Skip to main content



Actions is the standard way to perform any side-effects in the application, including:

  • making requests
  • dispatching store events
  • redirects and navigation
  • logging and analytics

There is two ways to execute actions - automatically (we call it global actions, usually it is page actions) and manually, in React components or nested calls inside another actions.

Actions have a lot of optimizations for achieving the best latency for application server responses.


You can find declareAction interface in @tramvai/core package documentation


Actions can be executed at different stages:

  • Server-side, for every request
  • Client-side, after page initialization
  • Client-side, after SPA transition
  • Anywhere manually

By default, all global actions will be executed by router on the server in parallel, at resolvePageDeps commandLineRunner stage, before page rendering.

This behaviour can be changed by actions conditions or in the case of exceeded execution time or errors

Client-side actions behaviour is visualized in Navigation Flow page

Global actions

In short, action is global if added via createApp or a static property of a page component, such as MainPage.actions. Only for this kind of actions some features will be available, for example, parallel execution and execution deadline.

You can imagine global actions as flat list of async operations, and for every page their own list will be created and executed for every request.

Also, some conditions will be applied only to global actions.


Execution Deadline

Servers must respond quickly, so we must reduce the number of cases when global actions cause a delay in page loading, for example, if an API fails. To do this, there is a time limit (500ms) on the server for executing global actions, and if this time passes, then waiting for actions ends and these actions must be executed on the client side.

Synchronizing between server and client

Information about all successfully executed actions will be transferred to the client, which will start the execution of global actions based on this information. At the same time, if an action, for example, fell out of the deadline or fell by mistake, then it will be re-executed on the client side.


Unique actions names are required for correct synchronization between server and client


  • For API calls, always try to separate the business and requests logic from actions in favour of specific services, which can be used in actions or React-Query
  • Actions is the good place to combine fetching and state management logic
  • Try to create actions with unique names, because they will be used for synchronization between server and client
  • Subscribe to abortSignal if you make requests in actions

Quick Start

⌛ Create a new action:

import { declareAction } from '@tramvai/core';

const logAction = declareAction({
name: 'log',
fn() {

⌛ Register action for page component:

import type { PageComponent } from '@tramvai/react';

const MainPage: PageComponent = () => <h1>Main Page</h1>;

MainPage.actions = [logAction];

export default MainPage;

This action will be executed only for this page and only on the server-side (because of synchronization).


Page actions

The most common way to use actions - is to fetch some data from API for specific page.

You need to use actions static property of page components to declare global actions which will be executed only for this page.

Actions execution result will not be connected with page directly, and you need to create specific reducer for this data as transport layer between action and page component.

Page actions are aborted on route changing, so you may control your action behavior via abortSignal.


React-Query is the great tool to reduce boilerplate for fetching data with interface similar to declareAction

Example with data fetching and custom reducer, let's imagine that we create a chat page:

import type { PageComponent } from '@tramvai/react';
import { createReducer, createEvent, useStore } from '@tramvai/state';
import { declareAction } from '@tramvai/core';

export const messagesLoaded = createEvent('messages loaded');

// data will be saved here on the server and transferred to the client automatically
export const MessagesStore = createReducer<Message[]>('messages', []).on(
(state, payload) => payload

// action will be executed at server-side
const loadMessagesAction = declareAction({
name: 'load-messages',
async fn() {
const messages = await loadMessagesFromApi();


// it's also possible to subscribe on pageAction abortion due to route changing
this.abortSignal.addEventListener('abort', () => {
this.context.dispatch(/* ... */);

const ChatPage: PageComponent = () => {
// page will be rendered with the same data on the server and hydrated on the client
const messages = useStore(MessagesStore);

return (
{ => (
<Message key={} text={message.text} />

// reducer and action will be registered only for this page
ChatPage.reducers = [MessagesStore];
ChatPage.actions = [loadMessagesAction];

export default ChatPage;

Application actions

Another way to register global actions is to register them in the application level in createApp method. This actions will be executed for every application page:

name: 'awesome-app',
actions: [loadSomeGlobalConfiguration],

Also you can register application-wide action with ACTIONS_LIST_TOKEN - it is useful when you want to separate some global logic in independed module:

import { provide } from '@tramvai/core';
import { ACTIONS_LIST_TOKEN } from '@tramvai/core';

const provider = provide({
multi: true,
useValue: [loadSomeGlobalConfiguration],

Execute actions in React components

Use useActions hook if you want to execute actions manually.


When you declare action, you can declare fn(...args: any[]) property with any amount of arguments, and then you can pass it to action when you execute it:

import { declareAction } from '@tramvai/core';
import { useActions } from '@tramvai/state';

const loadProductAction = declareAction({
name: 'load-product',
// declare typed params
async fn(id: number, timeout = 1000) {
return fetchProductById(id, { timeout });

const ProductCard = ({ id }: { id: number }) => {
const [product, setProduct] = useState(null);
// bind action to the context
const loadProduct = useActions(loadProductAction);

useEffect(() => {
// pass nessesary params
loadProduct(id).then((response) => {
}, [id]);

return <>...</>;


ConsumerContext is available in the action fn function in this context.

With Context for example you can execute nested actions inside global, or dispatch store events:

import { declareAction } from '@tramvai/core';
import type { PageComponent } from '@tramvai/react';

const nestedAction = declareAction({
name: 'nested',
async fn() {
await doAsyncStuff();

const rootAction = declareAction({
name: 'root',
async fn() {
await this.executeAction(nestedAction);


const MainPage: PageComponent = () => <h1>Main Page</h1>;

MainPage.actions = [rootAction];

export default MainPage;


Actions has full Dependency Injection support, so you can declare dependencies like in DI providers, in deps property. This dependencies will be available in the action fn function in this.deps property.

For example, let's use tramvai logger inside action:

import { declareAction } from '@tramvai/core';
import { LOGGER_TOKEN } from '@tramvai/tokens-common';

const logAction = declareAction({
name: 'log',
fn() {
const { logger } = this.deps;'logged!');
deps: {


Any actions has their own execution context, which is regulated by the AbortController API. What does it mean for application developers:

  • You can subscribe to parent abort signal through this.abortSignal in action fn function
  • You can cancel requests in nested actions through this.abortController.abort() in action fn function

Subscription to this.abortSignal is important, because when execution timeout will be exceeded, this actions will be ignored, and can make some unnecessary work on the server-side. Also it allows to control nested actions.

For example, we want to subscribe this.abortSignal, and force some deadline for nested action:

const innerAction = declareAction(/** ... */);

const action = declareAction({
name: 'root',
async fn() {
// after timeout exceeds, first request or request in innerAction will be aborted
setTimeout(() => {
}, 1000);

const { payload } = await this.deps.httpClient.request({
url: '',
// pass signal to the request
signal: this.abortSignal,

// if innerAction1 ends after abortController.abort was called
// then calling innerAction2 will throw an instance of ExecutionAbortError
await this.executeAction(innerAction, payload);
deps: {
httpClient: HTTP_CLIENT,


Not all actions can be executed under all circumstances, i.e. we can have actions that should be executed only on the server, others only in the browser, and having any other conditions. There is a conditions property to solve this problem:

name: 'fetch-data',
fn() {
return fetch(payload.url);
conditions: {
onlyBrowser: true,

In the example above, we create an action that will be executed only in the browser.

Adding new conditions to the application

You can implement your own execution conditions in an application or module. To do this, we must create an object with an interface:

interface Condition {
key: string;
fn(checker: ActionConditionChecker): void;
  • key - restriction identifier
  • fn - a validation function that will be called for each action

The function will receive in the argument checker, which has an interface

interface ActionConditionChecker {
payload: any;
parameters: any;
type: 'global' | 'local';
conditions: Record<string, any>;
forbid(): void;
setState(value: any): void;
getState(): any;
allow(): void;
  • payload - data that was transferred with the action
  • parameters - parameters that were passed when creating the action
  • conditions - conditions for the current action
  • type - type of the executed action, can be global or simple execution via executeAction
  • forbid - prohibits the execution of the action. If at least one checker calls this function, the action execution will be stopped
  • setState - allows you to write the check data. It is necessary for cases when we need to know with what data it was executed before and whether it needs to be repeated, for example, conditions on the authorization role
  • getState - get the previously recorded state
  • allow - tell the action to be executed always. The action will execute unless execution is forbidden via forbid

Example of a constraint

const isServer = typeof window === 'undefined';

export const onlyServer: ActionCondition = {
key: 'onlyServer',
fn: (checker) => {
if (checker.conditions.onlyServer && !isServer) {

After connecting, the constraint will look if the action has a onlyServer field in conditions, and if so, it will change the action's behavior

Connecting conditions to the app

To do this, you need to add the multi provider ACTION_CONDITIONALS and pass a function that will have an interface

import { provide } from '@tramvai/core';

const provider = provide({
multi: true,
useValue: [onlyServer],

Predefined presets

  • dynamic - the action is executed on the server and on each SPA transition within the application
  • always - the action is executed on the server, then in the browser and on each SPA transition within the application - completely without caches
  • onlyBrowser - the action is executed only in the browser
  • onlyServer - the action is executed only on the server
  • pageBrowser - the global action is executed only in the browser
  • pageServer - the global action is executed only on the server
  • dynamic + onlyBrowser - the action is executed in the browser and for each SPA transition within the application

Dynamic parameters and SPA-transitions

Keep in mind that actions are cached by default and are only executed once during the life cycle of the application.

The following feature follows from this.

Let's assume the following situation:

  • we have a page of the concert venue at the url /concertvenue-[objectId] - where objectId is a parameter that corresponds to the concert venue identifier;
  • on this page we have one component ConcertVenuePage and one page action preparePageAction;
  • objectId in the url of the page is used to get data in preparePageAction, as well as to fetch data for rendering the page;
  • we have a concert page /concert on which there are links to concert venues -/concertvenue-1, /concertvenue-2, /concertvenue-1. We can navigate to all these links with a SPA transition;
  • Transitions between pages are client-side (SPA), not server-side;


  1. On the concert page we click on /concertvenue-1, the page of the concert venue opens, the page action is performed for the first time.
  2. We go back to the SPA concert page by transition.
  3. Click on /concertvenue-2.
  4. We get to an empty page, since the page action has already been executed, new data has not been requested, and the data selection for drawing the page was made according to ID from url - 2.

If you want a page action to be executed every time you visit the page, you need to pass it the appropriate condition:

const preparePageAction = creareAction({
name: 'preparePageAction',
fn: () => {
// ...
conditions: {
// with dynamic: true, the action will always be called for SPA-transitions and not cached
dynamic: true,

ConcertVenuePage.actions = [preparePageAction];

Also, you can't pass parameters to the global actions, because they are called automatically. Better way is to get nessessary parameter from DI, for example if action depends on route dynamic parameter, you can use PAGE_SERVICE_TOKEN:

const preparePageAction = creareAction({
name: 'preparePageAction',
async fn() {
const { pageService } = this.deps;
const { id } = pageService.getCurrentRoute().params;

await loadConcert(id);
conditions: {
dynamic: true,

ConcertVenuePage.actions = [preparePageAction];