Skip to main content
Hello from ARGOS Omni. Here is what shipped in September.

​
Download an analysis report as PDF

New Feature You can now download a completed analysis as a PDF report. Use the Download PDF button on the analysis results screen.
PDF analysis report

The downloaded PDF report.

It lays out the score, the summary, and the decision variables as a document, along with the analysis ID, completion time, and profile ID — ready to file as evidence for a review, or to submit as-is. Report generation moved to a dedicated server, which substantially improved the layout and the reliability. Reports are available in Korean and English. To get the same file through the API, use GET /analyses/:analysisId/report.

​
An analysis completes only once its report is ready

Feature Improvement Previously, selecting Download PDF right after an analysis turned complete could tell you the report was still being prepared. Now an analysis is marked complete at the point its report is ready. When the status reads complete, Download PDF works immediately. No waiting and retrying.

​
Report download history

Feature Improvement When a report was generated and when it was downloaded are now recorded, along with generation failures — which makes it far easier to trace what went wrong when something does.

​
The 5-item folder limit is gone

New Feature The limit of 5 items per folder has been removed. A folder now holds as many documents as you need. In its place, one profile can hold 100MB of uploads in total. Splitting them across folders does not change the total, and going over returns 409 OMNI_2005. You no longer have to split folders just because a case involves many documents. Folders are now purely for organizing documents by type.

​
Unsupported formats are rejected at upload

Breaking change Previously any format uploaded successfully and then failed quietly during processing, when no text could be extracted. Now an unsupported format is rejected as you upload it — 400 OMNI_2105. Rejected: HWP and HWPX, the legacy doc, xls, and ppt formats, RAW, ICO, ZIP, and video and audio files. Convert legacy Office documents to docx, xlsx, or pptx first. A file whose extension disagrees with its actual content is rejected under the same code. Renaming a PDF to .md does not get it through.
This cuts down on discovering “the value was never extracted” only after running an analysis. The full supported list is in adding an item.

​
Empty lists now count in Output Completeness

Feature Improvement A list (array) field that came back completely empty barely lowered the Output Completeness part of the Analysis Score. That is fixed. Each item in the output schema now counts as one slot, and items inside a list count by the share that is filled. An empty list scores 0 for that item. A missing required item now shows up in the score. The weights (Step Execution 40 · Step Quality 20 · Output Completeness 40) are unchanged, and the change applies to analyses run after it shipped.

​
The reasoning reaches the same conclusion as the value

Feature Improvement A field’s value and its reasoning could point to opposite conclusions. That is fixed. The reasoning is now built from the same grounds used to set the value — the analysis result and the field definition — so it only ever supports the value shown on screen and in the API.
  • If a field description says what TRUE and FALSE mean, the reasoning is written against that definition.
  • A field with no grounds found does not get a made-up reasoning for the opposite conclusion.
This applies to the Report tab, the decision-variable detail on the Summary tab, and the analysis API.

​
The reasoning follows the workflow’s language

Feature Improvement Field-level reasoning now follows the language the workflow was written in. A workflow written in English gives reasoning in English; one written in Korean gives it in Korean. The language of the documents and the UI language setting do not change this. Previously the reasoning came out in Korean regardless of the workflow’s language. This applies to the Report tab, the decision-variable detail on the Summary tab, and the analysis API. Reasoning on analyses that already completed does not change — reading analysis results.