feat(sandbox): add Tensorlake sandbox backend

[cooleel]

Summary

Adds a Tensorlake-backed sandbox provider underagents.extensions.sandbox.tensorlake, following the existing extension pattern used by the other cloud backends. The tensorlake SDK is imported lazily so users without the extra can still import the package.

Highlights

• Installable via uv sync --extra tensorlake (or pip install "openai-agents[tensorlake]").
• pause_on_exit for suspend/resume of named sandboxes.
• Two workspace persistence modes: "tar"and "snapshot".
• Additive change — no existing public APIs are affected.

Includes a docs reference page, a manual runner example at examples/sandbox/extensions/tensorlake_runner.py, and a unit test suite.

Demo

import asyncio
import sys
from pathlib import Path

from agents import ModelSettings, Runner
from agents.extensions.sandbox import (
  DEFAULT_TENSORLAKE_WORKSPACE_ROOT,
  TensorlakeSandboxClient,
  TensorlakeSandboxClientOptions,
)
from agents.run import RunConfig
from agents.sandbox import Manifest, SandboxAgent, SandboxRunConfig
from agents.sandbox.entries import File

if __package__ is None or __package__ == "":
  sys.path.insert(0, str(Path(__file__).resolve().parents[3]))

from examples.sandbox.misc.workspace_shell import WorkspaceShellCapability 


async def main() -> None:
  manifest = Manifest(
      root=DEFAULT_TENSORLAKE_WORKSPACE_ROOT,
      entries={
          "README.md": File(content=b"# Tensorlake demo workspace\n"),
          "todo.md": File(content=b"1. Inspect files.\n2. Summarize.\n"),
      },
  )

  agent = SandboxAgent(
      name="Tensorlake Demo",
      model="gpt-5.5",
      instructions="Inspect the workspace and summarize it in two sentences.",
      default_manifest=manifest,
      capabilities=[WorkspaceShellCapability()],
      model_settings=ModelSettings(tool_choice="required"),
  )

  run_config = RunConfig(
      sandbox=SandboxRunConfig(
          client=TensorlakeSandboxClient(),
          options=TensorlakeSandboxClientOptions(
              workspace_persistence="snapshot",
              timeout_secs=300,
          ),
      ),
      workflow_name="Tensorlake demo",
  )

  result = await Runner.run(agent, "Summarize the workspace.", run_config=run_config)
  print(result.final_output)


if __name__ == "__main__":
  asyncio.run(main())

Run it with OPENAI_API_KEY and TENSORLAKE_API_KEY exported.

Test plan

• make lint, make format, make typecheck, make tests pass locally.
• Manual validation via examples/sandbox/extensions/tensorlake_runner.py
  against a live Tensorlake account.

Checks

• I've added new tests (if relevant)
• I've added/updated the relevant documentation
• I've run make lint and make format
• I've made sure tests pass