> ## Documentation Index
> Fetch the complete documentation index at: https://docs.recordengine.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Uploading Documents
> Every way to get documents into RecordEngine — web upload, email, watched folder, and API.
RecordEngine accepts documents through four channels. Use whichever fits your workflow — they all result in the same AI processing, confidence scoring, and structured extraction.
***
## Supported File Types
| Category | Formats | Notes |
| ------------------ | ----------------------- | ----------------------------------------------------------------------------- |
| **PDF** | `.pdf` | Standard and scanned — OCR fallback activates automatically for scanned files |
| **Images** | `.jpg`, `.jpeg`, `.png` | Vision AI reads directly from the image |
| **Word documents** | `.docx` | Full text extraction |
| **Spreadsheets** | `.xlsx`, `.csv` | Tabular data extraction |
| **Text files** | `.txt`, `.html` | Full text extraction |
| **Audio** | `.mp3`, `.wav`, `.m4a` | Transcribed locally, then extracted |
For documents with Chinese characters, stamps, or seals — including Fapiao and Business Licenses — RecordEngine uses vision-first extraction rather than text OCR, which handles stamps and handwriting that OCR-only tools miss.
***
## Method 1 — Web Upload (UI)
The most common upload method for individual documents or small batches.
Click **Upload** in the left sidebar.
Drag and drop files onto the upload zone, or click **Browse** to select from your file system. You can select multiple files at once — each becomes a separate document.
Set:
* **Contact** — the client or vendor this document belongs to
* **Folder** — the folder within that contact (e.g. Q4 Invoices, Documents, Review)
Choose the profile that matches your document type. If the contact has a default profile set, it's pre-selected automatically.
Click **Upload**. Documents appear in the list immediately with **New** status, then transition to **Processing** automatically.
***
## Method 2 — Email Intake
Send any file as an email attachment to your RecordEngine intake address. No login required — documents land directly in the System → Inbox folder and process automatically.
Best for:
* Clients or vendors submitting documents without access to RecordEngine
* Forwarding attachments from other email threads
* Mobile capture (photograph a receipt, email it)
Find your intake address at **Settings → Email Intake**.
See the [Email Intake](/user-guide/email-intake) guide for full details.
***
## Method 3 — Hot Folder (Watch Folder)
RecordEngine monitors a designated folder on the server's file system. Any file copied into that folder is automatically ingested — no UI interaction needed.
Best for:
* Bulk imports from a scanner or MFP (multi-function printer)
* Automated file drops from other software
* Legacy systems that export files to a directory
The hot folder is configured in your `docker-compose.yml` and is active whenever the `xr-watcher` container is running. Contact your administrator for the hot folder path on your server.
***
## Method 4 — REST API
Upload programmatically from any external system using the API. Ideal for automated pipelines where documents are generated or received by software — for example, uploading attachments received in a CRM, or exporting reports from an ERP.
```bash theme={null}
curl -X POST https://YOUR-INSTANCE/api/documents/upload \
-H "Authorization: Bearer YOUR_TOKEN" \
-F "file=@invoice.pdf" \
-F "contact_id=5" \
-F "folder_id=12" \
-F "profile_id=2"
```
See the [Documents API](/api-reference/documents) for the full request schema and code examples.
***
## Duplicate Detection
RecordEngine checks for duplicate filenames within the same folder on upload. If a file with the same name already exists in the target folder, you're prompted to either:
* **Upload as a new version** — links the new file to the existing document as version 2
* **Upload as a separate document** — creates an independent new document
See [Document Versioning](/user-guide/document-versioning) for how versioning works.
***
## Processing Time
Once uploaded, documents transition through statuses automatically:
| Status | Typical duration |
| ---------------- | ------------------------------------------------- |
| **New** | A few seconds — queued waiting for AI worker |
| **Processing** | 10–45 seconds for clean PDFs and images |
| **Processing** | 1–3 minutes for audio files (transcription first) |
| **Processing** | 2–5 minutes for large multi-page PDFs |
| **Needs Review** | Ready for human review |
The very first document after a server restart takes longer — typically 2–5 minutes — because the AI model loads into GPU memory. Every subsequent document processes much faster. Run a warmup document after any server restart to pre-load the model.
***
## Uploading Multiple Files
You can upload multiple files in a single web upload operation:
1. On the Upload page, select multiple files at once using Ctrl+Click or Shift+Click in the file picker
2. All selected files share the same contact, folder, and profile assignment
3. Each file becomes an independent document and processes in parallel
For large batches (50+ files), the hot folder method is more reliable than web upload.
***
## What Happens After Upload
Once a file is uploaded, RecordEngine automatically:
1. Stores the original file in the document archive
2. Detects the file type and chooses the appropriate extraction method (vision, text, audio transcription)
3. Runs the AI extraction against the selected profile
4. Generates an AI summary
5. Calculates a confidence score
6. Evaluates any matching Rules Engine rules
7. Sets status to **Needs Review**
The original file is always preserved — it is never modified or deleted by processing.