Declarative, typed, URL-aware state management for complex Streamlit apps.
st-page-state is an architectural pattern for Streamlit that replaces loose dictionary keys with strict, typed classes. It solves the friction of managing complex state, URL synchronization, and component communication in large-scale applications.
Stop writing boilerplate. Turn messy, hard-to-maintain state logic into clean, declarative code.
The old way is brittle, requires manual URL parsing, and lacks type safety.
import streamlit as st
# 1. Initialize logic scattered across the script
if "status" not in st.session_state:
# 2. Manual URL parsing
url_val = st.query_params.get("status", "pending")
st.session_state["status"] = url_val
# 3. No type safety
status = st.session_state["status"]
# 4. Manual URL updating on change
def update_status():
st.query_params["status"] = st.session_state["status"]
st.selectbox("Status", ["pending", "active"], key="status", on_change=update_status)Declarative, typed, and automatically synchronized with the URL.
from st_page_state import PageState, StateVar
class FilterState(PageState):
status: str = StateVar(default="pending", url_key="status")
st.selectbox("Status", ["pending", "active"], **FilterState.bind("status"))The #1 reason to use this library: st.session_state dies on page refresh.
When a user filters a app and shares the URL, their colleague sees a completely different view. st-page-state solves this by making the URL the source of truth.
- Shareable URLs: Copy the URL, send it to a colleague, and they'll see the exact same state.
- Bookmarkable Views: Users can bookmark a specific filtered view and come back to it days later.
- Deep Linking: Programmatically generate links (e.g., in an email report) that open your app in a pre-configured state.
This is powered by Automatic Bi-Directional Sync:
- On Load: The app state is initialized from the URL.
- On Change: Any interaction with a bound widget instantly updates the URL.
pip install st-page-stateWith optional Redis support:
pip install st-page-state[redis](Requires Python 3.8+ and Streamlit >= 1.30)
Handling lists, sets, tuples, or dates in query parameters usually requires manual parsing logic. st-page-state handles serialization automatically using efficient Base64 encoding for lists to keep URLs clean and safe.
class SearchState(PageState):
# Serialized as Base64 JSON in URL
tags: list[str] = StateVar(default=[], url_key="tags")
# Sets work too (great for unique selections)
categories: set[int] = StateVar(default={1, 2}, url_key="cats")
# Automatically handles ISO format dates
start_date: datetime.date = StateVar(default=datetime.date.today(), url_key="start")See examples/06_list_handling.py for a full demo.
Decouple your internal logic (integers/IDs) from your public URLs (friendly strings).
class TaskState(PageState):
status: int = StateVar(
default=0,
url_key="status",
value_map={
0: "todo", # URL shows ?status=todo
1: "in_progress", # URL shows ?status=in_progress
2: "done" # URL shows ?status=done
}
)
# Your code works with clean integers
if TaskState.status == 1:
st.info("Task is in progress")Stop cluttering your UI code with side effects. Define on_init and on_change logic directly where the state lives.
class AppState(PageState):
theme: str = StateVar(default="light")
@classmethod
def on_init(cls):
"""Runs once when the session starts."""
print("State initialized.")
@classmethod
def on_change(cls, field, old_value, new_value):
"""Runs whenever a specific field changes."""
if field == "theme":
print(f"Theme changed from {old_value} to {new_value}")
# Trigger external analytics, logging, or database updates hereThe .bind() method returns the exact dictionary (key and on_change) Streamlit widgets expect.
st.text_input("Search", **SearchState.bind("query"))Sometimes you want a widget to start with a specific value without changing the global StateVar default. You can pass a value argument to .bind():
# The widget will start at 10 on the first run, even if the StateVar default is 0
st.number_input("Counter", **MyState.bind("count", value=10))See examples/08_bind_initial_value.py for a full demo.
For multi-page or complex apps, you might need finer control over the URL query string. The inner Config class provides this.
class DashboardState(PageState):
class Config:
# 1. Add a prefix to all URL keys from this class
url_prefix = "dash_"
# 2. If True (default), remove any URL params not managed by this class
url_selfish = True
# 3. If True (default), restore any missing URL params from session state on every change
restore_url_on_touch = True
# 4. List of other PageState classes (by name or Class obj) to share the URL with
share_url_with = ["OtherState"] # Or share_url_with = [OtherState]
tab: str = StateVar(default="overview", url_key="tab") # -> ?dash_tab=overview
show_details: bool = StateVar(default=False, url_key="details") # -> ?dash_details=trueurl_prefix: Prevents key collisions between differentPageStatemodels.url_selfish: (Default: True) When any variable in a selfish state changes, it clears all other query parameters from the URL. This ensures a clean URL that only reflects the active component.restore_url_on_touch: (Default: True) Guarantees URL completeness. It ensures that whenever you access (read or write) a state variable, its corresponding URL parameter is present. This restores parameters if they were missing due to:- Manual removal by the user from the browser URL.
- Clearing by another
url_selfishstate.
share_url_with: (List[str] | List[Type[PageState]]) A list of otherPageStateclass names (strings) or class objects that this state should share the URL with. Even ifurl_selfishis True, parameters belonging to the specified classes will be preserved in the URL instead of being cleared.
See examples/07_config_class.py for a full demo.
State lives in st.session_state by default β if the server restarts, it's gone. For apps that need durability, enable Redis-backed persistence with two lines:
pip install st-page-state[redis]from st_page_state import PageState, StateVar, RedisBackend
r_backend = RedisBackend(
host="localhost",
default_ttl=3600,
# Tie state to a user identity instead of Streamlit's ephemeral session ID:
session_id=lambda: st.session_state.get("user_email", "anonymous"),
)
class Counter(PageState):
count: int = StateVar(default=0, url_key="count")
with r_backend.session():
st.number_input("Count", **Counter.bind("count"))r_backend.session() loads all stored state from Redis on entry and saves it back on exit. The global default_ttl controls key expiry; override it per class via Config.ttl.
By default, Redis keys are scoped to Streamlit's ephemeral session ID (one per browser tab). Pass session_id as a string or callable to tie state to a stable user identity so it persists across tabs and restarts:
class EphemeralState(PageState):
draft: str = StateVar(default="")
class Config:
ttl = 120 # expires after 2 minutesIdentity transition β when using a callable session_id (e.g. lambda: st.session_state.get("user_email", "anonymous")), if the returned value changes between reruns (anonymous β logged-in user), the in-memory state is automatically cleared and re-loaded from Redis under the new identity. No stale data leaks across users.
URL priority β fields whose url_key appears in the current query_params are skipped during Redis LOAD, so shared URLs like ?page=settings&tab=2 always take precedence over whatever was last saved.
See examples/09_redis_persistence.py for a full demo.
Sometimes you need to claim URL focus without user interaction. The .focus() method programmatically triggers the url_selfish and restore_url_on_touch logic.
# Clears any non-DashboardState params and ensures all its own params are in the URL
DashboardState.focus()For complex apps, visibility into your state is critical.
# Get a pure dictionary of current values (great for API payloads)
payload = UserState.dump()
# Reset specific fields or the entire state to defaults
UserState.reset()We welcome contributions from the community. If you are solving complex state problems in Streamlit, we want to hear from you.
- Fork the repository
- Create your feature branch
- Install dev dependencies (
pip install -e .[dev]) - Run tests (
pytest) - Open a Pull Request
Distributed under the MIT License. See LICENSE for more information.