Reactor

Reactor provides two sibling packages in one repository, with no python/typescript subpackage folders:

• datalayer_reactor (Python): FastAPI + pluggy plugin platform for modular and SaaS-style extensibility.
• @datalayer/reactor (TypeScript): Lexical-inspired extension runtime with a framework-agnostic core and separate React integration.

Why Reactor

This project targets a full extension platform, not only hook callbacks:

• Platform architecture with lifecycle phases and dependency graph
• Plugin marketplace metadata and discovery primitives
• Third-party ecosystem support through explicit manifest contracts
• Dynamic feature loading and runtime enable/disable
• Modular app concerns: interdependencies, lifecycle management, compatibility checks
• SaaS extensibility primitives: tenant-specific plugin activation, sandboxed execution, versioned compatibility

Repository Layout

• src/: TypeScript package source for @datalayer/reactor
• datalayer_reactor/: Python package source
• examples/frontend/: frontend-only React + Primer UI demo
• examples/frontend-backend/: combined React + Python demo

TypeScript Package: @datalayer/reactor

Lexical-style design applied

The TypeScript runtime implements a similar model to Lexical Extensions docs:

• defineExtension and configExtension
• dependencies, peerDependencies, conflictsWith
• ordered phases: init -> build -> register -> afterRegistration
• runtime lifecycle control: start, stop, enable, disable
• signal primitives for reactive extension outputs:
  • signal, computed, effect, batch, untracked
  • namedSignals, watchedSignal

Core vs React split

• Core runtime exports from @datalayer/reactor
• React bindings export from @datalayer/reactor/react

React bindings include:

• ReactorProvider: platform lifecycle in React tree
• ReactorSlot: render plugin-provided components by named slot
• useReactorPlatform: platform access for runtime toggles

Build

npm install
npm run build

Minimal TypeScript usage

import { buildPlatformFromExtensions, defineExtension } from '@datalayer/reactor';

const DemoExtension = defineExtension({
	name: '@demo/core',
	build() {
		return { message: 'hello' };
	},
});

const platform = buildPlatformFromExtensions([DemoExtension]);
platform.start();

Python Package: datalayer_reactor

Capabilities

• Pluggy-powered plugin registration (register_plugin)
• Compatibility and dependency checks via PluginManifest
• Runtime enable/disable globally and by tenant
• Marketplace publication/listing (PluginMarketplace)
• Sandboxed execution option for plugin calls
• FastAPI control plane with plugin/tenant endpoints

Install and run

python -m venv .venv
source .venv/bin/activate
pip install -e .
python -m datalayer_reactor

API app endpoints

• GET /plugins
• POST /plugins/{plugin_name}/toggle
• POST /tenants/plugins/{plugin_name}/toggle
• GET /tenants/{tenant_id}/features
• GET /tenants/{tenant_id}/routes
• GET /marketplace

Minimal Python usage

from datalayer_reactor import PluginManifest, PluginCompatibility, PluginPlatform

platform = PluginPlatform()
platform.register_plugin(
		PluginManifest(
				name="greeting-plugin",
				version="1.0.0",
				compatibility=PluginCompatibility(api_version="v1"),
		),
		plugin_impl=object(),
)

Example: Two Primer React Plugins

The frontend demo at examples/frontend/ contains two plugins:

• @demo/welcome-card
• @demo/status-banner

It renders them through ReactorSlot and exposes runtime enable/disable controls.

Run it:

npm install
npm run example:dev

Run the combined frontend-backend demo frontend:

npm run example:dev:frontend-backend

Python Demo

Run the Python app for the combined frontend-backend demo:

python -m uvicorn --app-dir examples/frontend-backend python_platform_demo:app --reload --port 8788

Next Steps

• Add a persisted marketplace backend (database + signed plugin artifacts)
• Add stronger sandbox isolation (subprocess/container boundaries)
• Add semantic version range support (^, ~) for compatibility contracts
• Add plugin state migrations for compatibility layer upgrades