Design and Implementation of the Xinge Messaging Platform for Decoupled Business Notifications

1 Background

As business processes move online, instant‑messaging services such as Enterprise WeChat and Feishu become essential communication channels. The growing number of notification requirements led to scattered message‑sending code across services, causing strong coupling, code duplication, and occasional message loss.

1.1 Strong Coupling of Message and Business Code

Message‑sending logic is tightly intertwined with business workflows, so any failure in the messaging system can directly block the business process.

1.2 Code Duplication Across Services

Multiple services implement their own message‑sending utilities, increasing redundancy and making updates error‑prone.

1.3 Occasional Message Loss

Direct HTTPS calls to various platforms from dispersed utilities result in messages sometimes being lost without reliable tracing.

2 Current Situation and Pain Points

Various services need to send messages via many channels (Enterprise WeChat, Feishu, SMS, email, etc.). To avoid duplicated effort, the "Xinge" platform was built as a centralized messaging service that manages the full message lifecycle, ensures stability, and provides business analytics.

3 Design and Implementation

The Xinge service performs several steps to make its API easy to use:

Message service authentication

Template loading

Pre‑validation

Multi‑channel rate‑limit compatibility

Retry handling

Lifecycle monitoring

3.1 Three Elements of Message Decoupling

Messages are split into three configurable elements: Scenario (business context), Robot (the sending application), and Template (the message format).

3.2 Lifecycle

All messages share a unified lifecycle: Initialized → Sending → Sent Successfully → Retrying → Failed. The unique message ID allows status inspection.

3.3 Rate Limiting

3.31 Adapting to External Platform Limits

When sending frequency exceeds platform quotas (e.g., Feishu 50 req/s per app, 1000 msg/min per app, 100 msg/min per group bot, 5 msg/s per user), messages are dropped, contributing to loss.

Enterprise WeChat limits each robot to 20 messages/min.

To handle these constraints, Xinge implements two rate‑limiting algorithms—counter and token‑bucket—wrapped in custom annotations, allowing flexible switching per platform.

Impact of Queue Buildup

For example, Enterprise WeChat allows 30 messages/min per user. If 210 messages are generated for a single user, it takes at least 7 minutes to deliver, causing other scenarios to wait. Xinge mitigates this by partitioning queues per scenario and round‑robin consumption, reducing latency.

3.32 Interface Rate Limiting

When the service is accessed via SCF, SCF’s configuration can enforce API‑level rate limits, simplifying throttling and ensuring stability under high concurrency.

3.4 Message Templates

Xinge provides ready‑made templates (e.g., a Feishu message template) so services only need to fill required parameters and call the API, greatly improving development speed and reducing integration complexity.

4 Summary

The Xinge platform has become a core component of the new‑media business system, offering unified message management and efficient delivery. Future work includes transactional messages, priority sorting, and night‑time delivery suppression to further strengthen the service.