Skip to contents

Tensorlake Setup Guide

This guide will help you get started with Tensorlake, the recommended OCR provider for ohseer.

Why Tensorlake?

Tensorlake offers the best accuracy for document parsing: - 91.7% accuracy for structured data extraction - 86.8% accuracy for table structure preservation - No 5 MB file size limit (unlike AWS Textract synchronous API) - Simple API key authentication (no IAM setup required) - Competitive pricing at $0.01 per page

Accuracy Comparison

Service Accuracy File Limit Auth Pricing
Tensorlake 91.7% None API Key $0.01/page
AWS Textract 88.4% 5 MB (sync) IAM Keys $1.50/1000 pages
Google Doc AI ~89% 20 MB OAuth2 $1.50/1000 pages

Getting Started

1. Sign Up

Visit https://cloud.tensorlake.ai and create an account.

2. Get Your API Key

  1. Log into your Tensorlake dashboard
  2. Navigate to API Keys section
  3. Click “Create New API Key”
  4. Copy your API key

3. Set Environment Variable

Option 1: In R Session

Sys.setenv(TENSORLAKE_API_KEY = "your-api-key-here")

Option 2: In .env File (Recommended)

Create a .env file in your project directory:

TENSORLAKE_API_KEY=your-api-key-here

⚠️ Security: Never commit .env files to version control! Add .env to your .gitignore.

4. Test Your Setup

library(ohseer)

# Process a document
result <- tensorlake_ocr("path/to/document.pdf")

# Check the result
str(result, max.level = 2)

Usage Examples

Basic Document Parsing

# Parse entire document
result <- tensorlake_ocr("document.pdf")

# Parse specific pages
result <- tensorlake_ocr("document.pdf", pages = "1-5")

# Save output to file
result <- tensorlake_ocr("document.pdf", output_file = "result.json")

Working with Results

# Access parsed content
text <- result$result$content

# Access metadata
metadata <- result$metadata

# Check parse status
status <- result$status  # Should be "completed"

Advanced Options

Custom Wait Time

By default, tensorlake_ocr() waits up to 60 seconds for parsing to complete:

# Wait up to 120 seconds
result <- tensorlake_ocr("large-document.pdf", max_wait_seconds = 120)

Poll Interval

Adjust how frequently to check parse status:

# Check every 5 seconds instead of default 2
result <- tensorlake_ocr("document.pdf", poll_interval = 5)

Low-Level API Access

For more control, use the low-level functions:

# Submit parse job
parse_response <- tensorlake_parse_document(
  file_path = "document.pdf",
  tensorlake_api_key = Sys.getenv("TENSORLAKE_API_KEY"),
  pages = "1-10"
)

parse_id <- parse_response$parse_id

# Wait a bit...
Sys.sleep(5)

# Get results
result <- tensorlake_get_parse_result(
  parse_id = parse_id,
  tensorlake_api_key = Sys.getenv("TENSORLAKE_API_KEY")
)

Supported File Formats

Tensorlake supports: - PDF - Microsoft Word (DOCX, DOC) - Microsoft PowerPoint (PPTX, PPT) - Microsoft Excel (XLSX, XLS) - Images (PNG, JPEG, TIFF) - HTML - Plain text (TXT) - CSV

Pricing

  • Document Parsing: $0.01 per page
  • No additional fees for storage or bandwidth
  • Pay only for what you process

For enterprise or on-premises deployments, contact Tensorlake support.

Troubleshooting

API Key Not Found

Error: Tensorlake API key not found

Solution: Set the TENSORLAKE_API_KEY environment variable:

Sys.setenv(TENSORLAKE_API_KEY = "your-key")

Timeout Errors

Error: Parse job timed out after 60 seconds

Solution: Increase max_wait_seconds:

result <- tensorlake_ocr("document.pdf", max_wait_seconds = 120)

Parse Job Failed

Error: Parse job failed

Solution: Check the error details in the result:

result$error

Common causes: - Unsupported file format - Corrupted document - Invalid API key

Support