First analyze the existing information on the Posit site, then consider in the context of what skills current undergraduate computer science and data science students to get that first entry job in a tough job market the the back half of 2026 and throughout 2027. Finally, consider the workflow of a Posit future enterprise customer, not just the great person using Posit free tools.
- ensure teams are using same dataset regardless of how the analysis was done
- use FastAPI as data gateway for Jupyter Notebook, Shiny app
benefits summarized from article:
- code simplification
- make data accessible to different apps and workflows (both R and Python developers)
- increased security due to the API authentication layer
my assessment
- data security needs to be treated seriously. It is a pain for many people to deal with, but I believe that data access must be managed. Using FastAPI authentication with Shiny token-based authentication helps with security. In this architecture, Shiny sends username/password to FastAPI
/login using HTTPS and then FastAPI provides validation and sends back a JWT. This is the common Authorization: Bearer <token> format
- many stories of the data analysis workflow gloss over how the data gets inputted into the data set and how the data analyst confirms they have the most recent dataset. If the data is in manufacturing or ERP, the data is streaming into a central system constantly. In this case, it is going to be painful to constantly attempt to download the data as CSV for every analysis. Additionally, the human is likely doing massive filtering on the dataset with SQL or other query mechanism prior to loading it into pandas or other framework
- most data workflow stories do not address mobile. Mobile may be relevant in both the presentation and the data input phases. Although mobile web on phones and tablets address parts of these problems, it is often not the best user experience. Shiny does use reactive screen updates, so it is going to be better than Streamlit, but I still think that having the data as a secured API endpoint opens the market for companies that need a mobile device in the field.
I believe that FastAPI is a supported deploy type in Posit Connect. However, when I used it with Posit Connect Cloud, I was not able to deploy it.
The document also covers outputting HTML from the FastAPI server in addition to the API data.
The features of Posit Connect (the on-premise enterprise one) look quite excellent, but I can't find any pricing information online.
As a thought exercise, describe the type of person that buys Posit Connect (the on-premise version). If you think about it, I believe you'll see that the "data science" workflow is expanding into new areas.
It is possible to deploy Shiny using wasm and pyodide. This allows Shiny to run in a web browser without the Shiny server.
Although I have not used Shinylive with wasm and pyodid, I have a good amount of experience with wasm and pyodide in Flet wasm deployments. As an aside, Flet when deployed as a server, uses FastAPI itself as the backend while Shiny uses Starlette (which FastAPI itself is based on).
The Shinylive examples use Pyodide. It's impressive that they got it working.
My assessment
Although it is initially seems simpler to run Shiny without a server and use Pyodide with wasm, I do not recommend this approach in general. I do use this approach to host Flet apps on GitHub pages. However, there are differences in compatibility in what Pyodide can use as dependencies versus the local development environment.
