hopx-ai 0.2.6

Official Python SDK for HOPX.AI Sandboxes

Hopx Python SDK

Official Python SDK for Hopx.ai - Cloud sandboxes for AI agents and code execution.

What is Hopx.ai?

Hopx.ai provides secure, isolated cloud sandboxes that start in seconds. Each sandbox is a lightweight VM with root access, pre-installed development tools, and configurable network access.

Use cases:

• AI agent code execution
• Running untrusted code safely
• Integration testing
• Data processing and analysis
• Browser automation
• Educational coding environments

Sandbox features:

• Full root access
• Pre-installed development tools
• Configurable network access
• Persistent filesystem during session
• Automatic cleanup after timeout

Installation

pip install hopx-ai

Quick Start

Basic Usage

from hopx_ai import Sandbox

# Create sandbox
sandbox = Sandbox.create(
    template="python",
    api_key="your-api-key"  # or set HOPX_API_KEY env var
)

# Execute code
result = sandbox.run_code("""
import sys
print(f"Python {sys.version}")
print("Hello from Hopx!")
""")

print(result.stdout)
# Output:
# Python 3.11.x
# Hello from Hopx!

# Clean up
sandbox.kill()

Context Manager

The context manager handles cleanup automatically:

from hopx_ai import Sandbox

with Sandbox.create(template="python") as sandbox:
    result = sandbox.run_code("print(2 + 2)")
    print(result.stdout)  # "4"

Async Support

from hopx_ai import AsyncSandbox
import asyncio

async def main():
    async with AsyncSandbox.create(template="python") as sandbox:
        result = await sandbox.run_code("print('Async!')")
        print(result.stdout)

asyncio.run(main())

Examples

AI Code Execution

Execute code generated by LLM agents:

from hopx_ai import Sandbox

agent_code = """
import pandas as pd
df = pd.DataFrame({'A': [1, 2, 3], 'B': [4, 5, 6]})
print(df.describe())
"""

sandbox = Sandbox.create(template="python")
result = sandbox.run_code(agent_code)

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

sandbox.kill()

Data Analysis with Charts

Capture matplotlib charts as base64-encoded PNG:

code = """
import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 10, 100)
plt.plot(x, np.sin(x))
plt.title('Sine Wave')
plt.show()
"""

sandbox = Sandbox.create(template="python")
result = sandbox.run_code(code)

# Access chart data
if result.rich_outputs:
    png_data = result.rich_outputs[0].data  # Base64 PNG

sandbox.kill()

Multi-Step Workflow

Run commands sequentially in the same environment:

from hopx_ai import Sandbox

sandbox = Sandbox.create(template="nodejs", timeout_seconds=600)

# Clone and install
sandbox.commands.run("git clone https://github.com/user/project.git /app")
sandbox.commands.run("cd /app && npm install")

# Run tests
result = sandbox.commands.run("cd /app && npm test")
print(f"Tests: {'PASSED' if result.exit_code == 0 else 'FAILED'}")

# Build
sandbox.commands.run("cd /app && npm run build")

# List artifacts
files = sandbox.files.list("/app/dist/")
for file in files:
    print(f"Built: {file.name} ({file.size} bytes)")

sandbox.kill()

File Processing

Upload files, process data, and download results:

sandbox = Sandbox.create(template="python")

# Upload data
sandbox.files.write("/tmp/data.csv", csv_content)

# Process
result = sandbox.run_code("""
import pandas as pd
df = pd.read_csv('/tmp/data.csv')
result = df.groupby('category').sum()
result.to_csv('/tmp/output.csv')
print(f"Processed {len(df)} rows")
""")

# Download result
output = sandbox.files.read("/tmp/output.csv")
print(output)

sandbox.kill()

Core Features

Code Execution

Execute code in Python, JavaScript, or Bash:

# Python
result = sandbox.run_code("print('Hello')", language="python")

# JavaScript
result = sandbox.run_code("console.log('Hello')", language="javascript")

# Bash
result = sandbox.run_code("echo 'Hello'", language="bash")

# With environment variables
result = sandbox.run_code(
    "import os; print(os.environ['API_KEY'])",
    env={"API_KEY": "secret"}
)

File Operations

# Write files
sandbox.files.write("/app/config.json", '{"key": "value"}')

# Read files
content = sandbox.files.read("/app/config.json")

# List directory
files = sandbox.files.list("/app/")
for file in files:
    print(f"{file.name}: {file.size} bytes")

# Delete files
sandbox.files.delete("/app/temp.txt")

Commands

# Run synchronously
result = sandbox.commands.run("ls -la /app")
print(result.stdout)

# Run in background
cmd_id = sandbox.commands.run_async("python long_task.py")
result = sandbox.commands.get_result(cmd_id)

Environment Variables

# Set single variable
sandbox.env.set("DATABASE_URL", "postgresql://...")

# Set multiple variables
sandbox.env.set_many({
    "API_KEY": "key123",
    "DEBUG": "true"
})

# Get variable
value = sandbox.env.get("API_KEY")

# Delete variable
sandbox.env.delete("DEBUG")

Template Building

Build custom sandbox environments with your dependencies:

from hopx_ai import Template, wait_for_port
from hopx_ai.template import BuildOptions

# Define template
template = (
    Template()
    .from_python_image("3.11")
    .copy("requirements.txt", "/app/requirements.txt")
    .copy("src/", "/app/src/")
    .run("cd /app && pip install -r requirements.txt")
    .set_workdir("/app")
    .set_env("PORT", "8000")
    .set_start_cmd("python src/main.py", wait_for_port(8000))
)

# Build template
result = await Template.build(
    template,
    BuildOptions(
        name="my-python-app",
        api_key="your-api-key",
        on_log=lambda log: print(f"[{log['level']}] {log['message']}")
    )
)

print(f"Template ID: {result.template_id}")

# Create sandbox from template
sandbox = Sandbox.create(template_id=result.template_id)

Authentication

Get your API key at hopx.ai/dashboard

Set via environment variable:

export HOPX_API_KEY="your-api-key"

Or pass directly to methods:

sandbox = Sandbox.create(
    template="python",
    api_key="your-api-key"
)

Configuration

Template Build Timeout

Template.build() waits for templates to reach "active" status before returning. Templates transition through: building → publishing → active. The publishing phase takes 60-120 seconds.

Default timeout: 2700 seconds (45 minutes)

Environment variable:

export HOPX_TEMPLATE_BAKE_SECONDS=2700

BuildOptions parameter:

result = await Template.build(
    template,
    BuildOptions(
        name="my-template",
        api_key="...",
        template_activation_timeout=1800  # 1800 seconds = 30 minutes
    )
)

Priority: BuildOptions.template_activation_timeout > HOPX_TEMPLATE_BAKE_SECONDS > Default (2700s)

Pre-built Templates

Template	Description
python	Python 3.11 with pip, numpy, pandas, requests
nodejs	Node.js 20 with npm, common packages
code-interpreter	Python with data science stack (pandas, numpy, matplotlib, seaborn, scikit-learn)
go	Go 1.21
rust	Rust with Cargo
java	Java 17 with Maven

Build custom templates with Template.build() for specific requirements.

Advanced Features

Rich Output Capture

Capture matplotlib charts and visualizations:

result = sandbox.run_code("""
import matplotlib.pyplot as plt
plt.plot([1, 2, 3], [1, 4, 9])
plt.title('My Chart')
plt.show()
""")

# Access PNG data
for output in result.rich_outputs:
    if output.type == "image/png":
        import base64
        with open("chart.png", "wb") as f:
            f.write(base64.b64decode(output.data))

Process Management

# List processes
processes = sandbox.processes.list()
for proc in processes:
    print(f"{proc.pid}: {proc.name} (CPU: {proc.cpu_percent}%)")

# Kill process
sandbox.processes.kill(1234)

Desktop Automation

Control sandboxes with graphical interfaces:

# Get VNC connection info
vnc = sandbox.desktop.get_vnc_info()
print(f"Connect to: {vnc.url}")

# Take screenshot
screenshot = sandbox.desktop.screenshot()  # Returns PNG bytes

# Control mouse
sandbox.desktop.mouse_click(100, 200)

# Type text
sandbox.desktop.keyboard_type("Hello, World!")

Health and Metrics

Monitor sandbox resource usage:

# Check health
health = sandbox.get_health()
print(health.status)

# Get metrics
metrics = sandbox.get_metrics()
print(f"CPU: {metrics.cpu_percent}%")
print(f"Memory: {metrics.memory_mb}MB")
print(f"Disk: {metrics.disk_mb}MB")

Error Handling

from hopx_ai import (
    HopxError,
    AuthenticationError,
    CodeExecutionError,
    FileNotFoundError,
    RateLimitError
)

try:
    sandbox = Sandbox.create(template="python")
    result = sandbox.run_code("1/0")

except AuthenticationError:
    print("Invalid API key")
except CodeExecutionError as e:
    print(f"Code execution failed: {e.stderr}")
except RateLimitError:
    print("Rate limit exceeded")
except HopxError as e:
    print(f"API error: {e.message}")

Best Practices

Resource management:

• Use context managers for automatic cleanup
• Set timeout_seconds to prevent runaway sandboxes
• Call .kill() explicitly when not using context managers

Error handling:

• Wrap sandbox operations in try/except blocks
• Check result.success before accessing output
• Review result.stderr for error details

Performance:

• Use pre-built templates when possible (faster than custom builds)
• Group related operations to reduce API calls
• Monitor resource usage for long-running tasks

Troubleshooting

Sandbox creation fails:

• Verify API key is valid
• Check network connectivity
• Review error message for specific issue

Code execution errors:

• Check result.stderr for error details
• Verify required packages are installed
• Confirm file paths are correct

File not found errors:

• Use absolute paths (e.g., /app/file.txt)
• Verify file upload completed successfully
• Check current working directory

Documentation

• Full Documentation
• API Reference
• Examples
• Cookbook

Support

• Email: [e-mail]
• Discord: discord.gg/hopx
• GitHub Issues: github.com/hopx-ai/hopx/issues

License

MIT License - see LICENSE file for details.

Links

• Website
• Dashboard
• GitHub
• Discord
• Twitter