Question/Feature Request: Exploring ways to pass variables between host and sandbox environment

[amn-max]

Description
I'm trying to understand if there's a way to pass variables, especially pandas DataFrames, from the host environment to the sandboxed container when using PythonContainerREPL. If this isn't currently possible, it would be a valuable feature to have.

Use Case Example

# Host environment
import pandas as pd
df = pd.DataFrame({'A': [1, 2, 3], 'B': ['a', 'b', 'c']})

# Need to use this DataFrame in sandbox
repl = PythonContainerREPL(port=7121)
# Ideally something like:
repl.exec(code, context={'df': df})  # Currently not possible?

Current Challenges

• Need to pass initial state/context from host environment
• Share DataFrames and other variables between host and sandbox
• Maintain compatibility with existing code that relies on shared state

Potential Solution Ideas

1. Direct variable transfer methods:

   repl.set_variables({'df': df})  # Send variables to sandbox
   repl.get_variables(['result_df'])  # Retrieve variables from sandbox

2. File-based transfer for larger objects:

   # Host
   df.to_parquet('temp.parquet')
   repl.exec('df = pd.read_parquet("temp.parquet")')

3. Network-based transfer protocol

Questions

1. Is there currently any way to achieve this or any workarounds being used?

Would appreciate any insights or guidance on this topic!