Proxy Server

ML data is rarely useful unless it can be exposed to other services.

This is where the proxy server simplifies the entire process.

CLI command

To expose all models and data, use the following command:

aligned proxy-server

You can also use the --expose-tag option to expose only specific data and model contracts that contain a particular tag.

By default, the command reads contracts from contracts_store.json. However, you can customise this by using the --contracts option to specify a different file.

Customising the server

Often, you may want to customise the API further by adding authentication, custom logging, and more. You can achieve this using the following code to obtain the FastAPI router:

from fastapi import FastAPI, Request, Depends
from contextlib import asynccontextmanager
from aligned import ContractStore
from aligned.proxy_api import router_for_store

async def authenticate_user(request: Request):
    ...

@asynccontextmanager
async def lifespan(app: FastAPI):
    store: ContractStore = ...

    contracts_router = router_for_store(
        store, 
        expose_tag="expose-through-api"  # Use None to expose all contracts
    )
    app.include_router(
        contracts_router,
        dependencies=[Depends(authenticate_user)]
    )
    yield

app = FastAPI(lifespan=lifespan)