Semantic document search
Search your documents by meaning, not keywords. Ask a question in plain language and find the right document — invoices, contracts, emails, notes — even when the exact words don't match.
Applications → Qdrant
Find anything by meaning, not keywords.
Qdrant stores vector embeddings of your documents so you can search by meaning, not just keywords. Ask 'show me the contract from last quarter' and find it instantly — even if the exact words don't appear anywhere. It's the memory layer behind ReefOffice's private document search.
What Qdrant does
Qdrant turns your documents into vectors — mathematical representations of meaning — and stores them for instant semantic search. When you search for something, Qdrant finds documents that match your intent, even if they use completely different words. Every ReefOffice plan includes Qdrant for automatic document indexing.
Why Qdrant on ReefOffice
Automatic indexing
Every file you add to Nextcloud, Paperless or Dolibarr is automatically indexed. No manual setup — your document corpus stays searchable from day one as new files arrive.
AI memory layer
Qdrant powers the memory behind Hermes Agent and Open WebUI document Q&A. When your AI answers questions about your files, it retrieves the right context from Qdrant to give accurate, sourced answers.
Fully private
All embeddings are stored on your ReefOffice server — no data is sent to external search or AI services. Your document search stays completely within your infrastructure.
How it works
Under the hood
Qdrant is a high-performance vector database written in Rust. On ReefOffice it runs as a Docker container with persistent storage. Documents are embedded using bge-m3 — a multilingual, MIT-licensed model — and stored as vectors for fast similarity search.
How it fits your stack
Qdrant is the search backbone for ReefOffice. It powers document search in Open WebUI, provides context for Hermes Agent actions, and can be triggered by Activepieces automations. When you ask "which invoices are unpaid?", Qdrant finds the relevant documents in milliseconds.
Frequently asked questions
How is this different from regular search? +
Regular search matches exact words. If you search for "office lease" but the document says "workspace rental agreement", regular search won't find it. Qdrant understands meaning — it finds documents that are semantically similar, even with different wording.
How much storage do vector embeddings use? +
Roughly 1 GB per 100,000 pages of text. Most clients use far less. Qdrant stores only the vectors, not the original documents — those stay in Nextcloud, Paperless or Dolibarr where they belong.
Does Qdrant work without Ollama? +
Yes. Qdrant is included on every ReefOffice plan for document indexing and search. Ollama runs on every plan too — for embeddings — but only AI plans load the generation models that answer questions from your documents.
Ready for smarter document search?
Book a demo and see how Qdrant on ReefOffice makes your documents findable by meaning — not just keywords.