Every document you upload to Relm goes through an indexing pipeline before it's available to chat or downstream agents. The status pill on each document tells you where the document is in that pipeline.
The states
- Uploaded — the file is on disk, not yet processed.
- Processing — extraction, OCR, chunking, embedding in flight.
- Indexed — done, fully available to chat and agents.
- Indexed-with-warnings — done, but some pages or sheets had issues.
- Failed — couldn't process; click for details.
Most documents move from Uploaded → Processing → Indexed in 30–90 seconds. Big PDFs (200+ pages, scanned) can take 2–3 minutes.
What "Indexed-with-warnings" means
The indexer hit an issue on at least one page or sheet but completed enough of the document to be useful. Common warnings:
- OCR low-confidence on certain pages — usually heavily-stamped or watermarked PDFs.
- A sheet in an Excel file is empty.
- A column header was unrecognized.
- A row in a rent roll had no unit number.
The document is still searchable in chat. Click the warning indicator to see the specific issues and decide whether to re-upload a cleaner copy.
What "Failed" means
The pipeline could not extract usable text or structure from the document. Common causes:
- Encrypted / password-protected PDF. Strip the password and re-upload.
- Image-only PDF that's heavily watermarked. OCR can't see through the watermark.
- Very large file (>500 MB) — split and re-upload.
- Corrupted file — try opening it locally first to verify it's valid.
Click "Retry processing" on the failed entry to try again. If it consistently fails, report the issue and attach a redacted copy of the document.
Re-processing
You can re-process any document — useful if you've manually corrected its type (e.g. the auto-detector classified your rent roll as an OM):
- Click the document.
- Open the actions menu.
- Click Re-process.
Re-processing replaces the existing index for that document.
Can I use the document while it's processing
Mostly no — chat won't return results from a not-yet-indexed document. You can read the source file directly in the document viewer while it indexes.
Multiple documents at once
You can upload many documents in one drag — they queue and process in parallel (up to a small concurrency limit per org). The Documents list shows each document's status independently.