如何通过单次调用FCM API向多个用户发送不同消息？批量高频个性化推送场景咨询

Great question—this is a super common pain point when you need personalized, dynamic FCM notifications at scale. Let’s break down the best solution that fits all your constraints:

Why This Works #

FCM's HTTP v1 API supports batch requests, which let you pack multiple independent FCM messages (each with unique content and target tokens) into a single HTTP call. This cuts your API requests from 1000 per minute to just 2 (since FCM caps batch requests at 500 sub-requests each) — way more efficient, and perfect for your dynamic user groups and personalized content needs.

Step-by-Step Implementation #

1. Prepare Your Batch Request Structure #

Each batch request uses a multipart/mixed format with a unique boundary string to separate individual FCM message requests. Here’s a simplified example:

POST /batch HTTP/1.1
Host: fcm.googleapis.com
Authorization: Bearer <YOUR_SERVICE_ACCOUNT_ACCESS_TOKEN>
Content-Type: multipart/mixed; boundary=unique_batch_boundary_123

--unique_batch_boundary_123
Content-Type: application/http
Content-Transfer-Encoding: binary

POST /v1/projects/your-firebase-project-id/messages:send HTTP/1.1
Content-Type: application/json

{
  "message": {
    "token": "user_1_fcm_token",
    "notification": {
      "title": "Hey User 1!",
      "body": "Your personalized update: Your order #123 is on its way!"
    },
    "data": {
      "order_id": "123",
      "tracking_url": "app://track/123"
    }
  }
}

--unique_batch_boundary_123
Content-Type: application/http
Content-Transfer-Encoding: binary

POST /v1/projects/your-firebase-project-id/messages:send HTTP/1.1
Content-Type: application/json

{
  "message": {
    "token": "user_2_fcm_token",
    "notification": {
      "title": "Hi User 2!",
      "body": "Your personalized update: New discount just for you!"
    },
    "data": {
      "discount_code": "USER2_SAVE20"
    }
  }
}

--unique_batch_boundary_123--

2. Handle Batch Limits & Splitting #

• FCM allows a maximum of 500 sub-requests per batch. For 1000 users, split your target list into two separate batch requests each minute.
• Generate a fresh boundary string for each batch to avoid parsing errors.

3. Key Optimizations & Notes #

• Dynamic User Filtering: Each minute, run your logic to fetch the 1000 target users, then generate a personalized message payload for each user’s token.
• Access Token Management: Use a service account to generate OAuth2 access tokens (valid for 1 hour) — cache these tokens to avoid re-generating them for every batch request.
• Error Handling: Batch responses include results for each sub-request. Parse the response to catch failures (e.g., invalid tokens, quota issues) and implement retry logic for transient errors.
• Quota Compliance: Batch requests don’t bypass FCM’s sending quotas, but they reduce HTTP overhead significantly, making your workflow more efficient.

Why Other Options Don’t Fit #

• Topics/Groups: These require static user groups, which doesn’t work for your dynamic, minute-by-minute user selection.
• Single API Calls: 1000 calls per minute is inefficient, increases latency, and raises the risk of hitting rate limits.

内容的提问来源于stack exchange，提问作者Dhruv Pahuja