Product · CompressAI
CompressAI slims down the document archive of your FileMaker database: scanned PDFs and images are compressed, Outlook e-mails turned into readable PDFs with their attachments extracted, and duplicates detected by checksum.
Everything runs locally on your server — your documents never leave the building.
Why slim down
Over the years, mature FileMaker databases accumulate tens of thousands of attachments: scans at 300–600 dpi, e-mails as opaque .msg blobs, files stored multiple times. That makes backups slow, storage expensive and the archive hard to oversee.
CompressAI cleans up — and is deliberately non-destructive about it.
What CompressAI processes
One HTTP call per file — CompressAI detects the type automatically and picks the right treatment.
High-resolution scans with uncompressed layers are reduced in two stages — with a safety net: if the result is not smaller, the original stays.
Opaque mail blobs become readable PDF documents with sender, recipients, date and subject — and every attachment is extracted as its own file.
Field photos and scanned receipts are brought down to a sensible size — measured up to 99% smaller, without hurting readability.
Every file gets a SHA256 checksum. Identical content is recognised even under a different file name — the basis for a clean archive.
In preparation: OCR text recognition and AI-assisted classification — turning the slimmed-down archive into a searchable one.
Privacy
Contracts, receipts, e-mails — archive content is often highly sensitive. Cloud compressors upload all of it to third-party servers. CompressAI runs entirely on your own machine.
The service listens locally on your server only — no external access, no uploads, no third-party APIs.
Where no data leaves the building, the privacy assessment stays simple: no US services, no transfers, no consent chains.
E-mail originals go into a local archive, neatly sorted by year and month. Duplicates are only flagged — nothing is lost.
A SHA256 checksum per file, a processing footer in every e-mail PDF, structured logs with sizes and durations.
Honestly: already optimised PDFs barely shrink — they simply remain unchanged.
Included in the service
CompressAI is a lean background service next to your FileMaker Server — these building blocks are included as standard.
two-stage via Ghostscript, with a safety net
.msg and .eml → readable PDF plus attachments
downsize and re-encode, never worse
SHA256 by content, in record context
signature logos and tracking pixels sorted out
original mails filed by year and month
one HTTP call per file, JSON back
every run logged with size and duration
Built for large collections: the pipeline was designed for migrating an archive of roughly 158,000 attachments — in stages, with a clear status per record.
Technology
CompressAI is a lean local service (Python/FastAPI) next to your FileMaker Server: FileMaker exports a file, calls the service via HTTP and receives structured JSON back — compressed file, checksum, extracted attachments.
Compression is handled by proven tools such as Ghostscript — deterministic and predictable. In preparation: OCR text recognition and AI-assisted classification, so the slimmed-down archive becomes searchable as well.
No. CompressAI runs entirely on your own server and listens locally only. There are no uploads, no external APIs and no cloud dependency.
The system is deliberately non-destructive: if a file does not get smaller, the original stays untouched. E-mail originals are archived, duplicates only flagged — deletion happens only on your decision.
It depends on the source material. Measured: a 59 MB scan shrank by 92%, a 19 MB photo by 99%. Already optimised PDFs, on the other hand, barely shrink — and then simply remain unchanged.
The ready-made integration currently exists for FileMaker (via scripting and HTTP calls). The service itself is an open HTTP interface — other systems can be connected; talk to us.
Scanned PDFs remain images — OCR text recognition is planned as the next expansion step. Today, CompressAI delivers readable e-mail PDFs and much smaller scans.
The compression itself is deliberately deterministic — proven tools, predictable results, no model ever sees your documents. The AI expansion (OCR, automatic classification) is in preparation; CompressAI itself is developed AI-assisted throughout.
Getting started & contact
CompressAI is currently being built in production use, migrating an archive of roughly 158,000 attachments — and it can be adapted to your FileMaker installation. Tell us about your database, and we will tell you honestly what slimming it down would achieve.
CompressAI is a software product by Pabeschitz Computer Consulting GmbH. Compression results depend on the source material; processing happens entirely locally on your infrastructure.
