RapidCloud Files
Cloud storage that becomes searchable knowledge
RapidCloud Files is OneDrive-style cloud storage with folders, sharing, and desktop sync, hosted entirely in the EU. Its difference is what happens after upload: every file can be classified, parsed (OCR, Office, PDF, audio, video), chunked, embedded, and turned into searchable knowledge through the RapidCloud document pipeline. It is built for teams that need ordinary file storage and the ability to ask questions across everything they store. Access is gated per user, down to allowed IP addresses.
Every file, instantly searchable
RapidCloud Files is cloud storage with a document pipeline built in. Upload once; find anything.
- ⌘K command palette — jump to any file, folder, or action without leaving the keyboard.
- Recursive search — search across every nested folder at once, by name or by meaning.
- Rich previews — images, video, audio, PDFs, code, CSV-as-table, and JSON render inline.
- Folders, projects, tags — organise the way your team actually works.
- Granular sharing — password-protected, expiring links with gallery views and one-click zip download.
- Desktop sync — a OneDrive-style folder on Windows that stays in step.
- EU-hosted, IP-gated — your data stays in Europe, reachable only from addresses you approve.
Start storing files that answer questions. Request access to RapidCloud Files →
Under the hood▾
Under the hood
Frontend is a Vite + React + TypeScript SPA (ky + react-query + zustand for a framework-agnostic API/auth/data layer), served as a static bundle. The redesigned build adds a ⌘K command palette, recursive search, persistent column sort, bulk-selection, and inline previews for images, video, audio, PDF, code (Prism), CSV, and JSON.
Backend is a FastAPI service backed by MinIO object storage. Files register into the document pipeline orchestrator, which runs a six-step workflow — classify → parse → chunk → embed → knowledge — across 16 microservices (docling, unstructured, OCR, Whisper audio, video, entity extraction to Neo4j). Chunks and embeddings land in Qdrant for semantic retrieval.
Access: portal JWT for the SPA, X-API-Key for external clients (Windows sync), and an internal-CIDR bypass for pipeline calls. Per-user IP allowlisting is enforced via Cloudflare Access. Public delivery runs over a Cloudflare Tunnel with chunked uploads up to 10 GB. Share pages are server-rendered with password, expiry, and zip endpoints.
