paulroberttaylor / code-interpreter

Python & JS SDK for building custom code interpreters. Built with E2B - Cloud Runtime for AI Agents.

Stateful code interpreter

The repository contains a template and modules for the code interpreter sandbox. It is based on the Jupyter server and implements the Jupyter Kernel messaging protocol. This allows for sharing context between code executions and improves support for plotting charts and other display-able data.

Motivation

The code generated by LLMs is often split into code blocks, where each subsequent block references the previous one. This is a common pattern in Jupyter notebooks, where each cell can reference the variables and definitions from the previous cells. In the classical sandbox each code execution is independent and does not share the context with the previous executions.

This is suboptimal for a lot of Python use cases with LLMs. Especially GPT-3.5 and 4 expects it runs in a Jupyter Notebook environment. Even when ones tries to convince it otherwise. In practice, LLMs will generate code blocks which have references to previous code blocks. This becomes an issue if a user wants to execute each code block separately which often is the use case.

This new code interpreter template runs a Jupyter server inside the sandbox, which allows for sharing context between code executions. Additionally, this new template also partly implements the Jupyter Kernel messaging protocol. This means that, for example, support for plotting charts is now improved and we don't need to do hack-ish solutions like in the current production version of our code interpreter.

The current code interpreter allows to run Python code but each run share the context. That means that subsequent runs can reference to variables, definitions, etc from past code execution runs.

Installation

Python

pip install e2b-code-interpreter

JavaScript

npm install @e2b/code-interpreter

Examples

Minimal example with the sharing context

Python

from e2b_code_interpreter import CodeInterpreter

with CodeInterpreter() as sandbox:
    sandbox.notebook.exec_cell("x = 1")

    execution = sandbox.notebook.exec_cell("x+=1; x")
    print(execution.text)  # outputs 2

JavaScript

import { CodeInterpreter } from '@e2b/code-interpreter'

const sandbox = await CodeInterpreter.create()
await sandbox.notebook.execCell('x = 1')

const execution = await sandbox.notebook.execCell('x+=1; x')
console.log(execution.text)  // outputs 2

await sandbox.close()

Get charts and any display-able data

Python

import base64
import io

from matplotlib import image as mpimg, pyplot as plt

from e2b_code_interpreter import CodeInterpreter

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

x = np.linspace(0, 20, 100)
y = np.sin(x)

plt.plot(x, y)
plt.show()
"""

with CodeInterpreter() as sandbox:
    # you can install dependencies in "jupyter notebook style"
    sandbox.notebook.exec_cell("!pip install matplotlib")

    # plot random graph
    execution = sandbox.notebook.exec_cell(code)

# there's your image
image = execution.results[0].png

# example how to show the image / prove it works
i = base64.b64decode(image)
i = io.BytesIO(i)
i = mpimg.imread(i, format='PNG')

plt.imshow(i, interpolation='nearest')
plt.show()

JavaScript

import { CodeInterpreter } from '@e2b/code-interpreter'

const sandbox = await CodeInterpreter.create()

const code = `
import matplotlib.pyplot as plt
import numpy as np

x = np.linspace(0, 20, 100)
y = np.sin(x)

plt.plot(x, y)
plt.show()
`;

// you can install dependencies in "jupyter notebook style"
await sandbox.notebook.execCell("!pip install matplotlib")

const execution = await sandbox.notebook.execCell(code)

// this contains the image data, you can e.g. save it to file or send to frontend
execution.results[0].png

await sandbox.close()

Streaming code output

Python

from e2b_code_interpreter import CodeInterpreter

code = """
import time
import pandas as pd

print("hello")
time.sleep(3)
data = pd.DataFrame(data=[[1, 2], [3, 4]], columns=["A", "B"])
display(data.head(10))
time.sleep(3)
print("world")
"""

with CodeInterpreter() as sandbox:
    sandbox.notebook.exec_cell(code, on_stdout=print, on_stderr=print, on_result=(lambda result: print(result.text)))

JavaScript

import { CodeInterpreter } from '@e2b/code-interpreter'

const code = `
import time
import pandas as pd

print("hello")
time.sleep(3)
data = pd.DataFrame(data=[[1, 2], [3, 4]], columns=["A", "B"])
display(data.head(10))
time.sleep(3)
print("world")
`

const sandbox = await CodeInterpreter.create()

await sandbox.notebook.execCell(code, {
  onStdout: (out) => console.log(out),
  onStderr: (outErr) => console.error(outErr),
  onResult: (result) => console.log(result.text)
})

await sandbox.close()

Pre-installed Python packages inside the sandbox

The full and always up-to-date list can be found in the requirements.txt file.

# Jupyter server requirements
jupyter-server==2.13.0
ipykernel==6.29.3
ipython==8.22.2

# Other packages
aiohttp==3.9.3
beautifulsoup4==4.12.3
bokeh==3.3.4
gensim==4.3.2
imageio==2.34.0
joblib==1.3.2
librosa==0.10.1
matplotlib==3.8.3
nltk==3.8.1
numpy==1.26.4
opencv-python==4.9.0.80
openpyxl==3.1.2
pandas==1.5.3
plotly==5.19.0
pytest==8.1.0
python-docx==1.1.0
pytz==2024.1
requests==2.26.0
scikit-image==0.22.0
scikit-learn==1.4.1.post1
scipy==1.12.0
seaborn==0.13.2
soundfile==0.12.1
spacy==3.7.4
textblob==0.18.0
tornado==6.4
urllib3==1.26.7
xarray==2024.2.0
xlrd==2.0.1

Custom template using Code Interpreter

The template requires custom setup. If you want to build your own custom template and use Code Interpreter, look at README.md in the template folder.