Project Archive Search
The files never move - a layer of memory is added on top
Turning more than a decade of project data on a company NAS from a pile of folders into something that answers in a few keystrokes. The files themselves stay exactly where they are.
- Type
- Internal file management
- Scope
- Index architecture
- Search engine
- Scanner
- Interface
- Industry
- In-house operations

The challenge
Given enough years, accumulated project data turns into a liability. Everyone knows the file is in some folder, but finding a particular file, for a particular client, from a particular year is close to impossible. Operating system search is poor and there is no way to tag or annotate a project. Worse, any proposal that moves files into a system gets rejected outright - nobody gambles fifteen years of work.
What we built
The files never move
This is the foundation of the whole design. If the system breaks, gets switched off, or is abandoned entirely, every file must still be sitting intact on the NAS and openable in a file browser. Any proposal requiring files to move is blocked first and discussed second.
The importer may read but never write
The scanner never moves, renames or deletes an existing file. Give a system like this write access and a mistake becomes irreversible.
Feasibility proven before building
Two technical risks were tested first - whether a browser can open a file on the NAS directly, and whether the search engine segments Traditional Chinese well enough, including tolerance for typos. Implementation only began once both passed.
Search as you type
Scanning 14,854 projects takes 1.0 second. Under load testing, sixteen queries returned a p95 between 2.6 and 4.4 milliseconds against a 300 millisecond target - roughly sixty to a hundred times the headroom required.
Wondering whether a system like this would fit your company?
No specification needed. Tell us which part of your work takes the most time, and we will look at the right approach together.


