How do Dash, Shiny, and Streamlit compare as low-code,
UI layers for AI and ML models?
| Dash | Shiny | Streamlit |
|---|
| Languages Supported | Python | R only | Python only |
| OSS License | MIT | AGPL | Apache 2.0 |
| Back-end Architecture | Stateless | Stateful | Stateful |
| Downloads/month | 350,000 | 370,000 | 100,000 |
| Web protocol | HTTP(S) | Websockets | Websockets |
| Recommended deployment | Dash Enterprise | Shiny Server Pro (RStudio Connect) | Heroku |
| User experience | Web app | Model output with controls in a Web page | Notebook with controls |
| App structure | Multi-page | Single page | Notebook with code |
| Front end | React | jQuery | React |
| Interactivity | Complete: any component can be an input/output, including tables | Partial: some components can be inputs/outputs | Limited: only widgets as inputs, graphs and tables can only be outputs |
| Styling control | Dash Design Kit, Basic Bootstrap, or CSS | Basic Bootstrap or CSS | Limited |
| Crossfiltering | ✔️ | ✔️ | |
| Dynamic Layouts | Pattern-Matching Callbacks | | |
| Long Running Tasks | ✔️ | | |
| Year founded | 2017 | 2012 | 2019 |
| Funding (USD) | $20M | $8M | $27M |
| GitHub contributors | 47 | 55 | 54 |
| GitHub Stars | 12.5k | 3.9k | 10.1k |
