Feature request: managed Python image for AI data and document agents

Vercel Sandbox currently provides vercel/sandbox/python:3.14, containing Python, pip, venv, and uv. It does not include common packages needed by AI chatbots that analyze attachments and generate downloadable files.

I tested the current image and confirmed that none of these imports are available:

• NumPy
• pandas
• matplotlib
• openpyxl
• reportlab
• python-pptx
• python-docx
• pypdf

Would Vercel consider publishing a managed image such as:

vercel/sandbox/python-agent:3.14

With:

numpy
pandas
matplotlib
openpyxl
reportlab
python-pptx
python-docx
pypdf

Common use cases include:

• Reading attached CSV and Excel files
• Performing calculations and statistical analysis
• Generating Excel workbooks
• Creating charts
• Creating and manipulating PDFs
• Creating PowerPoint and Word documents
• Returning generated files from AI chat applications

Project-owned snapshots and custom VCR images can solve this, but every application must independently install, version, snapshot, and maintain essentially the same baseline dependencies.

A Vercel-managed image would provide:

• Faster startup without runtime package installation
• A documented and reproducible dependency baseline
• Centralized package and security updates
• Less setup for common AI code-interpreter applications

If a managed image is considered too broad, an official example image or Dockerfile under images/ would also be useful.

I can contribute an initial Dockerfile if this direction fits the project.

Start by inspecting the existing Python 3.14 image setup and the images/ directory mentioned in the issue. Define whether the project should publish vercel/sandbox/python-agent:3.14 or an official example Dockerfile, then verify that the proposed image includes the listed Python packages and has a documented, reproducible update path.