Self-hosted XLSX upload indexes ZIP/OpenXML bytes instead of spreadsheet cell text

[Konan69]

Summary

Self-hosted supermemory-server v0.0.3 accepts an .xlsx upload and eventually marks the document done, but the stored chunks are raw ZIP/OpenXML bytes (PK\u0003\u0004...xl/worksheets/sheet1.xml...) rather than extracted spreadsheet cell text. Search then returns binary/OpenXML fragments instead of useful spreadsheet content.

Environment

• Release: server-v0.0.3
• Binary: supermemory-server-linux-x64
• OS: Linux/WSL
• Upload endpoint: POST /v3/documents/file
• File MIME: application/vnd.openxmlformats-officedocument.spreadsheetml.sheet

Reproduction

Upload an Excel workbook:

curl -sS -X POST http://127.0.0.1:6767/v3/documents/file \
  -H 'Authorization: Bearer <api-key>' \
  -F 'file=@MAG Agent Data.xlsx;type=application/vnd.openxmlformats-officedocument.spreadsheetml.sheet' \
  -F 'containerTag=debug:xlsx' \
  -F 'customId=debug-xlsx-1'

The document is accepted and later reports done:

{
  "id": "E63cFP2EuiVgMefznAxADD",
  "status": "done",
  "title": "MAG Agent Data.xlsx",
  "type": "text",
  "metadata": {
    "mimeType": "application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
  }
}

But GET /v3/documents/{id}/chunks returns chunks like this:

{
  "position": 0,
  "content": "PK\u0003\u0004...xl/worksheets/sheet1.xml..."
}

and later chunks include more OpenXML/ZIP internals instead of cell values:

xl/sharedStrings.xml
xl/styles.xml
xl/workbook.xml
[Content_Types].xml

Search also returns those binary/OpenXML chunks instead of spreadsheet rows/cells.

Suspected cause

The self-hosted content detection appears to classify Office MIME types as generic text:

application/vnd.openxmlformats-officedocument.spreadsheetml.sheet -> text

Then the text extractor decodes the raw .xlsx ZIP bytes as text and embeds that, rather than using an XLSX/OpenXML extractor to unpack worksheets/shared strings and serialize cells.

Docs ambiguity

The docs are inconsistent:

• Supported Content Types says Microsoft Office Excel .xlsx is supported with content type xlsx.
• Upload Files / file upload docs list spreadsheets as CSV / Google Sheets, but not XLSX.

Either way, marking an uploaded .xlsx as done while indexing ZIP bytes is misleading. If XLSX is unsupported for file upload/self-hosted, the document should fail with a clear unsupported-content error instead of indexing binary content.

Expected behavior

One of:

1. .xlsx uploads extract worksheet cell text into searchable chunks, preserving useful sheet/row context; or
2. .xlsx uploads are rejected/marked failed with an explicit unsupported file type error.

Suggested fix

Add an XLSX/OpenXML extractor for uploaded files, or treat Excel MIME/extension as unsupported rather than text. A regression test could upload a minimal workbook with a unique cell value and assert that GET /v3/documents/{id}/chunks or document search contains that cell value, not PK ZIP bytes.

[linear-code]

ENG-888