Sandbox: Pydantic MCP (Local)

PydanticExecutorSandbox runs code locally inside a WebAssembly sandbox powered by Pyodide. It communicates with a Pydantic MCP server over stdin/stdout using JSON-RPC.

Prerequisites

• Deno installed and available on PATH

# Install Deno
curl -fsSL https://deno.land/install.sh | sh

Setup

from gllm_tools.code_interpreter.code_sandbox.pydantic_executor_sandbox import PydanticExecutorSandbox

# Uses 'deno' from PATH by default
sandbox = PydanticExecutorSandbox()

# Custom path to deno executable
sandbox = PydanticExecutorSandbox(mcp_server_path="/usr/local/bin/deno")

Note: Unlike the other sandboxes, PydanticExecutorSandbox is synchronous in its constructor — the MCP server process is started immediately on instantiation.

Constructor Parameters

Parameter	Type	Default	Description
mcp_server_path	str | None	None	Path to the deno executable. Defaults to "deno" (from PATH).
language	str	"python"	Programming language (only Python supported)

Execution Constraints

Constraint	Value
Max timeout	300 seconds
Max output length	10,000 characters
File download	Not supported (returns None)
Persistent state	No — each execute_code call is isolated

Full Example

import asyncio
from gllm_tools.code_interpreter.code_sandbox.pydantic_executor_sandbox import PydanticExecutorSandbox

async def main():
    sandbox = PydanticExecutorSandbox()

    try:
        result = await sandbox.execute_code("""
def factorial(n):
    return 1 if n <= 1 else n * factorial(n - 1)

for i in range(1, 8):
    print(f"{i}! = {factorial(i)}")
""", timeout=30)

        if result.status.value == "success":
            print(result.stdout)
        else:
            print(f"Error: {result.error}")

    finally:
        await sandbox.terminate()

asyncio.run(main())

With SandboxCodeInterpreter

import asyncio
from gllm_tools.code_interpreter.code_sandbox.pydantic_executor_sandbox import PydanticExecutorSandbox
from gllm_tools.code_interpreter.code_interpreter.sandbox_code_interpreter import SandboxCodeInterpreter

async def main():
    sandbox = PydanticExecutorSandbox()
    interpreter = SandboxCodeInterpreter(sandbox=sandbox, lm_invoker=lm_invoker)

    result = await interpreter.execute("Write a function to check if a number is prime and test it on numbers 1–20")
    print(result.text)

    await sandbox.terminate()

asyncio.run(main())

Notes

• Each code execution is fully isolated — variables and imports do not persist between calls.

• File upload via files= is supported (content is base64-encoded and written to /files/<filename> inside the sandbox at execution time).

• File download is not supported — download_file() always returns None.

• Runs entirely locally — no cloud calls, no API keys required.