Run Python | Substrate

Run custom code using RunPython

Substrate has a built-in Python sandbox, which you can use to run custom Python code on Substrate. Similar functionality for the TypeScript SDK is coming soon.

In the Python SDK, create a RunPython node with:

function – The function to run remotely.
kwargs – Keyword arguments to your function. This can include future references to the output of previous nodes.


start = 1
def add(n: int):
    print("hello")
    return start + n
add1 = RunPython(
    function=add,
    kwargs={"n": 1},
)
add2 = RunPython(
    function=add,
    kwargs={"n": add1.future.output},
)
res = substrate.run(add1, add2)
add1_out = res.get(add1)
print(add1_out.output) # 2
add2_out = res.get(add2)
print(add2_out.output) # 3
print(add2_out.stdout) # hello

Any outputs printed to stdout are returned in the stdout field. If your code does not run successfully, the contents of stderr are returned in the stderr field.

Substrate's code interpreter runs Python in a sandbox with network access. You can add packages to the sandbox by passing a list of Python dependencies to pip_install.


def markdown(url: str):
    import requests
    from bs4 import BeautifulSoup
    from markdownify import markdownify
    res = requests.get(url)
    soup = BeautifulSoup(res.content, "html.parser")
    return markdownify(str(soup))
md = RunPython(
    function=markdown,
    kwargs={
        "url": "https://metrograph.com/film/?vista_film_id=9999001208",
    },
    pip_install=["requests", "beautifulsoup4", "markdownify"],
)
res = substrate.run(md)
out = res.get(md)
print(out.output) # (markdown content of the web page)

Here, we define a local function markdown that retrieves the markdown content of a web page. Then, we run the function remotely on Substrate using RunPython.

If Streaming