Our existing Flask + graphql-ws + gevent implementation of dagit has a number of issues
I did an exploration of starlette in D8040 which was promising. I believe we can move to this modern ASGI framework which gives us greater control and testability. This will be key in both resolving current issues and evolving dagit more effectively moving forward.
This first diff gets basic scaffolding in place and the dagit_info endpoint. Subsequent diffs will add other endpoints until we have parity with the flask app. After some further experimentation with it once working we can figure out a cut over strategy.