// 5 apps · deploy as one
AI Knowledge Base
Chat with your company documents. Self-hosted ChatGPT + vector RAG + OCR archive + voice transcription + file sync. CPU-friendly entry tier into private AI.
What this stack is
AI Knowledge Base is the CPU-friendly entry tier into private AI. One-click deploys a self-hosted ChatGPT (Open WebUI), a vector database for retrieval (Qdrant), an OCR-powered document archive (Paperless), voice-to-text transcription (Whisper), and a file sync hub (Nextcloud).
Your customer drops PDFs, scanned contracts, meeting recordings, internal wikis, and emails into the system. The AI indexes everything and lets users ask questions across the entire corpus — answers cite the source document and page. No data leaves the VPS, no per-seat ChatGPT subscription, no upload limit, no cloud-LLM telemetry.
This is the "Chat with your company documents" product. Sells itself once a prospect sees the demo.
What it's for
- Internal company knowledge base — every employee can ask the AI questions about HR policies, technical docs, customer history, past projects
- Consulting firm knowledge management — partners and associates query across decades of project deliverables, proposals, case files
- Sales enablement — reps query a unified corpus of product specs, pricing sheets, competitor analysis, won/lost deal notes
- Customer support augmentation — agents get instant answers from past tickets, documentation, internal runbooks
- Research synthesis — academics, analysts, and R&D teams index papers, reports, datasets and query across them with citations
Who it's for
- SMBs (10-100 employees) wanting "internal ChatGPT trained on our docs" without enterprise procurement
- Consulting firms with years of accumulated client deliverables that nobody has time to read
- Departments inside larger companies that can't get IT to deploy an LLM solution corporate-wide
- Privacy-bound organizations (legal, healthcare, finance, government) that need AI-on-our-docs but can't push to cloud LLMs
- Researchers and analysts who want a personal RAG system over their bibliography and notes
Apps in this stack
| App | Role |
|---|---|
| Open WebUI | ChatGPT-style frontend with multi-user auth, conversation history, citations |
| Qdrant | Vector database — stores document embeddings for fast semantic search |
| Paperless | OCR-powered document archive — scans PDFs, extracts text, auto-tags |
| Whisper | Voice-to-text — transcribes recordings (meetings, calls, voice memos) into searchable text |
| Nextcloud | File sync + sharing — the inbox where users drop documents for indexing |
Sizing & deployment
- Recommended VPS: 8 GB RAM, 4 vCPU, 150 GB NVMe storage
- NO GPU required — small LLMs (Qwen 7B Q4, Llama 3 8B Q4) run acceptably on CPU for RAG workloads
- Storage scales with corpus size — 1 GB of source documents ≈ 100-300 MB Qdrant index
- Deploy time: ~15-20 min one-click, all SSL automatic via Let's Encrypt
- Backup profile: nightly Nextcloud data + Paperless archive + Qdrant collections snapshot
Why hosters sell this
This stack is the easiest demo to close in the entire BluixApps catalog. Drop 10 of the prospect's own documents into Nextcloud, wait 60 seconds for indexing, ask "summarize the key points across these contracts" — the prospect's eyes light up. From there it's a one-call close.
Common resale shape:
- Solo professional tier — 1 user, single VPS, modest LLM, personal knowledge base
- Team tier — 5-30 users, larger VPS, multiple knowledge collections, SSO
- Department tier — 30-100 users, dedicated VPS, multi-collection RAG, audit logging, branding
Stack retail pricing is set by the hoster based on their cost stack. The Bluix module license layer is detailed at bluix.net/bluixapps.php. The CPU-only nature of this stack makes it the lowest-cost AI deploy in the catalog — ideal entry product.
The pitch is "private ChatGPT trained on your documents, no upload limits, no monthly seat tax". Competes directly with ChatGPT Enterprise / Copilot / Glean at a fraction of the cost.
How it differs from Private AI Plus Workspace
| AI Knowledge Base | Private AI Plus Workspace | |
|---|---|---|
| Apps | 5 | 7 (adds OnlyOffice, Ollama) |
| LLM runtime | optional (BYOK or cloud) | included (Ollama) |
| Collaborative editing | no | yes (OnlyOffice) |
| GPU | not required | recommended |
| Target deploy size | smaller VPS | mid-size VPS |
| Customer profile | "internal knowledge Q&A" | "full self-hosted AI office" |
| Position | entry AI tier | flagship AI suite |
AI Knowledge Base is the starter that lets a hoster get a prospect into the AI category with low VPS cost. When the customer outgrows it, the upgrade path is Private AI Plus Workspace.
Compliance profile
- No data leaves the VPS — Ollama optional, default LLM is BYOK to customer's own OpenAI/Anthropic key
- Per-user audit trail — Open WebUI logs every prompt, every cited document, every response
- At-rest encryption for Nextcloud data + Paperless archive + Qdrant snapshots
- Right-to-be-forgotten — delete a user's documents from Nextcloud + Paperless + re-index Qdrant = full removal
- Deploy region picked at provisioning for EU/US/APAC/LATAM data-residency requirements