Use Cases

AI AGENTS

Code Interpreter for LLMs

Let your AI assistant write and execute Python code to answer questions, analyze data, and solve problems.

ChatGPT-style code execution
Data analysis and visualization
Mathematical computations
File processing and transformations

from cerver import Sandbox
from openai import OpenAI

sandbox = Sandbox()
client = OpenAI()

# AI generates code, Cerver runs it safely
response = client.chat.completions.create(
    model="gpt-4",
    messages=[{
        "role": "user",
        "content": "Calculate compound interest"
    }],
    tools=[{"type": "function", ...}]
)

result = sandbox.run(generated_code)
print(result.output)

DATA SCIENCE

Interactive Data Analysis

Run pandas, numpy, and matplotlib in isolated environments. Perfect for data pipelines and exploratory analysis.

CSV/JSON processing
Statistical analysis
Chart generation
Machine learning inference

with Sandbox(engine="containers") as sandbox:
    sandbox.install("pandas matplotlib")
    sandbox.write_file("data.csv", csv_content)

    result = sandbox.run("""
import pandas as pd
import matplotlib.pyplot as plt

df = pd.read_csv(__sandbox_dir__ + '/data.csv')
print(df.describe())

df.plot(kind='bar')
plt.savefig(__sandbox_dir__ + '/chart.png')
""")

    chart = sandbox.read_file("chart.png")

EDUCATION

Coding Playgrounds

Build interactive coding environments for learning platforms, bootcamps, and technical interviews.

Real-time code execution
Isolated student environments
Automated grading
No infrastructure to manage

# Student submits solution
student_code = request.json["code"]

with Sandbox() as sandbox:
    # Run with timeout for safety
    result = sandbox.run(student_code)

    # Check against test cases
    test_result = sandbox.run("""
assert fibonacci(10) == 55
assert fibonacci(0) == 0
print("All tests passed!")
""")

    return {
        "output": result.output,
        "passed": test_result.exit_code == 0
    }

AUTOMATION

ETL & Data Pipelines

Run user-defined transformations on data without risking your infrastructure.

Custom data transformations
User-defined functions
Webhook processors
Scheduled jobs

# User defines their transformation
user_transform = """
def transform(data):
    return [
        {**row, 'total': row['qty'] * row['price']}
        for row in data
    ]
"""

with Sandbox() as sandbox:
    sandbox.write_file("input.json", json.dumps(data))

    result = sandbox.run(f"""
import json
{user_transform}

with open(__sandbox_dir__ + '/input.json') as f:
    data = json.load(f)

output = transform(data)
print(json.dumps(output))
""")

DEVTOOLS

Code Review & Testing

Automatically run and validate code snippets from PRs, documentation, or user submissions.

PR code validation
Documentation testing
Syntax checking
Security scanning

# Validate code from a pull request
for snippet in extract_code_blocks(pr_diff):
    with Sandbox() as sandbox:
        result = sandbox.run(snippet)

        if result.exit_code != 0:
            post_pr_comment(
                f"Code block failed:\n{result.error}"
            )
        else:
            print(f"✓ Snippet passed")

REAL-TIME

Streaming Output

Show users real-time progress for long-running tasks. Perfect for data processing, training, and batch jobs.

Live progress updates
Server-Sent Events
Long-running computations
Interactive debugging