Skip to main content
E2B’s cloud environments are great runtime sandboxes for LLMs. E2B’s Data Analysis sandbox allows for safe code execution in a sandboxed environment. This is ideal for building tools such as code interpreters, or Advanced Data Analysis like in ChatGPT. E2B Data Analysis sandbox allows you to:
  • Run Python code
  • Generate charts via matplotlib
  • Install Python packages dynamically during runtime
  • Install system packages dynamically during runtime
  • Run shell commands
  • Upload and download files
We’ll create a simple OpenAI agent that will use E2B’s Data Analysis sandbox to perform analysis on a uploaded files using Python. Get your OpenAI API key and E2B API key here and set them as environment variables. You can find the full API documentation here. You’ll need to install e2b to get started: 
pip install -qU  langchain e2b langchain-community

from langchain_community.tools import E2BDataAnalysisTool

import os

from langchain.agents import AgentType, initialize_agent
from langchain_openai import ChatOpenAI

os.environ["E2B_API_KEY"] = "<E2B_API_KEY>"
os.environ["OPENAI_API_KEY"] = "<OPENAI_API_KEY>"
When creating an instance of the E2BDataAnalysisTool, you can pass callbacks to listen to the output of the sandbox. This is useful, for example, when creating more responsive UI. Especially with the combination of streaming output from LLMs. 
# Artifacts are charts created by matplotlib when `plt.show()` is called
def save_artifact(artifact):
    print("New matplotlib chart generated:", artifact.name)
    # Download the artifact as `bytes` and leave it up to the user to display them (on frontend, for example)
    file = artifact.download()
    basename = os.path.basename(artifact.name)

    # Save the chart to the `charts` directory
    with open(f"./charts/{basename}", "wb") as f:
        f.write(file)


e2b_data_analysis_tool = E2BDataAnalysisTool(
    # Pass environment variables to the sandbox
    env_vars={"MY_SECRET": "secret_value"},
    on_stdout=lambda stdout: print("stdout:", stdout),
    on_stderr=lambda stderr: print("stderr:", stderr),
    on_artifact=save_artifact,
)
Upload an example CSV data file to the sandbox so we can analyze it with our agent. You can use for example this file about Netflix tv shows. 
with open("./netflix.csv") as f:
    remote_path = e2b_data_analysis_tool.upload_file(
        file=f,
        description="Data about Netflix tv shows including their title, category, director, release date, casting, age rating, etc.",
    )
    print(remote_path)

name='netflix.csv' remote_path='/home/user/netflix.csv' description='Data about Netflix tv shows including their title, category, director, release date, casting, age rating, etc.'
Create a Tool object and initialize the LangChain agent. 
tools = [e2b_data_analysis_tool.as_tool()]

llm = ChatOpenAI(model="gpt-4", temperature=0)
agent = initialize_agent(
    tools,
    llm,
    agent=AgentType.OPENAI_FUNCTIONS,
    verbose=True,
    handle_parsing_errors=True,
)
Now we can ask the agent questions about the CSV file we uploaded earlier. 
agent.run(
    "What are the 5 longest movies on netflix released between 2000 and 2010? Create a chart with their lengths."
)

> Entering new AgentExecutor chain...

Invoking: `e2b_data_analysis` with `{'python_code': "import pandas as pd\n\n# Load the data\nnetflix_data = pd.read_csv('/home/user/netflix.csv')\n\n# Convert the 'release_year' column to integer\nnetflix_data['release_year'] = netflix_data['release_year'].astype(int)\n\n# Filter the data for movies released between 2000 and 2010\nfiltered_data = netflix_data[(netflix_data['release_year'] >= 2000) & (netflix_data['release_year'] <= 2010) & (netflix_data['type'] == 'Movie')]\n\n# Remove rows where 'duration' is not available\nfiltered_data = filtered_data[filtered_data['duration'].notna()]\n\n# Convert the 'duration' column to integer\nfiltered_data['duration'] = filtered_data['duration'].str.replace(' min','').astype(int)\n\n# Get the top 5 longest movies\nlongest_movies = filtered_data.nlargest(5, 'duration')\n\n# Create a bar chart\nimport matplotlib.pyplot as plt\n\nplt.figure(figsize=(10,5))\nplt.barh(longest_movies['title'], longest_movies['duration'], color='skyblue')\nplt.xlabel('Duration (minutes)')\nplt.title('Top 5 Longest Movies on Netflix (2000-2010)')\nplt.gca().invert_yaxis()\nplt.savefig('/home/user/longest_movies.png')\n\nlongest_movies[['title', 'duration']]"}`


stdout:                              title  duration
stdout: 1019                        Lagaan       224
stdout: 4573                  Jodhaa Akbar       214
stdout: 2731      Kabhi Khushi Kabhie Gham       209
stdout: 2632  No Direction Home: Bob Dylan       208
stdout: 2126          What's Your Raashee?       203
{'stdout': "                             title  duration\n1019                        Lagaan       224\n4573                  Jodhaa Akbar       214\n2731      Kabhi Khushi Kabhie Gham       209\n2632  No Direction Home: Bob Dylan       208\n2126          What's Your Raashee?       203", 'stderr': ''}The 5 longest movies on Netflix released between 2000 and 2010 are:

1. Lagaan - 224 minutes
2. Jodhaa Akbar - 214 minutes
3. Kabhi Khushi Kabhie Gham - 209 minutes
4. No Direction Home: Bob Dylan - 208 minutes
5. What's Your Raashee? - 203 minutes

Here is the chart showing their lengths:

![Longest Movies](sandbox:/home/user/longest_movies.png)

> Finished chain.

"The 5 longest movies on Netflix released between 2000 and 2010 are:\n\n1. Lagaan - 224 minutes\n2. Jodhaa Akbar - 214 minutes\n3. Kabhi Khushi Kabhie Gham - 209 minutes\n4. No Direction Home: Bob Dylan - 208 minutes\n5. What's Your Raashee? - 203 minutes\n\nHere is the chart showing their lengths:\n\n![Longest Movies](sandbox:/home/user/longest_movies.png)"
E2B also allows you to install both Python and system (via apt) packages dynamically during runtime like this: 
# Install Python package
e2b_data_analysis_tool.install_python_packages("pandas")

stdout: Requirement already satisfied: pandas in /usr/local/lib/python3.10/dist-packages (2.1.1)
stdout: Requirement already satisfied: python-dateutil>=2.8.2 in /usr/local/lib/python3.10/dist-packages (from pandas) (2.8.2)
stdout: Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.10/dist-packages (from pandas) (2023.3.post1)
stdout: Requirement already satisfied: numpy>=1.22.4 in /usr/local/lib/python3.10/dist-packages (from pandas) (1.26.1)
stdout: Requirement already satisfied: tzdata>=2022.1 in /usr/local/lib/python3.10/dist-packages (from pandas) (2023.3)
stdout: Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.10/dist-packages (from python-dateutil>=2.8.2->pandas) (1.16.0)
Additionally, you can download any file from the sandbox like this: 
# The path is a remote path in the sandbox
files_in_bytes = e2b_data_analysis_tool.download_file("/home/user/netflix.csv")
Lastly, you can run any shell command inside the sandbox via run_command. 
# Install SQLite
e2b_data_analysis_tool.run_command("sudo apt update")
e2b_data_analysis_tool.install_system_packages("sqlite3")

# Check the SQLite version
output = e2b_data_analysis_tool.run_command("sqlite3 --version")
print("version: ", output["stdout"])
print("error: ", output["stderr"])
print("exit code: ", output["exit_code"])

stderr:
stderr: WARNING: apt does not have a stable CLI interface. Use with caution in scripts.
stderr:
stdout: Hit:1 http://security.ubuntu.com/ubuntu jammy-security InRelease
stdout: Hit:2 http://archive.ubuntu.com/ubuntu jammy InRelease
stdout: Hit:3 http://archive.ubuntu.com/ubuntu jammy-updates InRelease
stdout: Hit:4 http://archive.ubuntu.com/ubuntu jammy-backports InRelease
stdout: Reading package lists...
stdout: Building dependency tree...
stdout: Reading state information...
stdout: All packages are up to date.
stdout: Reading package lists...
stdout: Building dependency tree...
stdout: Reading state information...
stdout: Suggested packages:
stdout:   sqlite3-doc
stdout: The following NEW packages will be installed:
stdout:   sqlite3
stdout: 0 upgraded, 1 newly installed, 0 to remove and 0 not upgraded.
stdout: Need to get 768 kB of archives.
stdout: After this operation, 1873 kB of additional disk space will be used.
stdout: Get:1 http://archive.ubuntu.com/ubuntu jammy-updates/main amd64 sqlite3 amd64 3.37.2-2ubuntu0.1 [768 kB]
stderr: debconf: delaying package configuration, since apt-utils is not installed
stdout: Fetched 768 kB in 0s (2258 kB/s)
stdout: Selecting previously unselected package sqlite3.
(Reading database ... 23999 files and directories currently installed.)
stdout: Preparing to unpack .../sqlite3_3.37.2-2ubuntu0.1_amd64.deb ...
stdout: Unpacking sqlite3 (3.37.2-2ubuntu0.1) ...
stdout: Setting up sqlite3 (3.37.2-2ubuntu0.1) ...
stdout: 3.37.2 2022-01-06 13:25:41 872ba256cbf61d9290b571c0e6d82a20c224ca3ad82971edc46b29818d5dalt1
version:  3.37.2 2022-01-06 13:25:41 872ba256cbf61d9290b571c0e6d82a20c224ca3ad82971edc46b29818d5dalt1
error:
exit code:  0
When your agent is finished, don’t forget to close the sandbox 
e2b_data_analysis_tool.close()
Edit the source of this page on GitHub
⌘I