Convex Push Notifications Component

This is a Convex component that integrates with Expo's push notification API to allow sending mobile push notifications to users of your app. It will batch calls to Expo's API and handle retrying delivery.

Demo GIF

Example usage:

// App.tsx
<Button
  onPress={() => {
    void convex.mutation(api.example.sendPushNotification, {
      to: otherUser,
      title: `Hi from ${currentUser.name}`,
    });
  }}
>
  <Text>Say hi!</Text>
</Button>

// convex/example.ts
export const sendPushNotification = mutation({
  args: { title: v.string(), to: v.id("users") },
  handler: async (ctx, args) => {
    // Sending a notification
    return pushNotifications.sendPushNotification(ctx, {
      userId: args.to,
      notification: {
        title: args.title,
      },
    });
  },
});

Pre-requisite: Convex

You'll need an existing Convex project to use the component. Convex is a hosted backend platform, including a database, serverless functions, and a ton more you can learn about here.

Run npm create convex or follow any of the quickstarts to set one up.

Installation

Install the component package:

npm i @convex-dev/expo-push-notifications

Create a convex.config.ts file in your app's convex/ folder and install the component by calling use:

// convex/convex.config.ts
import { defineApp } from "convex/server";
import pushNotifications from "@convex-dev/expo-push-notifications/convex.config.js";

const app = defineApp();
app.use(pushNotifications);
// other components

export default app;

Instantiate the PushNotifications client in your Convex functions:

// convex/example.ts
import { PushNotifications } from "@convex-dev/expo-push-notifications";

const pushNotifications = new PushNotifications(components.pushNotifications);

It takes in an optional type parameter (defaulting to Id<"users">) for the type to use as a unique identifier for push notification recipients:

import { PushNotifications } from "@convex-dev/expo-push-notifications";

export type Email = string & { __isEmail: true };

const pushNotifications = new PushNotifications<Email>(
  components.pushNotifications,
);

Registering a user for push notifications

Get a user's push notification token following the Expo documentation here, and record it using a Convex mutation:

// convex/example.ts
export const recordPushNotificationToken = mutation({
  args: { token: v.string() },
  handler: async (ctx, args) => {
    const userId = await getAuthUserId(ctx);
    await pushNotifications.recordToken(ctx, {
      userId,
      pushToken: args.token,
    });
  },
});

You can pause and resume push notification sending for a user using the pausePushNotifications and resumePushNotifications methods.

To determine if a user has a token and their pause status, you can use getStatusForUser.

Send notifications

// convex/example.ts
export const sendPushNotification = mutation({
  args: { title: v.string(), to: v.string() },
  handler: async (ctx, args) => {
    const pushId = await pushNotifications.sendPushNotification(ctx, {
      userId: args.to,
      notification: {
        title: args.title,
      },
    });
  },
});

You can use the ID returned from sendPushNotifications to query the status of the notification using getNotification. Using this in a query allows you to subscribe to the status of a notification.

You can also view all notifications for a user with getNotificationsForUser.

How delivery works

1. How individual notifications are inserted

   sendPushNotification and sendPushNotificationBatch first look up the user's Expo token. If the user has no token, the component either throws or returns null depending on allowUnregisteredTokens. If the user's token is paused, the notification is dropped and null is returned. Otherwise the component inserts a row into the notifications table with state = "awaiting_delivery", a failure count of 0, and a short time-bucket marker used for batching.

2. How the batcher stays alive while work remains

   Every successful insert calls an internal helper that ensures exactly one future batch run is scheduled for the earliest eligible notification segment. If a newly inserted notification becomes eligible sooner than the currently scheduled run, the component cancels the later wake-up and reschedules it earlier. When that batch run wakes up, it either: finds work and enqueues one or more workpool jobs, or finds no work and clears the pending batch-run marker. After each workpool job completes, the component schedules the next batch pass again if there is still work left.

3. How workpool threads parallelize the work

   Each batch becomes one workpool job. The workpool limits how many of those jobs may run at once via maxParallelism, and it keeps its own pending and running bookkeeping internally. Before a batch is handed to the workpool, the notifications in that batch are marked in_progress, which prevents the component from selecting the same notification for another batch at the same time.

4. How partial failures are reflected

   Expo can accept some notifications in a batch and reject others in the same response. The component records that result per notification: delivered notifications are finalized as delivered, non-retryable ticket failures are finalized as failed, and documented temporary ticket failures such as MessageRateExceeded are moved to needs_retry. Failed notifications are not retried inline in the same HTTP request; they are picked up by a later batch pass. Request-level failures are handled separately by the workpool retry policy.

5. How temporary failures are deferred

   There are two retry layers. If the whole Expo request fails, the workpool retries that batch with exponential backoff. If Expo returns a successful batch response with some individual temporary message failures, those notifications are deferred by moving them back into the queue as needs_retry and advancing their next eligible segment using exponential backoff. The next batch pass prioritizes needs_retry notifications over brand new ones once their scheduled segment becomes eligible, and it sleeps until that exact segment instead of polling.

6. How shutdown and paused tokens interact

   Paused tokens short-circuit before insertion, so paused users do not create queued notifications. shutdown() cancels future batch scheduling and cancels pending workpool jobs that have not started yet. Batches already marked in_progress are allowed to drain, and restart() resumes scheduling once those in-flight notifications have finished.

Troubleshooting

To add more logging, provide PushNotifications with a logLevel in the constructor:

const pushNotifications = new PushNotifications(components.pushNotifications, {
  logLevel: "DEBUG",
});

The push notification sender can be shutdown gracefully, and then restarted using the shutdown and restart methods.