Demonstrate Bluesky in a single process

[danielballan]

Show how to run Tiled and RunEngine in a single process.

[przluca]

Hi, what do you mean by single process? Can you describe your intentions?

[danielballan]

One of the design goals of Bluesky is to "scale down", to be easy to set up and run on a laptop without configuring a bunch of services.

For example, Bluesky can be used in IPython, with EPICS connections (via ophyd and pyepics), visualization (via interactive matplotlib figures in Qt windows), and data persistence (via writes to MongoDB via pymongo) happening in a single process.

Tiled/Databroker 2.0 introduces an HTTP server into this stack, which would typically be launched as a separate service, and would sit between Bluesky and MongoDB. For secure, production, scalable deployments, that process-separation is standard and encouraged. For experimentation or minimalist deployments, it is possible to run the Tiled server in-process. (The event loop of the server is run on a background thread, and HTTP messages are passed over the ASGI protocol via Python function calls rather than network send/recv calls.)

[untzag]

I'm more interested in documenting workflows including suitcase-csv and custom domain-specific serializers. Even if everything is in one process, applications like tiled still really want to run persistently. At the smallest scale, I want researchers to be able to run bluesky plans in scripts and monolithic graphical applications with ZERO persistent services.

[untzag]

I'm not a tiled user at this time, so I might just not understand what the advantages are at the smallest scale of instrumentation. Workflows are hodgepodge, and for better or worse people really love having a local "raw data" folder on their hard-drive.