File handling
SecondLoop keeps original attachments, then optionally runs enrichment pipelines based on type and settings.
Processing by file type
| File type | What SecondLoop does |
|---|---|
image/* | Stores original image, reads EXIF metadata when available, and can run image caption/OCR enrichment if enabled. |
audio/* | Stores attachment, optionally creates a normalized proxy format, and can run transcription (local / BYOK / Cloud). |
video/* | Can convert to a segmented video manifest (application/x.secondloop.video+json) with preview frames and optional extracted audio; then runs transcript/OCR style enrichment. |
application/pdf | Extracts text layer first; if text is missing, marks it as OCR-needed and can run OCR. |
application/vnd.openxmlformats-officedocument.wordprocessingml.document | Extracts document text; OCR can use embedded media (for scan-like content). |
text/* | Decodes UTF-8 text and normalizes it for search/readability. |
Structured text-like (json/xml/yaml/toml/ini/csv) | Treated as text extraction targets for searchable content. |
URL share manifest (application/x.secondloop.url+json) | Fetches page metadata/readable text with security checks, then stores extracted result. |
| Other binary files | Stored as attachments and available for system open/download, without guaranteed text enrichment. |
PDF and OCR behavior
- PDF text extraction runs first.
- If extracted text is empty, the attachment enters needs OCR state.
- OCR can be run automatically or manually (depending on settings and limits).
Video behavior
- Video processing may create:
- segmented playback proxy
- poster + keyframes
- linked audio payload for transcript
- The viewer can combine transcript and OCR/keyframe text into one readable payload.
Practical notes
- Processing depends on enabled settings (for example OCR/transcription toggles).
- If a format is stored but not fully enriched, you can still keep and sync the original file.