Skip to content

04_process_binary_files.md

Latest commit

 

History

History
75 lines (62 loc) · 2.22 KB

04_process_binary_files.md

File metadata and controls

75 lines (62 loc) · 2.22 KB
Raw

Example 4: Process Binary Files

When to Use

Read this when you need to upload, process, or download binary files (images, PDFs, executables) in a sandbox.

User Request Pattern

Resize this image in a sandbox
Process this PDF file
Convert this image format
Generate an image/PDF from data

Workflow

Step 1: Validate Environment

cd .claude/skills/agent-sandboxes/sandbox_cli
grep "E2B_API_KEY" ../../../../.env

Step 2: Initialize Sandbox and Capture ID

uv run sbx init
# YOU capture and remember: sandbox_id = "sbx_img456resize"

Step 3: Upload the Binary File

Use sbx files upload for binary files (images, PDFs, executables):

uv run sbx files upload sbx_img456resize ./input.jpg /home/user/input.jpg

Important: Use upload/download for binary files, not write/read (which are for text).

Step 4: Install Processing Tools

Install uv and the required library (e.g., Pillow for images):

uv run sbx exec sbx_img456resize "curl -LsSf https://astral.sh/uv/install.sh | sh" --shell --timeout 120
uv run sbx exec sbx_img456resize "/home/user/.local/bin/uv pip install --system pillow"

Step 5: Write and Run Processing Script

uv run sbx files write sbx_img456resize /home/user/resize.py "from PIL import Image; img = Image.open('/home/user/input.jpg'); img.resize((800,600)).save('/home/user/output.jpg')"
uv run sbx exec sbx_img456resize "python3 /home/user/resize.py"

Step 6: Download the Result

uv run sbx files download sbx_img456resize /home/user/output.jpg ./output.jpg

Step 7: Clean Up

uv run sbx sandbox kill sbx_img456resize

Key Points

  • Use upload/download for binary files (images, PDFs, executables)
  • Use write/read for text files only
  • Upload source files before processing
  • Install required libraries (PIL, reportlab, etc.)
  • Download processed files back to local system
  • Common libraries:
    • Images: pillow (PIL)
    • PDFs: reportlab, pypdf2
    • Data processing: pandas, numpy
  • Always clean up the sandbox when done

Binary File Types Supported

  • Images: jpg, png, gif, bmp, tiff, webp
  • Documents: pdf, docx, xlsx
  • Archives: zip, tar, gz
  • Executables: any binary format