This page lists highlights, bug fixes, and known issues for official Streamlit releases. If you’re looking for information about nightly releases, beta features, or experimental features, see Try pre-release features.


To upgrade to the latest version of Streamlit, run:

$ pip install --upgrade streamlit

Version 0.67.0

Release date: September 16, 2020


  • 🦷 Streamlit Components can now return bytes to your Streamlit App. To create a component that returns bytes, make sure to upgrade to the latest streamlit-component-lib.

Notable Changes

  • 📈 Deprecation warning: Beginning December 1st, 2020 st.pyplot() will require a figure to be provided. To disable the deprecation warning, please set deprecation.showPyplotGlobalUse to False

  • 🎚 st.multiselect and are now lightning fast when working with large datasets. Thanks masa3141!

Version 0.66.0

Release date: September 1, 2020


  • ✏️ st.write is now available for use in the sidebar!

  • 🎚 A slider for distinct or non-numerical values is now available with st.select_slider.

  • ⌗ Streamlit Components can now return dataframes to your Streamlit App. Check out our SelectableDataTable example.

  • 📦 The Streamlit Components library used in our Streamlit Component template is now available as a npm package (streamlit-component-lib) to simplify future upgrades to the latest version. Existing components do not need to migrate.

Notable Changes

  • 🐼 Support StringDtype from pandas version 1.0.0

  • 🧦 Support for running Streamlit on Unix sockets

Version 0.65.0

Release date: August 12, 2020


  • ⚙️ Ability to set page title, favicon, sidebar state, and wide mode via st.beta_set_page_config(). See our documentation for details.

  • 📝 Add stateful behaviors through the use of query parameters with st.experimental_set_query_params and st.experimental_get_query_params. Thanks @zhaoooyue!

  • 🐼 Improved pandas dataframe support for, st.selectbox, and st.multiselect.

  • 🛑 Break out of your Streamlit app with st.stop.

  • 🖼 Inline SVG support for st.image.


  • 🚨Deprecation Warning: The st.image parameter format has been renamed to output_format.

Version 0.64.0

Release date: July 23, 2020


  • 📊 Default matplotlib to display charts with a tight layout. To disable this, set bbox_inches to None, inches as a string, or a Bbox

  • 🗃 Deprecation warning for automatic encoding on st.file_uploader

  • 🙈 If gatherUserStats is False, do not even load the Segment library. Thanks @tanmaylaud!

Version 0.63.0

Release date: July 13, 2020


  • 🧩 Support for Streamlit Components!!! See documentation for more info.

  • 🕗 Support for datetimes in st.slider. And, of course, just like any other value you use in st.slider, you can also pass in two-element lists to get a datetime range slider.

Version 0.62.0

Release date: June 21, 2020


  • 📨 Ability to turn websocket compression on/off via the config option server.enableWebsocketCompression. This is useful if your server strips HTTP headers and you do not have access to change that behavior.

  • 🗝️ Out-of-the-box support for CSRF protection using the Cookie-to-header token technique. This means that if you’re serving your Streamlit app from multiple replicas you’ll need to configure them to to use the same cookie secret with the server.cookieSecret config option. To turn XSRF protection off, set server.enableXsrfProtection=false.

Notable bug fixes:

  • 🖼️ Added a grace period to the image cache expiration logic in order to fix multiple related bugs where images sent with st.image or st.pyplot were sometimes missing.

Version 0.61.0

Release date: June 2, 2020


  • 📅 Support for date ranges in st.date_picker. See docs for more info, but the TLDR is: just pass a list/tuple as the default date and it will be interpreted as a range.

  • 🗣️ You can now choose whether st.echo prints the code above or below the output of the echoed block. To learn more, refer to the code_location argument in the docs.

  • 📦 Improved @st.cache support for Keras models and Tensorflow saved_models.

Version 0.60.0

Release date: May 18, 2020


  • ↕️ Ability to set the height of an st.text_area with the height argument (expressed in pixels). See docs for more.

  • 🔡 Ability to set the maximimum number of characters allowed in st.text_area or st.text_input. Check out the max_chars argument in the docs.

  • 🗺️ Better DeckGL support for the H3 geospatial indexing system. So now you can use things like H3HexagonLayer in st.pydeck_chart.

  • 📦 Improved @st.cache support for PyTorch TensorBase and Model.

Version 0.59.0

Release date: May 05, 2020


  • 🎨 New color-picker widget! Use it with st.beta_color_picker()

  • 🧪 Introducing st.beta_* and st.experimental_* function prefixes, for faster Streamlit feature releases. See docs for more info.

  • 📦 Improved @st.cache support for SQL Alchemy objects, CompiledFFI, PyTorch Tensors, and builtins.mappingproxy.

Version 0.58.0

Release date: April 22, 2020


  • 💼 Made st.selectbox filtering case-insensitive.

  • ㈬ Better support for Tensorflow sessions in @st.cache.

  • 📊 Changed behavior of st.pyplot to auto-clear the figure only when using the global Matplotlib figure (i.e. only when calling st.pyplot() rather than st.pyplot(fig)).

Version 0.57.0

Release date: March 26, 2020


  • ⏲️ Ability to set expiration options for @st.cache’ed functions by setting the max_entries and ttl arguments. See docs.

  • 🆙 Improved the machinery behind st.file_uploader, so it’s much more performant now! Also increased the default upload limit to 200MB (configurable via server.max_upload_size).

  • 🔒 The server.address config option now binds the server to that address for added security.

  • 📄 Even more details added to error messages for @st.cache for easier debugging.

Version 0.56.0

Release date: February 15, 2020


  • 📄 Improved error messages for st.cache. The errors now also point to the new caching docs we just released. Read more here!

Breaking changes:

  • 🐍 As announced last month, Streamlit no longer supports Python 2. To use Streamlit you’ll need Python 3.5 or above.

Version 0.55.0

Release date: February 4, 2020


  • 📺 Ability to record screencasts directly from Streamlit! This allows you to easily record and share explanations about your models, analyses, data, etc. Just click ☰ then “Record a screencast”. Give it a try!

Version 0.54.0

Release date: January 29, 2020


  • ⌨️ Support for password fields! Just pass type="password" to st.text_input().

Notable fixes:

  • ✳️ Numerous st.cache improvements, including better support for complex objects.

  • 🗣️ Fixed cross-talk in sidebar between multiple users.

Breaking changes:

  • If you’re using the SessionState hack Gist, you should re-download it! Depending on which hack you’re using, here are some links to save you some time:

Version 0.53.0

Release date: January 14, 2020


  • 🗺️ Support for all DeckGL features! Just use Pydeck instead of st.deck_gl_chart. To do that, simply pass a PyDeck object to st.pydeck_chart, st.write, or magic.

    Note that as a preview release things may change in the near future. Looking forward to hearing input from the community before we stabilize the API!

    The goals is for this to replace st.deck_gl_chart, since it is does everything the old API did and much more!

  • 🆕 Better handling of Streamlit upgrades while developing. We now auto-reload the browser tab if the app it is displaying uses a newer version of Streamlit than the one the tab is running.

  • 👑 New favicon, with our new logo!

Notable fixes:

  • Magic now works correctly in Python 3.8. It no longer causes docstrings to render in your app.

Breaking changes:

  • Updated how we calculate the default width and height of all chart types. We now leave chart sizing up to your charting library itself, so please refer to the library’s documentation.

    As a result, the width and height arguments have been deprecated from most chart commands, and use_container_width has been introduced everywhere to allow you to make charts fill as much horizontal space as possible (this used to be the default).

Version 0.52.0

Release date: December 20, 2019


  • 📤 Preview release of the file uploader widget. To try it out just call st.file_uploader!

    Note that as a preview release things may change in the near future. Looking forward to hearing input from the community before we stabilize the API!

  • 👋 Support for emoji codes in st.write and st.markdown! Try it out with st.write("Hello :wave:").

Breaking changes:

  • 🧹 st.pyplot now clears figures by default, since that’s what you want 99% of the time. This allows you to create two or more Matplotlib charts without having to call pyplot.clf every time. If you want to turn this behavior off, use st.pyplot(clear_figure=False)

  • 📣 st.cache no longer checks for input mutations. This is the first change of our ongoing effort to simplify the caching system and prepare Streamlit for the launch of other caching primitives like Session State!

Version 0.51.0

Release date: November 30, 2019


  • 🐕 You can now tweak the behavior of the file watcher with the config option server.fileWatcherType. Use it to switch between:

    • auto (default) : Streamlit will attempt to use the watchdog module, and falls back to polling if watchdog is not available.

    • watchdog : Force Streamlit to use the watchdog module.

    • poll : Force Streamlit to always use polling.

    • none : Streamlit will not watch files.

Notable bug fixes:

  • Fix the “keyPrefix” option in static report sharing #724

  • Add support for getColorX and getTargetColorX to DeckGL Chart #718

  • Fixing Tornado on Windows + Python 3.8 #682

  • Fall back on webbrowser if xdg-open is not installed on Linux #701

  • Fixing number input spin buttons for Firefox #683

  • Fixing CTRL+ENTER on Windows #699

  • Do not automatically create credential file when in headless mode #467

Version 0.50.1

Release date: November 10, 2019


  • 👩‍🎓 SymPy support and ability to draw mathematical expressions using LaTeX! See st.latex, st.markdown, and st.write.

  • 🌄 You can now set config options using environment variables. For example, export STREAMLIT_SERVER_PORT=9876.

  • 🐱 Ability to call streamlit run directly with Github and Gist URLs. No need to grab the “raw” URL first!

  • 📃 Cleaner exception stack traces. We now remove all Streamlit-specific code from stack traces originating from the user’s app.

Version 0.49.0

Release date: October 23, 2019


  • 💯 New input widget for entering numbers with the keyboard: st.number_input()

  • 📺 Audio/video improvements: ability to load from a URL, to embed YouTube videos, and to set the start position.

  • 🤝 You can now (once again) share static snapshots of your apps to S3! See the S3 section of streamlit config show to set it up. Then share from top-right menu.

  • ⚙️ Use server.baseUrlPath config option to set Streamlit’s URL to something like

Notable bug fixes:

Version 0.48.0

Release date: October 12, 2019


  • 🔧 Ability to set config options as command line flags or in a local config file.

  • ↕️ You can now maximize charts and images!

  • ⚡ Streamlit is now much faster when writing data in quick succession to your app.

  • ✳️ Ability to blacklist folder globs from “run on save” and @st.cache hashing.

  • 🎛️ Improved handling of widget state when Python file is modified.

  • 🙈 Improved HTML support in st.write and st.markdown. HTML is still unsafe, though!

Notable bug fixes:

  • Fixes @st.cache bug related to having your Python environment on current working directory. Issue #242

  • Fixes loading of root url / on Windows. Issue #244

Version 0.47.0

Release date: October 1, 2019


  • 🌄 New showing off 4 glorious Streamlit apps. Try it out!

  • 🔄 Streamlit now automatically selects an unused port when 8501 is already in use.

  • 🎁 Sidebar support is now out of beta! Just start any command with st.sidebar. instead of st.

  • ⚡ Performance improvements: we added a cache to our websocket layer so we no longer re-send data to the browser when it hasn’t changed between runs

  • 📈 Our “native” charts st.line_chart, st.area_chart and st.bar_chart now use Altair behind the scenes

  • 🔫 Improved widgets: custom st.slider labels; default values in multiselect

  • 🕵️‍♀️ The filesystem watcher now ignores hidden folders and virtual environments

  • 💅 Plus lots of polish around caching and widget state management

Breaking change:

  • 🛡️ We have temporarily disabled support for sharing static “snapshots” of Streamlit apps. Now that we’re no longer in a limited-access beta, we need to make sure sharing is well thought through and abides by laws like the DMCA. But we’re working on a solution!

Version 0.46.0

Release date: September 19, 2019


  • ✨ Magic commands! Use st.write without typing st.write. See

  • 🎛️ New st.multiselect widget.

  • 🐍 Fixed numerous install issues so now you can use pip install streamlit even in Conda! We’ve therefore deactivated our Conda repo.

  • 🐞 Multiple bug fixes and additional polish in preparation for our launch!

Breaking change:

  • 🛡️ HTML tags are now blacklisted in st.write/st.markdown by default. More information and a temporary work-around at:

Version 0.45.0

Release date: August 28, 2019


  • 😱 Experimental support for sidebar! Let us know if you want to be a beta tester.

  • 🎁 Completely redesigned st.cache! Much more performant, has a cleaner API, support for caching functions called by @st.cached functions, user-friendly error messages, and much more!

  • 🖼️ Lightning fast st.image, ability to choose between JPEG and PNG compression, and between RGB and BGR (for OpenCV).

  • 💡 Smarter API for st.slider, st.selectbox, and

  • 🤖 Automatically fixes the Matplotlib backend – no need to edit .matplotlibrc

Version 0.44.0

Release date: July 28, 2019


  • ⚡ Lightning-fast reconnect when you do a ctrl-c/rerun on your Streamlit code

  • 📣 Useful error messages when the connection fails

  • 💎 Fixed multiple bugs and improved polish of our newly-released interactive widgets

Version 0.43.0

Release date: July 9, 2019


  • ⚡ Support for interactive widgets! 🎈🎉

Version 0.42.0

Release date: July 1, 2019


  • 💾 Ability to save Vega-Lite and Altair charts to SVG or PNG

  • 🐇 We now cache JS files in your browser for faster loading

  • ⛔ Improvements to error-handling inside Streamlit apps

Version 0.41.0

Release date: June 24, 2019


  • 📈 Greatly improved our support for named datasets in Vega-Lite and Altair

  • 🙄 Added ability to ignore certain folders when watching for file changes. See the server.folderWatchBlacklist config option.

  • ☔ More robust against syntax errors on the user’s script and imported modules

Version 0.40.0

Release date: June 10, 2019


  • Streamlit is more than 10x faster. Just save and watch your analyses update instantly.

  • We changed how you run Streamlit apps: $ streamlit run [script args]

  • Unlike the previous versions of Streamlit, streamlit run [script] [script args] creates a server (now you don’t need to worry if the proxy is up). To kill the server, all you need to do is hit Ctrl+c.

Why is this so much faster?

Now, Streamlit keeps a single Python session running until you kill the server. This means that Streamlit can re-run your code without kicking off a new process; imported libraries are cached to memory. An added bonus is that st.cache now caches to memory instead of to disk.

What happens if I run Streamlit the old way?

If you run $ python the script will execute from top to bottom, but won’t produce a Streamlit app.

What are the limitations of the new architecture?

  • To switch Streamlit apps, first you have to kill the Streamlit server with Ctrl-c. Then, you can use streamlit run to generate the next app.

  • Streamlit only works when used inside Python files, not interactively from the Python REPL.

What else do I need to know?

  • The strings we print to the command line when liveSave is on have been cleaned up. You may need to adjust any RegEx that depends on those.

  • A number of config options have been renamed:

    Old config

    New config





    proxy.runOnSave, proxy.watchFileSystem
























What if something breaks?

If the new Streamlit isn’t working, please let us know by Slack or email. You can downgrade at any time with these commands:

$ pip install --upgrade streamlit==0.37
$ conda install streamlit=0.37

What’s next?

Thank you for staying with us on this journey! This version of Streamlit lays the foundation for interactive widgets, a new feature of Streamlit we’re really excited to share with you in the next few months.

Version 0.36.0

Release date: May 03, 2019


  • 🚣‍♀️ st.progress() now also accepts floats from 0.0–1.0

  • 🤯 Improved rendering of long headers in DataFrames

  • 🔐 Shared apps now default to HTTPS

Version 0.35.0

Release date: April 26, 2019


  • 📷 Bokeh support! Check out docs for st.bokeh_chart

  • ⚡️ Improved the size and load time of saved apps

  • ⚾️ Implemented better error-catching throughout the codebase