FastAPI Integration

[mingjerli]

FastAPI Integration

We recently wrote a blog post to build a FastAPI app to forecast the regional gas price with LineaPy by running:

lineapy.to_pipeline(['price_change','all_regions'], input_parameters=['region'])

in the forecasting notebook to create a module file (price_change_all_regions_module.py) and use functions in the module within the app (app.py) as follows:

# app.py

from fastapi import FastAPI
from price_change_all_regions_module import get_all_regions
from price_change_all_regions_module import get_price_change
from price_change_all_regions_module import get_df_long_for_artifact_all_regions_and_downstream

app = FastAPI()

@app.get("/regions/")
def all_regions():
    df_long = get_df_long_for_artifact_all_regions_and_downstream()
    all_regions = get_all_regions(df_long)
    return '|'.join(all_regions)
    
@app.get("/{region}/pricechange/")
def price_change(region):
    df_long = get_df_long_for_artifact_all_regions_and_downstream()
    price_change = get_price_change(df_long, region)
    return f"Gas price in {region} is expected to change {price_change:.2f} next week."

Thus, app developers can focus on building the app but not refactoring the code.

We feel this is an interesting way to use LineaPy and we also see some potential to integrate into lineapy.to_pipeline as another pipeline framework, so anyone can put together an app directly from the LineaPy artifact store in no time.
Would this feature be helpful to you? We want to hear your voice. Let us know by voting or submitting a reply!