At a glance
| Input | Current action | Output | Boundary |
|---|---|---|---|
| .docx and exact Word MIME type | Validate the allow-listed pair | Accepted Word input | Legacy .doc and macro-enabled .docm are unsupported. |
| DOCX package | Inspect the first local ZIP header, decompress entries, and require word/document.xml | Parseable primary document body | A declared first-entry encryption flag and malformed packages fail. No broader authenticity validation runs. |
| Expanded package | Count entries and bytes | Bounded parser input | Maximum: 20,000 entries and 100 MiB expanded bytes. |
| Primary word/document.xml | Join w:t text within each non-empty w:p | Plain text and paragraphCount | Extracted text cannot exceed 2 MiB. Paragraph breaks become new lines. |
| Approved baseline terms | Count exact, case-insensitive token matches | Revision-specific usage observations | Maximum catalog: 5,000 terms. Each term count is capped at 100,000. Definition fit remains unchecked. |
| File identity and upload context | Record name, extension, MIME type, uploader, capture time, and parser metadata | Artifact provenance and immutable revision | Provenance does not establish document correctness or approval. |
| Source bytes | Calculate SHA-256 and append a content-addressed frozen-content object | Object Store content plus SQL hash and object-key references | Raw bytes are not stored in SQL evidence records. |
| Retention class and Retention until | Validate an allowed class and future timestamp | Retention class, retention date, legal-hold flag, deletion state, and archive state on the artifact record | Upload has no legal-hold control, so the flag is always off. Nothing enforces the date. |
| File up to 350 KiB | Send one direct resolver request | Persisted or duplicate result | Base64 transport remains bounded below Forge request limits. |
| File above 350 KiB and up to 25 MiB | Stage verified 256 KiB chunks | Final result after assembly and whole-file hash verification | Maximum: 100 chunks. Cleanup is attempted and reported after finalization. |
How it works
- Open Enterprise evidence workspace.
- Find Upload controlled evidence.
- Under Evidence file, select the DOCX file.
- Select a Retention class.
- Set Retention until.
- Select Upload evidence.
- Review the returned Artifact, Revision, and Usages values.
- Apply deletion and legal holds in your records system. The app stores the retention date but does not enforce it.
Example
A team uploads Policy.docx. Its primary document body contains the approved term “Risk” twice. The app stores the bytes in Object Store. It creates one immutable revision and one usage observation with mention count two. Uploading the same file again returns a duplicate result.
What this does not mean
Word support is limited to DOCX. Extraction reads paragraph text only from word/document.xml. It does not separately parse headers, footers, comments, footnotes, endnotes, relationships, embedded files, macros, fields, or signatures. It does not interpret tracked-change state or visual layout. Term matching finds literal tokens, not semantic agreement. The retention date is stored on the artifact record, but no workflow deletes or archives evidence when it passes. Uploads made before 6.9.0 have no stored date.
Related questions
Next step
Upload one controlled DOCX file. Review the returned artifact, revision, and usage count before relying on the evidence record.
View on Atlassian Marketplace