File types

Struxen accepts 20 file extensions for documents. What it does with a file depends heavily on the type: PDFs are processed and become searchable, images are looked at but not read for text, and Office and text files are stored without being parsed. Different upload surfaces accept different subsets.

This page is the authoritative list. If another page or a dropzone hint disagrees with it, this page is right.

The accepted extension list

These are the extensions the Documents upload path accepts. Anything else is refused before the transfer starts.

.pdf
.png  .jpg  .jpeg  .gif  .tif  .tiff  .bmp  .webp  .heic  .heif
.doc  .docx  .xls  .xlsx  .ppt  .pptx
.txt  .csv  .rtf

The check is on the extension, not on the file's contents. Renaming a .dwg to .pdf gets it past the dropzone and produces a document that never finishes processing, so do not do it.

What actually happens per type

Category	Extensions	What Struxen does
PDF	.pdf	Stored, rendered in the viewer, versioned, downloadable. Text is pulled from the PDF's own text layer, page 1 is read by a vision model, the document is classified, and the text is split into chunks and indexed so STRUX can search it and cite it.
Images	.png .jpg .jpeg .gif .tif .tiff .bmp .webp .heic .heif	Stored, rendered, versioned, downloadable. Read by a vision model for classification. No text extraction runs on an image, so text visible in the picture does not enter the search index and STRUX cannot cite it.
Office	.doc .docx .xls .xlsx .ppt .pptx	Accepted and stored. Struxen bundles no Word, Excel, or PowerPoint parser. The file uploads, versions, and downloads exactly as you sent it. Do not rely on its contents being searchable.
Text	.txt .csv .rtf	Accepted and stored. Same as Office: the file is kept and downloadable, and its contents are not something STRUX can be relied on to search or cite.

The short version: if you want STRUX to be able to answer from a document, upload it as a PDF with a real text layer.

How much of a file gets read

Text extraction does not read every page.

Document	Pages read for text
Specifications	First 100 pages
Everything else, including drawings	First 10 pages
Vision pass (all types)	Page 1 only

A drawing sheet is one page, so the 10-page cap is not a constraint on drawings in practice. It does constrain a long report uploaded as a single PDF: split it, or upload the section you need.

Limits differ by where you upload

The same file can be accepted in one place and refused in another. There is no single platform-wide upload rule.

Where	Types accepted	Size cap	Notes
Documents (Add documents)	The full 20-extension list above	500 MB per file	50 files per batch, 3 uploading at once. See Limits.
Drawing sets	.pdf only	No size check on this surface	Refused with "Drawing sets must be PDF files."
STRUX composer and mobile attachments	.pdf .jpg .jpeg .png .tiff .tif	100 MB per file	A stricter list than Documents. Office files are refused here.
Photos	Image files only	64 MB per photo	200 per batch. HEIC and HEIF are accepted.
Punch item attachments	Any file attaches	Per-item attachment cap applies	Only image/jpeg, image/png, image/heic, image/heif, and image/webp are treated as photos. Anything else attaches as a generic file rather than being rejected.
Daily log voice memos	Audio recordings (.m4a .mp3 .mp4 .aac .wav .caf .amr .ogg .opus .webm .3gp)	128 MB per memo	A memo sent for transcription is capped at 32 MB.
Autodesk import	.pdf only	Set by Autodesk	250 files per import. See Integrations.

Not supported

These formats are not accepted, and no conversion happens behind the scenes.

Format	Why	What to do instead
DWG, DXF	Struxen does not read CAD files natively. They are not on the accepted list and will be refused at the dropzone.	Publish the sheets to PDF from your CAD tool and upload the PDF set on the Drawings page.
RVT (Revit)	Not read natively and not on the accepted list.	Export the sheets to PDF and upload the PDF set.
IFC, NWD, NWC and other model formats	Not on the accepted list. There is no model viewer or model ingestion path.	Export drawings to PDF. Coordinate the model in your BIM tool.
Scanned-image-only PDFs	A PDF with no text layer, produced by a scanner or a photo of a drawing, uploads and stores fine, but text extraction returns nothing from it. Struxen does not run OCR.	Re-export the sheet from the source file so it carries a text layer, or run your own OCR on the file before uploading it.
Video	Not on any accepted list.	Not supported.

Scanned PDFs are the one that catches people out, because the upload succeeds and the document reaches Indexed. Indexed means the pipeline finished, not that it found text. If STRUX cannot answer from a document you can see is Indexed, check whether it is a scan.

Selecting files

The uploader takes files, not folders. Select or drag the files themselves. There is no folder upload, and nothing is derived from a folder name.

When a file is refused

Uploads are screened in the browser before any bytes move, so a bad file costs you nothing. Three of those checks are repeated on the server, which means a direct API call gets the same answer.

Server-enforced, and reported by both the browser and the API:

Code	What you see	Cause
UNSUPPORTED_TYPE	… is not a file type Struxen can read	The extension is not on the accepted list.
EMPTY_FILE	… is empty	The file is zero bytes.
TOO_LARGE	… is over the 500 MB limit	The file exceeds the per-file cap.

Browser-only, so they guard the upload card rather than the API:

Code	Cause
BATCH_LIMIT	More than 50 files were dropped at once. Files past the cap are named, not silently dropped.
DEMO_READ_ONLY	Demo projects accept no uploads.
NO_PROJECT	No project is selected.
UPLOAD_FAILED	The transfer itself failed.

One more refusal comes from the server regardless of the file:

• Upload not allowed, on a project that mirrors Autodesk. The project's documents come from Autodesk. Upload the file there and let it sync.

Related

• Limits for every size and count cap in one table
• Uploading documents for the upload workflow
• Processing for what happens after the upload
• Troubleshooting if a file uploaded but did not behave the way you expected