Introduction
Streamlit shipped a wave of features in early 2026 that make apps feel noticeably more polished — dropdown menus, loading skeletons, Mermaid diagrams, iframes, and a step-timeline mode for expanders. If you haven’t updated in a while, your apps are missing out.
This is a practical roundup: each feature with a minimal working snippet and when to use it. All examples assume pip install --upgrade streamlit (1.40+).
1. st.menu_button — dropdown action menus
Build toolbars and action lists without custom components:
import streamlit as st
choice = st.menu_button(
"Actions",
options=["Export CSV", "Export JSON", "Share link", "Delete"],
icons=[":material/download:", ":material/code:", ":material/share:", ":material/delete:"],
)
if choice == "Export CSV":
st.download_button("Download", data=csv_bytes, file_name="data.csv")
Use it for: row-level actions in tables, “more options” menus, compact toolbars. Replaces the old pattern of stacking 4 buttons horizontally.
2. st.skeleton — loading placeholders
Show animated placeholders while data loads, instead of a blank screen or spinner:
import time
with st.skeleton():
st.title("Dashboard")
st.write("Loading metrics...")
time.sleep(2)
st.metric("Revenue", "$42,000", "+12%")
Use it for: any app with a slow first load (DB queries, API calls, model inference). Pairs beautifully with @st.cache_data.
3. st.mermaid_chart — diagrams in Markdown
Render flowcharts and sequence diagrams directly:
st.mermaid_chart("""
graph TD
A[User uploads CSV] --> B[Pandas profiling]
B --> C{Missing values?}
C -->|Yes| D[Impute]
C -->|No| E[Train model]
D --> E
E --> F[Show results]
""")
Also works inside st.markdown via fenced mermaid code blocks. Use it for: architecture docs, ML pipelines, onboarding flows inside your app.
4. st.iframe — embed anything
Embed external URLs or raw HTML:
st.iframe("https://www.youtube.com/embed/dQw4w9WgXcQ", height=400)
# Or raw HTML
st.iframe("<h1 style='color:tomato'>Hello</h1>", height=100)
Use it for: embedding YouTube tutorials, Figma prototypes, external dashboards, or sandboxed widgets. Cleaner than components.html for simple cases.
5. filter_mode on selectbox/multiselect
Let users search and filter options by typing:
city = st.selectbox(
"City",
options=["Mumbai", "Delhi", "Bangalore", "Chennai", "Hyderabad", ...],
filter_mode="contains", # or "startswith"
)
Use it for: any dropdown with >10 options. Huge UX win for city/country/product pickers.
6. st.status with type="step" — agent timelines
Render a connected timeline — perfect for showing an agent’s or pipeline’s progress:
with st.status("Running analysis...", expanded=True, state="running") as status:
st.write("Step 1: Loading data")
time.sleep(1)
st.write("Step 2: Cleaning")
time.sleep(1)
st.write("Step 3: Training model")
time.sleep(1)
status.update(label="Done!", state="complete", expanded=False)
Add type="step" (on expander/status) for a connected timeline look. Use it for: multi-step AI agents, ETL pipelines, and any “here’s what I’m doing” feedback.
7. on_change="ignore" — buffer widget updates
Update a widget in the browser without rerunning the app; the value arrives on the next rerun (e.g., a button click):
name = st.text_input("Name", on_change="ignore")
if st.button("Submit"):
st.write(f"Hello, {name}!") # reads the buffered value
Use it for: forms where you don’t want a rerun on every keystroke. Huge performance win for text-heavy inputs.
Bonus: small but nice
st.expanderaccepts icons and renders cleaner.st.dataframegot faster rendering for large frames.- Theme tweaks: advanced theming + custom fonts landed late 2025 and matured in 2026.
Migration checklist
pip install --upgrade streamlit- Replace stacked button rows with
st.menu_button. - Wrap slow first loads in
st.skeleton. - Add
filter_mode="contains"to any long selectbox. - Swap
st.spinnerforst.status+type="step"in multi-step flows. - Add
on_change="ignore"to text inputs in forms.
Key Concepts
st.menu_button= compact action menusst.skeleton= perceived-performance winst.mermaid_chart= diagrams as codest.iframe= embed anythingfilter_mode= searchable dropdownsst.status+ steps = agent/pipeline timelines
What to Try Next
- Upgrade your chatbot with
st.statussteps showing “thinking → tool call → answer”. - Add
st.skeletonto your data dashboard for a snappier first load. - Embed a YouTube tutorial in your Streamlit course pages with
st.iframe.
FAQ
Do these features work on Streamlit Community Cloud?
Yes — Community Cloud runs the latest Streamlit release. Just update requirements.txt to streamlit>=1.40.
Will st.menu_button replace st.button?
No — st.button is still right for primary actions. st.menu_button is for secondary/overflow actions.
Is on_change="ignore" stable?
It shipped in 1.40+ and is stable. Older versions will error; pin your Streamlit version in production.