datasette-plugin-router

WIP router for Datasette plugins

Datasette plugins that have a lot of custom API endpoints can get tiresome to write by hand. datasette-plugin-router aims to be a small Python library that adds a FastAPI-like API for defining custom Datasette plugin endpoints.

• Define routes with familiar GET/POST decorators
• Define Pydantic-backed input/output schemas on JSON endpoints
• register_routes() compatability
• export to OpenAPI schema for codegen'ing clients

Sample usage:

from datasette import Response, hookimpl
from datasette_plugin_router import Router, Body
from pydantic import BaseModel

router = Router()

class Input(BaseModel):
    id: int
    name: str

class Output(BaseModel):
    id_negative: int
    name_upper: str

@router.POST(r"/-/demo1$", output=Output)
async def demo1(params: Body[Input]) -> Output:
    output = Output(
        id_negative=-1 * params.id,
        name_upper=params.name.upper(),
    )
    return Response.json(output.model_dump())


@router.GET(r"/-/hello/(?P<name>.*)$")
async def hello(name: str):
    return Response.html(f"<h1>Hello, {name}!</h1>")


@hookimpl
def register_routes():
    return router.routes()

Request body validation errors

If a Body()-injected request body fails Pydantic validation — including an empty body or malformed JSON — the router returns a 400 instead of a 500:

{
  "error": "id: Input should be a valid integer, unable to parse string as an integer",
  "errors": [{"type": "int_parsing", "loc": ["id"], "msg": "..."}]
}

• error joins each field error as "<loc>: <msg>" (just <msg> when there is no location, e.g. malformed JSON) with "; ".
• errors is ValidationError.errors() minus the url/ctx/input fields.
• Messages are Pydantic's msg verbatim, so custom validator messages pass through.