You don't have to run a separate FastAPI server to handle additional webhooks anymore!

With the January Streamlit 1.53 release, you can now mount a Streamlit app over Starlette, effectively replacing Tornado.

However, the more interesting part is, you have programmatic access to the Starlette instance to add your own endpoints, middleware and error handlers! This enables so many new use cases:

• Custom endpoints: move your ML library to a Starlette endpoint, create a Stripe endpoint to catch the payment successful callback, use Starlette RedirectResponse to send the user to a new location
• Mount FastAPI-like apps: FastAPI objects, FastMCP servers, LangServe routes, any ASGI-compatible object probably
• Intercept every request with Middleware: rate-limit your Streamlit app, blacklist IP ranges, add enterprise-level security headers to every authenticated Streamlit user
• ...and more, cookies, static assets, 404 error pages, font serving, websockets in Starlette...

So many use cases it deserved its own video, which I mistakenly uploaded 12h early from choosing 2:45AM CE/T instead of 2:45PM.

​

Other Nuggets

• I met Streamlit Ambassador Siddhant Sadangi last October at PyData Paris, where he told me he was finalizing a new book about Streamlit. This book is now out, I have not read it yet but you should have a look too.

• The #30DaysOfAI Streamlit Social Media challenge is over, you can binge-build the tutorial now.
• The redesign of Components Page, you can now add your own component through a PR here. I wrote a Component building tutorial a long time ago, though it's not for version 2 yet which I think is much more manageable. I should update this tutorial to version 2.
• This Reflex vs Streamlit blog post is one of the rare comparison blog posts I've ever appreciated... A grounded take with a good explanation on the core value proposition for each library.

Next video?

So what should we study next? I got a few suggestions in social media comments: React over FastAPI for dashboards (though I personally prefer Nuxt, and am curious about Astro), Livecoding the 30 Days of AI Challenge, Stripe in Streamlit using the new Starlette, deep dive into FastMCP.

Feel free to tell me your preferences. Maybe I should vibe-code a "Vote for the next video" website ✨

See you soon DataFan!