Observ 👁

Vue.js-style reactivity for Python. Wrap your state in a reactive proxy, and observ tracks every read and write for you: computed values invalidate themselves, watchers fire when something they depend on changes. No dirty flags, no manual invalidation, no dependency bookkeeping.

📖 Documentation · 📦 PyPI · 🧪 Examples (Qt, rendercanvas)

🖼️ Building a GUI?

Head straight to Collagraph — a full-featured, Vue-inspired declarative UI framework built on top of observ. If you're looking for a strong reactive GUI framework for your Python Qt/PySide applications, Collagraph is the main entrypoint; observ is the reactivity engine underneath it.

Why observ?

• Automatic dependency tracking — computed state knows exactly what it depends on and lazily re-evaluates only when needed.
• React to any change — watch plain state or computed state and build unidirectional data flow: state changes drive view changes, input events drive state changes.
• Zero dependencies, framework agnostic — a pure Python library (≥ 3.13) that plugs into any event loop: asyncio, Qt, or your own.
• Fully typed — a PEP 561 typed package, checked with ty in CI.

Quick start

pip install observ

from observ import computed, reactive, watch

state = reactive({
    "todos": [
        {"title": "groceries", "done": False},
        {"title": "dishes", "done": True},
    ],
})

@computed
def progress():
    done = sum(todo["done"] for todo in state["todos"])
    return f"{done}/{len(state['todos'])} done"

watcher = watch(progress, lambda new: print(new), sync=True)

# Mutate the plain dicts and lists you already have —
# observ sees every change, no matter how deeply nested:
state["todos"][0]["done"] = True                            # prints: 2/2 done
state["todos"].append({"title": "laundry", "done": False})  # prints: 2/3 done

# ...but you only ever react when a *result* actually changes:
state["todos"][1]["title"] = "do the dishes"                # (no print)

No subclasses to inherit, no observable fields to declare, no signals to wire up: reactive() takes your existing data, and dependencies are tracked automatically simply by using it.

Continue with the Quick Start tutorial, or dive into the Guide and API Reference.

Related projects

• Collagraph: reactive user interfaces in Python, built on top of observ.
• Reliev: the store (with undo/redo support) that used to be part of observ.
• Patchdiff: diffing and patching for plain Python data structures — the same dicts, lists and sets you'd wrap with observ.