From a dropped file to a search result.
Mia Documents is two halves: an automated intake pipeline that runs on the data machine, and this web app that lets the team search everything it produces. Here's the whole journey.
1. Drop files on the machine
The process starts the moment a file lands in the intake folder on the data PC. Scanned paper, PDFs, exported statements, letters — anything goes in, with no naming convention or manual sorting required. You don't upload anything through this website; the machine watches the folder and picks new files up automatically.
2. Read and extracted
Every file is opened and its full text is extracted — including text pulled from scans and images. That raw text is what makes a document findable later, so nothing is thrown away: the complete body of each document is preserved.
3. Structured automatically
On top of the raw text, the pipeline extracts the structured details that make a document useful to filter on: a short summary, the document type, its date, the counterparty, any amounts and currency, a reference number, the people mentioned, and a set of tags. It also records which entity the document belongs to — Voyage Manager, Personal, Blackwave, Mia Bazo, Mia IP, Cruises With Friends, All Aboard and more.
4. Loaded into Azure SQL
The text and structured fields are written to a secure Azure SQL database and added to a full-text index. Indexing happens as documents arrive, so a file is searchable within moments of being processed — there's no nightly batch to wait for.
5. Searchable from anywhere
Sign in to this site and you're searching the whole archive. Type precise keywords, or switch to natural-language mode and ask in plain English. Matches are ranked by relevance across the document body, summary, filename, counterparty and tags. Narrow things down with structured filters — entity, document type, date range, counterparty, amount — and page through the results.
Who can see it
The archive is private. Only signed-in members of the team can search it, using the same accounts that run the rest of the Mia platform — no separate password to remember, and no public access to the documents themselves.