Data Analysis

Write a CSV file and run analysis

Python example:

import os
import textwrap

from e2b_code_interpreter import Sandbox

sandbox = Sandbox.create(
    template="code-interpreter-v1",
    api_key=os.environ["E2B_API_KEY"],
    api_url=os.environ["E2B_API_URL"],
    domain=os.environ["E2B_DOMAIN"],
)

try:
    sandbox.files.write(
        "/tmp/sales.csv",
        "month,revenue\n2026-01,120\n2026-02,180\n2026-03,160\n",
    )

    execution = sandbox.run_code(
        textwrap.dedent(
            """
            import pandas as pd

            df = pd.read_csv("/tmp/sales.csv")
            print(df)
            df["revenue"].sum()
            """
        ),
        timeout=30,
        request_timeout=60,
    )

    if execution.error:
        raise RuntimeError(execution.error)

    stdout = "".join(execution.logs.stdout or [])
    stderr = "".join(execution.logs.stderr or [])

    print("stdout:")
    print(stdout.strip())
    print("stderr:")
    print(stderr.strip())
    print("text:")
    print(execution.text)
finally:
    sandbox.kill()

TypeScript example:

import { Sandbox } from "@e2b/code-interpreter";

const sandbox = await Sandbox.create("code-interpreter-v1", {
  apiKey: process.env.E2B_API_KEY,
  apiUrl: process.env.E2B_API_URL,
  domain: process.env.E2B_DOMAIN,
});

try {
  await sandbox.files.write(
    "/tmp/sales.csv",
    "month,revenue\n2026-01,120\n2026-02,180\n2026-03,160\n",
  );

  const execution = await sandbox.runCode(`
import pandas as pd

df = pd.read_csv("/tmp/sales.csv")
print(df)
df["revenue"].sum()
`, {
    timeoutMs: 30_000,
    requestTimeoutMs: 60_000,
  });

  if (execution.error) {
    throw new Error(`${execution.error.name}: ${execution.error.value}`);
  }

  console.log("stdout:");
  console.log(execution.logs.stdout.join("").trim());
  console.log("stderr:");
  console.log(execution.logs.stderr.join("").trim());
  console.log("text:");
  console.log(execution.text);
} finally {
  await sandbox.kill();
}

Generate a chart file

sandbox.run_code("""import matplotlib.pyplot as plt; plt.plot([1, 2, 3], [120, 180, 160]); plt.title("Revenue"); plt.savefig("/tmp/revenue.png")""")

content = sandbox.files.read("/tmp/revenue.png", format="bytes")

TypeScript example:

await sandbox.runCode(`import matplotlib.pyplot as plt; plt.plot([1, 2, 3], [120, 180, 160]); plt.title("Revenue"); plt.savefig("/tmp/revenue.png")`);

const content = await sandbox.files.read("/tmp/revenue.png", { format: "bytes" });

FC Agent Sandbox currently supports only the file-based fallback path for charts. For details, see Charts and Visualizations.

Integrate AI-generated code

Before you pass model-generated code into the sandbox, add basic guardrails on the business side:

• Define the path of each data file inside the sandbox.

• Limit code runtime, input file size, and output size.

• Return execution.error, stdout, stderr, and rich results to the model or user.

• Do not write user data, API keys, or long-lived business state into generated code.

Recommendations

• Small data files can be written with sandbox.files.write(). For large files, prefer upload URLs.

• Analysis code should explicitly read paths inside the sandbox, such as /tmp/input.csv.

• For charts and report files, write them to /tmp and then download them. For tables and text results, read from execution.results or output files.

• Call sandbox.kill() when the task is done to avoid holding resources longer than needed.