Document AI — On-Premises Document Processing Service
Service Quick Reference
| Field |
Details |
| Service URL |
document.ai.hpc.fau.edu |
| Category |
Research Computing |
| Audience |
FAU Faculty, Staff, and Researchers |
| Authentication |
FAU Microsoft Entra ID (single sign-on) |
| Max File Size |
100 MB per file |
| Supported Inputs |
PDF, DOCX, PPTX, XLSX, HTML, PNG, JPG, TIFF, and more |
| Output Formats |
Markdown (.md), JSON, HTML, Plain Text (.txt), DocTags |
What is Document AI?
-
Document AI is an on-premises document processing service hosted on FAU's High Performance Computing (HPC) cluster. It converts documents into machine-readable formats for downstream use in research workflows, AI pipelines, data extraction, and automated reporting. The service is powered by Docling, an open-source document conversion engine developed by IBM Research, and runs entirely within FAU's infrastructure — no files leave the university network.
The service is available to all FAU faculty, staff, and researchers who authenticate with their FAU Microsoft credentials. No special account request is required.
What You Can Do
- Convert PDFs, Word documents, PowerPoint files, spreadsheets, images, and HTML pages into structured text formats.
- Extract tables, headings, reading order, code blocks, and formulas from complex documents.
- Apply OCR (optical character recognition) to scanned documents and images.
- Process multiple documents at once using the web interface.
- Download results as Markdown, JSON, HTML, plain text, or DocTags.
- Generate API keys to integrate Document AI into your own scripts or pipelines.
- View your processing history and re-download past results.
Before You Begin
Ensure the following before accessing the service:
- You have an active FAU NetID and password.
- You are connected to the internet (VPN is not required from off campus).
- Your browser supports modern JavaScript (Chrome, Firefox, Edge, or Safari — current versions recommended).
Note: Document AI uses FAU's Microsoft Entra ID for authentication. When you visit the service URL, you will be redirected to the FAU single sign-on (SSO) page automatically.
Accessing the Service
- Open a web browser and navigate to document.ai.hpc.fau.edu.
- You will be redirected to the FAU SSO login page. Enter your FAU email address and password.
- Complete multi-factor authentication (MFA) if prompted.
- Once authenticated, you will be taken directly to the Document AI home page.
Note: Your name or email will appear in the top-right corner of the page once you are signed in. If the page shows an error instead of the sign-in prompt, see the Troubleshooting section.
Converting a Document
Step 1 — Choose Your Output Formats
At the top of the page, check one or more output formats:
- Markdown — best for AI/LLM ingestion, documentation pipelines, and readable structured text.
- JSON — full structured representation including page layout and metadata; ideal for programmatic processing.
- HTML — preserves visual structure; useful for web-based workflows.
- Text — plain extracted text with no markup.
- DocTags — Docling's native tagging format for document structure.
Note: Markdown is selected by default. You may select multiple formats; each will be available for download after conversion.
Step 2 — Select an OCR Option
OCR (optical character recognition) extracts text from scanned pages or image-based documents. Choose the option that best fits your document:
- On — Tesseract (default): Recommended for most documents. Uses Tesseract, a widely used open-source OCR engine.
- On — RapidOCR (ONNX): An alternative engine that may perform better on certain Asian-language documents or non-standard fonts.
- Off: Disables OCR. Use this only if your document already contains selectable, machine-readable text and you want to skip OCR processing to save time.
Step 3 — Upload Your Document(s)
- Click the drop zone (the dashed box that reads "Drop documents here") or drag and drop your files directly onto it.
- To upload multiple files at once, drag them all into the drop zone or use the file picker and select multiple files using Ctrl+Click (Windows) or Cmd+Click (Mac).
- The drop zone will confirm how many files are selected.
- Maximum file size: 100 MB per file.
- Supported file types: PDF, DOCX, PPTX, XLSX, HTML, PNG, JPG, TIFF, BMP, and other common document and image formats.
Step 4 — Submit for Conversion
- Click the Convert button.
- Each file is submitted as a separate job. The Jobs table will appear at the bottom of the page with a status badge for each job.
- Status badges update automatically — no need to refresh the page:
- Queued — the job is waiting for a processing slot.
- Processing — conversion is in progress.
- Succeeded — the converted file is ready to download.
- Failed — conversion encountered an error. An error message will appear in the Result column.
Downloading Results
When a job shows the Succeeded status, download links appear in the Download column of the Jobs table.
- Single format: Click the filename link (e.g.,
report.md) to download that format directly.
- Multiple formats: Individual links appear for each format you requested. Click ⬇ all to download all available formats sequentially.
Downloaded files are named after the original document you uploaded (for example, uploading research_paper.pdf and requesting Markdown will produce research_paper.md).
Note: Job history is preserved and reloads automatically when you return to the page. You can re-download results from any previous job at any time without reconverting the document.
Using the API
Document AI exposes a REST API for integration into scripts, automated workflows, and research pipelines. To use the API, you need a personal API key.
Generating an API Key
- Click Manage API Keys in the top-right corner of the page.
- Enter an optional Name to identify the key (for example, "ingest-pipeline" or "my-research-script").
- Optionally enter an expiration in the Expires in (days) field. Leave blank for a non-expiring key.
- Click Generate key.
- Copy the key displayed in the yellow box. This is the only time the full key will be shown. Store it securely (for example, in a password manager or environment variable).
- Click Close when finished.
Important: Each user is limited to one active API key at a time. Generating a new key immediately revokes the previous one. Any scripts or integrations using the old key will stop working.
Making API Requests
Include your API key in the X-API-Key request header. The base URL for all API endpoints is [https://document.ai.hpc.fau.edu](https://document.ai.hpc.fau.edu).
Example — submit a document for conversion (Markdown output):
curl -X POST [https://document.ai.hpc.fau.edu/v1/documents](https://document.ai.hpc.fau.edu/v1/documents) \ -H "X-API-Key: YOUR_API_KEY" \ -F "file=@document.pdf" \ -F "to_formats=md" \ -F "do_ocr=true"
Key API endpoints:
| Endpoint |
Description |
POST /v1/documents |
Submit a document for conversion. Returns a job ID. |
GET /v1/jobs |
List all your jobs and their current status. |
GET /v1/jobs/{id} |
Get the status and details of a specific job. |
GET /v1/jobs/{id}/result |
Download a converted result. Use ?format=md to specify format. |
GET /v1/whoami |
Returns information about your authenticated identity. |
GET /v1/usage |
Returns your document processing usage statistics. |
POST /v1/keys |
Create or rotate your API key (requires browser sign-in). |
DELETE /v1/keys/{id} |
Revoke an API key. |
Note: Full interactive API documentation is available at document.ai.hpc.fau.edu
Troubleshooting
| Symptom |
Likely Cause |
Resolution |
| Page shows "500 Internal Server Error" instead of the sign-in prompt. |
Authentication service may be restarting or temporarily unavailable. |
Wait 1–2 minutes and refresh the page. If the error persists for more than 10 minutes, contact Research Computing. |
| Job stays in "Queued" status for a long time. |
The conversion queue may be busy, or a processing node may be unavailable. |
Wait up to 20 minutes for CPU-based conversions. If the job does not move to "Processing" after that, contact Research Computing with your job ID. |
| Job shows "Failed" with an error message. |
The document may be corrupted, password-protected, or in an unsupported format. |
Verify the file opens correctly on your computer. Try a different output format. Remove any password protection before uploading. If the issue continues, submit a help ticket with the job ID and a description of the file. |
| Download link appears but file fails to save. |
Browser pop-up blocker may be preventing the download. |
Allow pop-ups for document.ai.hpc.fau.edu in your browser settings, then try again. |
| "Invalid API key" error when using the API. |
The key may have been rotated or revoked, or it is being sent in the wrong header. |
Generate a new API key from the web interface. Confirm the header name is X-API-Key (case-insensitive). |
| Signed in but no jobs appear in the Jobs table. |
Expected — jobs are loaded per session. Past jobs reload on page refresh. |
Refresh the page. Your previous jobs will reload automatically. If you still do not see them, you may be signed in with a different account. |
| Output is missing text that appears in the original document. |
The document may contain text as embedded images without selectable text (e.g., a scanned page). |
Re-submit the document with OCR enabled (On — Tesseract). If OCR is already on, try On — RapidOCR as an alternative. |
Data and Privacy
- All files are processed and stored within FAU's on-premises HPC infrastructure. No data is sent to external cloud services.
- Uploaded documents and converted results are stored in FAU's internal storage and are accessible only by the account that submitted them.
- Do not upload documents containing sensitive personal information (FERPA-protected student records, HIPAA-protected health information, or other regulated data) without first consulting with the FAU Office of Information Technology and confirming the appropriate data handling procedures.
- If you have questions about data classification or appropriate use, contact the FAU OIT Security team at security@fau.edu.
Related Resources
Need Further Help?
If you have issues or questions regarding this service, please contact Research Computing or our OIT Help Desk.
Contact our OIT Help Desk