cerver 0.1.0

Python SDK for Cerver - Sandboxed code execution for AI agents

Cerver Python SDK

Sandboxed code execution for AI agents.

Installation

pip install cerver

Quick Start

from cerver import Sandbox

# Set your API key
# export CERVER_API_KEY=sk_live_...

with Sandbox() as sandbox:
    result = sandbox.run("print('Hello from sandbox!')")
    print(result.output)  # Hello from sandbox!

Usage

Basic Execution

from cerver import Sandbox

with Sandbox() as sandbox:
    # Run code
    result = sandbox.run("x = 1 + 1; print(x)")
    print(result.output)  # 2
    print(result.exit_code)  # 0
    print(result.duration)  # execution time in ms

State Persistence

Variables persist across runs in the same sandbox:

with Sandbox() as sandbox:
    sandbox.run("x = 10")
    sandbox.run("y = 20")
    result = sandbox.run("print(x + y)")
    print(result.output)  # 30

Error Handling

from cerver import Sandbox, CerverError, TimeoutError

try:
    with Sandbox() as sandbox:
        result = sandbox.run("print(undefined_var)")
        if result.exit_code != 0:
            print(f"Error: {result.error}")
except TimeoutError:
    print("Code timed out")
except CerverError as e:
    print(f"Error: {e}")

Custom Configuration

sandbox = Sandbox(
    api_key="sk_live_...",  # Or use CERVER_API_KEY env var
    timeout=60  # Default timeout in seconds
)

API Reference

Sandbox

• run(code, timeout=None) - Execute Python code
• close() - Close the sandbox

ExecutionResult

• output - stdout as string
• error - stderr as string
• exit_code - Process exit code (0 = success)
• duration - Execution time in milliseconds

License

MIT