Exxec

Read the documentation!

Basic Usage

Use the get_environment() function to create execution environments:

from exxec import get_environment

# Local execution (same process)
env = get_environment("local")

# Subprocess execution (separate process when executing python code)
env = get_environment("local", isolated=True)

# Docker execution (containerized)
env = get_environment("docker")

# Execute code
async with env:
    result = await env.execute("""
    async def main():
        return "Hello from execution environment!"
    """)
    print(result.result)  # "Hello from execution environment!"

Available Providers

Local Provider

Executes code in the same Python process. Fastest option but offers no isolation.

env = get_environment("local", timeout=30.0)

Parameters:

• timeout (float): Execution timeout in seconds (default: 30.0)
• isolated (bool): Whether to execute code in a separate process (default: False)
• language (Language): Programming language (default: "python")

Docker Provider

Executes code in Docker containers for strong isolation and reproducible environments.

env = get_environment(
    "docker",
    image="python:3.13-slim",
    timeout=60.0,
    language="python"
)

Parameters:

• lifespan_handler: Tool server context manager (optional)
• image (str): Docker image to use (default: "python:3.13-slim")
• timeout (float): Execution timeout in seconds (default: 60.0)
• language (Language): Programming language (default: "python")

Daytona Provider

Executes code in remote Daytona sandboxes for cloud-based development environments.

env = get_environment(
    "daytona",
    api_url="https://api.daytona.io",
    api_key="your-api-key",
    timeout=300.0,
    keep_alive=False
)

Parameters:

• api_url (str): Daytona API URL (optional, uses env vars if not provided)
• api_key (str): API key for authentication (optional)
• target (str): Target configuration (optional)
• image (str): Container image (default: "python:3.13-slim")
• timeout (float): Execution timeout in seconds (default: 300.0)
• keep_alive (bool): Keep sandbox running after execution (default: False)

E2B Provider

Executes code in E2B sandboxes for secure, ephemeral execution environments.

env = get_environment(
    "e2b",
    template="python",
    timeout=300.0,
    keep_alive=False,
    language="python"
)

Parameters:

• template (str): E2B template to use (optional)
• timeout (float): Execution timeout in seconds (default: 300.0)
• keep_alive (bool): Keep sandbox running after execution (default: False)
• language (Language): Programming language (default: "python")

Beam Provider

Executes code in Beam cloud sandboxes for scalable, serverless execution environments.

env = get_environment(
    "beam",
    cpu=1.0,
    memory=128,
    keep_warm_seconds=600,
    timeout=300.0,
    language="python"
)

Parameters:

• cpu (float | str): CPU cores allocated to the container (default: 1.0)
• memory (int | str): Memory allocated to the container in MiB (default: 128)
• keep_warm_seconds (int): Seconds to keep sandbox alive, -1 for no timeout (default: 600)
• timeout (float): Execution timeout in seconds (default: 300.0)
• language (Language): Programming language (default: "python")

MCP Provider

Executes Python code with Model Context Protocol support for AI integrations.

env = get_environment(
    "mcp",
    dependencies=["requests", "numpy"],
    allow_networking=True,
    timeout=30.0
)

Parameters:

• dependencies (list[str]): Python packages to install (optional)
• allow_networking (bool): Allow network access (default: True)
• timeout (float): Execution timeout in seconds (default: 30.0)

Code Execution Patterns

All providers support two execution patterns:

1. Main Function Pattern

code = """
async def main():
    # Your code here
    return "result"
"""

2. Result Variable Pattern

code = """
import math
_result = math.pi * 2
"""

Error Handling

Execution results include comprehensive error information:

async with env:
    result = await env.execute(code)
    if result.success:
        print(f"Result: {result.result}")
        print(f"Duration: {result.duration:.3f}s")
    else:
        print(f"Error: {result.error}")
        print(f"Error Type: {result.error_type}")

Multi-Language Support

Some providers support multiple programming languages:

# JavaScript execution
env = get_environment("subprocess", language="javascript", executable="node")

# TypeScript execution
env = get_environment("docker", language="typescript", image="node:18")

Advanced Usage

Context Managers

All environments are async context managers for proper resource cleanup:

async with get_environment("docker") as env:
    result1 = await env.execute(code1)
    result2 = await env.execute(code2)  # Reuses same container
# Container automatically cleaned up

Custom Configurations

Each provider supports environment-specific customization:

# Docker with custom image and networking
env = get_environment(
    "docker",
    image="tensorflow/tensorflow:latest-py3",
    timeout=600.0
)

# Subprocess with specific Python version
env = get_environment(
    "subprocess",
    executable="/usr/bin/python3.11",
    timeout=120.0
)

Streaming Output

Some providers support streaming output line by line, useful for long-running processes:

from exxec import get_environment

# Stream output from subprocess execution
env = get_environment("subprocess")

async with env:
    async for line in env.execute_stream("""
    import time
    for i in range(5):
        print(f"Processing step {i+1}...")
        time.sleep(1)
    print("Done!")
    """):
        print(f"Live output: {line}")

# Also works with Docker execution
env = get_environment("docker")
async with env:
    async for line in env.execute_stream(code):
        # Process each line as it's produced
        if "ERROR" in line:
            print(f"⚠️  {line}")
        else:
            print(f"✓ {line}")

Supported providers: docker, local, beam, e2b, modal, vercel, ssh, daytona