Introduction

In usage scenarios of DeepSeek Harness, developers often need to understand the current load without interfering with the request flow. dsh-concurrency-meter is designed to address this pain point. It is a read-only concurrency monitoring tool for model requests.

Plugin Overview

This plugin is maintained by Wanbinyu and is licensed under the MIT License. It plugs into the official DSH llm/stream call chain to display current active requests, peak concurrency, execution results, and per-provider statistics. Unlike middleware that intercepts or modifies requests, this plugin only observes the call chain and does not intercept, queue, or change model request results.

Core Features

The plugin provides the following core capabilities:

  • Real-time Monitoring: Displays current active count, peak concurrency, started requests, and failure count.
  • Provider Statistics: Counts successes, failures, aborts, and streams not fully consumed by provider.
  • Request Details: Shows provider, model, use case, and elapsed running time for active requests.
  • Threshold Alerts: Shows an alert by default when concurrency reaches 3. The threshold can be changed in the configuration.
  • Performance Optimization: Refreshes every second while the settings page is open, and performs no browser polling after it is closed.
  • Memory Management: Supports resetting in-memory statistics. Resetting does not cancel running requests.
  • Privacy Protection: Does not read messages, prompts, model outputs, credentials, or session IDs.
  • Non-intrusive Design: Only observes the call chain without rate limiting or queuing requests.

Installation and Enablement

Before installing, ensure your environment meets the requirements: Node.js version >= 22.19 and DeepSeek Harness version >= 0.1.0-rc.6.

  1. Install the plugin:
    Use the following command to add the plugin to the Web configuration:
    dsh plugin --profile web add https://github.com/Wanbinyu/dsh-concurrency-meter/releases/download/v0.1.6/dsh-concurrency-meter-0.1.6.tgz
  1. Start the service:
    After installation or update, restart the Harness Web service:
    dsh web
  1. Access the monitor:
    Once started, open “Settings -> Concurrency Monitor” to view the statistics.

Configuration

The plugin supports adjusting the threshold and the number of details displayed through the configuration file.

Configuration Example:

- insert:
    - id: concurrency-meter
      name: dsh-concurrency-meter
      config:
        warningThreshold: 3
        maxActiveDetails: 20
Option Default Description
warningThreshold 3 Concurrency alert threshold, range 1-100. An alert is displayed when this value is reached, without limiting requests.
maxActiveDetails 20 Maximum number of active requests expanded for details, range 1-100. The total count is still fully tracked.

Notes

  • Unofficial Plugin: This is a community-maintained unofficial plugin and is not affiliated with or endorsed by DeepSeek.
  • Statistics Scope: All statistics are kept only within the current DSH Host process and are automatically cleared after restart. The plugin does not persist statistics, nor does it provide historical trends, rate limiting, or request queuing.
  • Lazy Streams: Lazy streams that are not actually consumed are not counted as “started requests.”