Skip to content

google-vision-setup.md

Latest commit

 

History

History
169 lines (120 loc) · 4.53 KB

google-vision-setup.md

File metadata and controls

169 lines (120 loc) · 4.53 KB

Google Cloud Vision Setup

Google Cloud Vision provides far superior handwriting recognition compared to Tesseract. Unless your handwriting is exceptionally clear and print-like, we strongly recommend using Google Vision.

Why Google Vision?

Feature Google Vision Tesseract
Handwriting ✅ Excellent ❌ Poor
Cursive ✅ Handles well ❌ Fails
Mixed content ✅ Text + drawings ❌ Confused
Languages ✅ Auto-detect Manual config
Setup API key System install

Quick Setup (API Key)

The easiest way to get started:

1. Create a Google Cloud Project

  1. Go to Google Cloud Console
  2. Click Select a project → New Project
  3. Name it (e.g., "remarkable-ocr") and click Create

2. Enable the Vision API

  1. Go to Cloud Vision API
  2. Click Enable

3. Create an API Key

  1. Go to Credentials
  2. Click Create Credentials → API Key
  3. Copy the key

4. (Optional) Restrict the Key

For security, restrict the API key:

  1. Click on the API key you just created
  2. Under API restrictions, select Restrict key
  3. Select only Cloud Vision API
  4. Click Save

5. Configure MCP

Add the key to your MCP configuration:

{
  "servers": {
    "remarkable": {
      "command": "uvx",
      "args": ["remarkable-mcp", "--ssh"],
      "env": {
        "GOOGLE_VISION_API_KEY": "your-api-key-here"
      }
    }
  }
}

Cost

Google Cloud Vision pricing:

Tier Price
First 1,000 requests/month Free
1,001 - 5,000,000 requests $1.50 per 1,000

For personal use, you'll likely stay within the free tier. Each notebook page counts as one request when OCR is enabled.

Alternative: Service Account Credentials

For production use or tighter security controls:

1. Create a Service Account

  1. Go to Service Accounts
  2. Click Create Service Account
  3. Name it and click Create
  4. Grant the role Cloud Vision API User
  5. Click Done

2. Download Credentials

  1. Click on the service account
  2. Go to Keys tab
  3. Click Add Key → Create new key
  4. Select JSON and click Create
  5. Save the downloaded file securely

3. Configure MCP

Point to your credentials file:

{
  "servers": {
    "remarkable": {
      "command": "uvx",
      "args": ["remarkable-mcp", "--ssh"],
      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/path/to/your/credentials.json"
      }
    }
  }
}

Or install the SDK with default credentials:

# Install gcloud CLI and authenticate
gcloud auth application-default login

# Install SDK dependency
pip install remarkable-mcp[ocr]

Environment Variables

Variable Description
GOOGLE_VISION_API_KEY API key (simplest setup)
GOOGLE_APPLICATION_CREDENTIALS Path to service account JSON
REMARKABLE_OCR_BACKEND Force backend: auto, google, tesseract

Troubleshooting

"API key not valid"

  • Double-check the key is copied correctly
  • Ensure the Vision API is enabled for your project
  • Check that API restrictions (if any) include Vision API

"Permission denied"

  • For service accounts, verify the Cloud Vision API User role
  • Check that the credentials file path is correct

"Quota exceeded"

  • You've exceeded the free tier
  • Enable billing or wait until next month
  • Consider caching results (remarkable-mcp does this automatically)

OCR Not Running

  • Check remarkable_status() to see which OCR backend is configured
  • Set REMARKABLE_OCR_BACKEND=google to force Google Vision
  • Ensure include_ocr=True when calling remarkable_read()

Performance Tips

  1. Caching: OCR results are cached per document — reading multiple pages doesn't re-run OCR
  2. Selective OCR: Only enable include_ocr when you need handwritten content
  3. Typed text: Notebooks with Type Folio extract text without OCR
  4. PDFs: Text is extracted directly, OCR only needed for scanned documents

Privacy Considerations

When using Google Vision:

  • Your handwritten content is sent to Google's servers for processing
  • Google's Cloud Vision Terms apply
  • Consider this when processing sensitive documents

For fully offline OCR, use Tesseract (with significantly reduced accuracy for handwriting).