NimbusVault
Building a Real CMDB
The challenge
The client's IT department was tracking servers, storage, and client environments in spreadsheets and scattered docs: no revision history, no way to see what depended on what, and credentials sitting in plain text wherever someone last pasted them. Off-the-shelf CMDB tools like Docusnap or i-doit are built for teams managing hundreds of assets across many customers, not a single in-house department that needed something closer to a structured wiki.
The approach
- Built a single-tenant Astro app from scratch: typed records (servers, databases, applications, and free-text docs) with structured fields, full diffed revision history, and a directed relationship graph with cycle detection
- Added an encrypted credentials layer (AES-256-GCM, masked by default, every reveal audit-logged) so license keys and passwords stop living in plain text
- Iterated against real client requirements: consolidated an early wiki-shaped model and a bolted-on asset table into one unified Collections + Items structure once real usage showed the split was more confusing than useful
Outcome
- In active use by the client's IT department, tracking real servers and infrastructure
- Every credential reveal audit-logged; every record change diffed and kept in revision history
- Deployed on Vercel + Neon Postgres, with a live English/German UI toggle