Last production measurement: 29 August 2026All runs used synthetic files in temporary directories on the production host. No customer data existed and no database jobs or result files remained afterwards.
Verified maximum and reference runs
Each run exercised the real worker code inside its isolated, resource-limited container — the same code path a paying job would use. Wall time is the complete job duration; peak RSS is the highest memory the worker process reached.
ToolProduction workloadWall timePeak RAMVerified output
Price List CompareTwo synthetic supplier lists, 100,000 rows each41.3 s929 MiBFull XLSX report and changes CSV
Product Image Batch Editor20 JPEG images, 12 MP, three variants4.2 s900 MiB shared peakAll variants, ZIP and manifest
Target Size Image CompressorTwo 12 MP JPEGs to AVIF under 500 KB8.1 s900 MiB shared peakBoth outputs under the target
Google Merchant Feed Validator100,000 TSV products with 15,000 safe fixes8.8 s211 MiBComplete audit and corrected TSV
Landlord Service Charge Calculator25 units, 100 costs, 20 repeated calculations0.3 s24 MiBDeterministic calculation package
Landlord document intake25 OCR images, 25 text PDFs, 25 CSV, 25 XLSX10.6 s66 MiB100 documents, no temporary files left
Bulk Image Watermark20 JPEG images, 12 MP, tiled rotated text mark7.9 s612 MiB combined peak20 outputs, ZIP and previews verified
Target Size PDF CompressorTwo 150-page PDFs, lossless pass0.1 s612 MiB combined peakBoth outputs under the 5 MB target
PDF Page Organizer300 selected and reversed pages with rotations0.1 s612 MiB combined peakExact 300-page output
PDF Page Numbering300 pages0.6 s612 MiB combined peakFirst and last label verified
Bates Number PDF300 pages0.8 s612 MiB combined peakContinuous labels and CSV manifest
Parallel acceptance run
Price List Compare, both image tools, the Merchant Feed Validator and the rental document pipeline also ran concurrently in their isolated containers. The complete parallel run finished in 47 seconds. All nine long-running services stayed healthy, all four worker queues involved returned to zero and no synthetic upload or result file remained on disk.
How these numbers relate to the published limits
The per-tool limits shown on each product page — for example 100,000 rows for Price List Compare or 20 images per batch — are the workloads these measurements actually verified. Production additionally keeps one active job per worker queue and enforces per-container CPU and memory limits, so a single heavy job cannot slow other tools down.
What this page does not claim
These are capacity measurements on the current host, not a performance guarantee, uptime promise or price basis. Individual jobs vary with file structure and content. Numbers on this page are updated only after a new documented production run.