combination Open WebUI + Ollama + RAG + collaborative docs maps directly to what enterprise IT departments are trying to assemble themselves out of OpenAI ChatGPT Enterprise + Box + Slack + Otter.ai — except self-hosted, single-vendor, and aligned with the customer's data-sovereignty mandate.
Pricing for end-customer subscriptions is set by the hoster based on their cost stack (VPS + GPU + Bluix license + support). The Bluix module license layer is detailed in bluix.net/bluixapps.php; stack retail pricing pending GPU partner agreement — contact sales for a current cost-build model.
Common resale angles validated in early hoster conversations:
- Internal AI assistant for SMBs — companies of 20-200 employees who'd otherwise sign for ChatGPT Team but want sole-tenancy
- Confidential research workspace for consulting + R&D departments
- University AI lab — single VPS per faculty, students access under SSO
- Regional government departments at municipal/regional scale
The AI vendor relationship (Ollama, Open WebUI) is open-source = the hoster keeps 100% of the gross margin above their cost stack. No third-party rev-share leak.
Compliance profile
- No data leaves the VPS unless the customer explicitly enables a cloud LLM endpoint (BYOK OpenAI/Anthropic — disabled by default)
- Audit log of every prompt + every document access — exportable for DPO review
- At-rest encryption for Nextcloud data folder + Paperless archive + Qdrant snapshots
- DPA-ready: hoster acts as processor, end-customer is controller, BluixApps software supplier only
- Deploy region picked at provisioning: EU regions for Schrems II / GDPR compliance, US/APAC/LATAM available for jurisdictions with their own data-residency rules
- Cloud-LLM-clean: no US-soil processing, no Microsoft/Google dependencies, no telemetry to third parties — by default