Introduction

The DeepSeek Harness (DSH) plugin ecosystem emphasizes modularity. dsh-period-report focuses on solving the archiving and review of session data, especially for scenarios that require periodic work reporting.

Plugin Positioning

This is a DeepSeek Harness plugin used to generate AI-written periodic work reports and schedule system reminders. It is maintained by user zhengjy01 under the MIT license. It allows extracting data from session conversations, generating complete statistical reports with natural language summaries (e.g., “Today at a glance”), and supports triggering automated reminders via system notifications.

Core Capabilities

  • Web UI (v0.2.0): Provides a session report settings page for configuring report generation and reminder features. A reminder prompt is shown in the bottom-right corner of the interface, and clicking it allows viewing the full report.
  • Arbitrary time range: Supports custom startDate and endDate, covering single-day, multi-day, or full-month ranges.
  • AI narrative summary: Uses the deployed default model to generate natural language summaries (e.g., overview, highlights, review, suggestions). If AI is unavailable or disabled, it falls back to a template-based highlighted list.
  • Complete statistics: Includes detailed data such as active session count, new sessions, user/assistant message counts, tool call counts, and token usage.
  • Single-session details: Displays title, ID, working directory, active window, message/tool/turn counts, etc.
  • Periodic reminders: Supports custom interval (intervalDays), anchor date (anchorDate), trigger time (hour, minute), and report range (previous or current).
  • System notifications: On macOS, triggers Notification Center banners via osascript; on Linux, triggers desktop notifications via notify-send.
  • Bilingual support: Reports support Chinese (zh) or English (en).
  • Persistent configuration: The configuration file is stored at ~/.dsh/dsh-period-report, with permissions 0600. Full reports are saved at ~/.dsh/dsh-period-report/reports/.

Installation

Use the official installation command to add the plugin.

dsh plugin --profile web add github:zhengjy01/dsh-period-report

Usage

The plugin provides two main tool commands.

report_generate

Generate a report for an arbitrary date range.

Parameter Type Description
startDate string (required) Start date, YYYY-MM-DD (inclusive)
endDate string (required) End date, YYYY-MM-DD (inclusive)
includeSubagents boolean Whether to include subagent sessions (default false)
withUsage boolean Whether to include token usage statistics (default true; disabling can improve speed)
language string Language zh or en (default zh)
narrate boolean Whether to enable AI narrative summary (default true; disabling uses a quick template summary)

report_config

View or update reminder configuration (persisted at ~/.dsh/dsh-period-report.json).

Parameter Type Description
enabled boolean Enable/disable scheduled reminders
anchorDate string Anchor date YYYY-MM-DD; the period starts from this day
intervalDays number Remind every N days (1-90; 7 is weekly)
hour / minute number Reminder time on that day
range string previous (report the previous period, default) or current
language string Report language zh / en
narrate boolean AI narrative summary toggle

Example: Set a reminder to generate Chinese reports every 2 days at 09:00, starting from 2026-06-01.

dsh report_config \
  enabled=true \
  anchorDate="2026-06-01" \
  intervalDays=2 \
  hour=9 \
  minute=0 \
  range="previous" \
  language="zh" \
  narrate=true

Scenarios and Notes

  • Suitable scenarios: Suitable for developers who need to periodically summarize AI interactions, track token consumption, review tool calls, or need automated workflow reminders.
  • Permission note: The plugin runs with the current DSH process permissions; ensure write access to the configuration directory ~/.dsh/dsh-period-report/.
  • Source code review: Before installation, review the GitHub repository source code and license (MIT).

Summary

dsh-period-report provides a complete workflow closed loop from data extraction and AI summary generation to system notification triggering. By configuring report_generate and report_config, you can quickly establish a reporting mechanism that fits personal or team habits.