PDF Ingestion

Upload and process documents directly inside Snowflake.

Use this page when you want to add new PDF or ZIP files to your dataset.

---

Upload & Processing Interface

Upload files and select the appropriate processing mode.

Supported formats:

• Single PDF file
• ZIP archive containing multiple PDFs

---

Processing Modes

Choose a mode based on document quality and structure.

Mode	Best For	Use When
Simple ($)	Clean digital PDFs	Text is selectable and well-formatted
Moderate ($$)	Mixed layouts	Tables, headers, multi-column text
Complex ($$$)	Scanned / noisy PDFs	Images, poor OCR, complex layouts

Start with Simple. Upgrade only if extraction quality is poor.

---

File Upload Options

Single PDF

Use when uploading one document.

ZIP File

Use when uploading multiple PDFs in batch.

Each PDF inside the ZIP is processed separately.

Limit: 200MB per file

---

Requirements

Before uploading, ensure:

• CREATE TABLE permissions on your target schema
• Sufficient data access grants if the tables are already created as outlined in access grants.

---

Large File Processing

Staging from Local Environment

For files (PDF/Zip) greater than 200MB, you can process them directly through SnowSQL staging:

1. Stage the file using SnowSQL:

   PUT file://path/to/your/large-file @my_stage;

2. Process staged files through the ZettaQuant application interface

Note: For detailed information on staging files from your local environment, refer to the Snowflake documentation on PUT command.

Best Practices for Large Files:

• Compress files before staging to reduce transfer time
• Use internal stages for better performance
• Consider splitting very large archives into smaller batches

---

Troubleshooting

• Upload or ingestion fails → Check execution details in Telemetry & Logs

• Permission errors → Confirm required privileges are applied in Data Access Grants

• No output tables created → Verify Data Configuration is completed and tables exist in Snowflake

• Poor text extraction / missing tables → Try switching to Moderate or Complex processing mode

• ZIP file not processed → Ensure the ZIP contains valid .pdf files only

• Duplicate document warning → File already exists in the dataset and was skipped

• Slow or stalled processing → Verify warehouse size and GPU compute pool status

• PDF preview not loading → Refresh the page and reselect the uploaded file

---

Next Steps

After ingestion:

1. ZQ Classify — Run analysis
2. Telemetry — Monitor processing

---