Add a REST server with FastAPI

[nelimee]

This is a draft PR trying to implement a REST server on the backend side.

This is the initial implementation that will allow a link between the frontend and the backend.

Currently, the idea is the following: all the QEC-related state is stored in the backend and the frontend is making requests through the provided REST API to 1. query the current QEC state or 2. change it. Here, "QEC-related state" means that:

• the frontend will not store the plaquettes, it will rather query the backend to have a description of the currently implemented plaquettes when needed,
• the frontend will not store the templates, it will rather query the backend to have a description of the currently implemented plaquettes when needed,

Because we want to avoid a communication at each user inputs, the frontend may store partial state.

Note that everything I wrote above and the code from this PR is a proposal, I am very much open to any other idea if it turns out to be easier to use.

To use the code from that PR and experiment with the REST server, follow the instructions:

1. checkout the branch from that PR (https://github.com/nelimee/tqec/tree/feat/fast-api-rest),
2. re-install the tqec package to get the new dependencies

   python -m pip install -e ./

   or install the added dependencies directly

   python -m pip install pydantic fastapi

3. Start the REST server with

   fastapi dev ./web/server.py

4. If everything succeeded, you can access the server on http://127.0.0.1:8000/ (try out http://127.0.0.1:8000/templates/list or http://127.0.0.1:8000/plaquettes/list). The REST api documentation is available at http://127.0.0.1:8000/docs.

Related issues:

• 170: if the decision to pursue in the direction explained above is taken, the backend will be responsible for that, and already implement the feature.

• 169: the backend would become responsible for the generation of the file, the frontend will still need to provide a way to save the file to the user.

• 61: nicely handled by using pydantic and implementing methods to convert from/to the defined pydantic.BaseModel.

• 38: the main related issue, this PR try to fix #38.

• 18: this PR introduces the possibility to export to Stim files, and so should fix that issue.

[Gistbatch]

Just as a reminder: We already had a Flask server in place, which was initially set up by @smburdick, and the front end has some unused functionality. The file server root is in /web/main.py, also deployed to gcloud. I don't have any strong preference between flask and fastapi, and most of the concepts are very similar.

• From a dev perspective, we might eventually think about hosting a proper database, which is straightforward with Pydantic.
• If we plan to host the backend and the frontend on different sites, we also need to enable CORS
• We should start to think about what information needs to be transferred at which points in time
• Also we should set up a proper structure for the webserver

As an alternative, there is also GraphQL which can be used to query data in a different way, I don't think it's useful for our purposes though.

[nelimee]

I have no personal preference on that. As I already said multiple times, I am not competent for web-related development task (and I try to stay as far away as possible from anything related to web-technologies).

My choice of pydantic + fastapi was based on the easy of use of pydantic and the ultra-fast integration of fastapi once you have pydantic models in place. There is really no smart thinking that happened on my side, an I will adopt whatever is judged to be the best solution by others.

Let's consider this PR as a playground to start exploring the integration between the frontend and backend, and see what is required from both sides. At any time (if possible, early), if the choice of pydantic + fastapi seems worse than another library, please tell me, or change the framework directly.

[Gistbatch]

I'm very much in favor of FastAPI because I have worked with it before. I think pydantic also works with flask and django.

[Gistbatch]

Could you move this to a repo branch instead of the fork? This would make it easier for others to check out / work on it.

[nelimee]

I will resolve all the conflicts and push to a repository branch in the afternoon.

[nelimee]

Replaced by #240