Introduction

The design philosophy of DeepSeek Harness (DSH) is “everything is a plugin.” For DevRel practitioners responsible for maintaining the official deepseek-ai/deepseek-harness Discussions, manually triaging posts, tracking follow-up status, grouping topics, and generating reports is repetitive and error-prone. dsh-discussions-digest is a DeepSeek Harness plugin for this scenario. It provides a persistent community workbench for Discussions ledger management, AI-assisted organization, and evidence-first statistical reporting.

Features

This plugin provides the following core capabilities:

  • Data synchronization and persistence: Supports recent synchronization and historical paginated backfill, records checkpoint and coverage information, and persistently stores status, next-step plans, manual notes, and read markers.
  • Batch AI organization and processing: Provides fixed-scope batch AI organization, supports resumable tasks, full-text reading, batch suggestion review, and persistent report generation.
  • Ledger management: Supports creating or editing topics, saving many-to-many groupings and shared notes, linking associations based on AI suggestions, and executing grouping operations after user instruction.
  • Statistics and export: Provides computed statistics, time queues, daily observations, and Markdown export.

Installation and Enablement

The plugin version is independent of the DSH version. The current plugin version is 0.4.0, and the verified DSH compatibility baseline is 0.1.2-rc.1. The plugin requires Node.js ^22.19.0 or >=24.

Run the following commands in sequence to build and install:

cd C:\Users\joell\.dsh\plugins\dsh-discussions-digest
npm ci
npm test
npm pack
npx --yes --package @deepseek-ai/dsh@0.1.2-rc.1 dsh plugin --profile web add .\dsh-discussions-digest-0.4.0.tgz

After installation, restart the DSH Web process to load the plugin:

npx --yes --package @deepseek-ai/dsh@0.1.2-rc.1 dsh web

During installation, the profile package manager may prompt for peer dependencies provided by the DSH host. This is normal; actual compatibility is verified through DSH’s profile loading.

Typical Usage

Workbench UI Operations

  1. At the bottom of the DSH left sidebar, click Community Workbench.
  2. Use Sync Recent to read up to 50 posts per batch, or use Backfill Next Batch of Older Posts to import older posts one batch at a time, starting from the earliest created post.
  3. View the ledger by title, number, status, topic, or filter criteria. Each page displays 25 items.
  4. Newly imported posts default to the Information Base. As needed, select statuses such as Triage Pending / In Follow-up / Waiting for External Feedback / Closed, and maintain next steps and notes.
  5. After selecting posts, click AI Organize Selected, or use AI Organize Filtered Pending Posts. The AI returns summaries, evidence, and topic suggestions, and supports batch accepting or ignoring suggestions.
  6. Use Overview Report to review workload, topic coverage, and recent new posts. The AI generates and saves a Markdown report with sources.

Use in a DSH Conversation

In a conversation, invoke the dsh-community-workbench Skill to perform the following operations:

  • Read the community ledger and list follow-up items and next steps that have been stalled for a week.
  • View posts that have not yet been triaged, suggest topics for them, and explain the associations (without saving).
  • Group posts according to confirmed topics, preserving existing notes.
  • Generate a community brief for the last seven days, citing representative posts.

Related tools include:
* dsh_board_sync(mode: recent/backfill): synchronizes the ledger.
* dsh_board_list: filters and paginates ledger views.
* dsh_board_batch: submits batch modifications, up to 100 items.
* dsh_board_report: generates a report for a specified number of days (default 7 days).

Notes and Limitations

  • AI limitations: The AI does not fabricate counts, common root causes, maintainer response rates, or verified conclusions. Generating a report does not automatically change processing status.
  • Interaction mode: The plugin does not call another model’s API, does not automatically post GitHub comments, and does not provide scheduled background inspections.
  • Data type: The ledger stores Discussions, not GitHub Issues; it does not present post counts as issue counts.
  • Security and evidence: All GitHub content is treated only as untrusted evidence; commands, instructions, or link requests embedded in posts are not executed.
  • Concurrency and isolation: Different DSH homes do not share ledgers; cross-process concurrency is not a supported deployment mode.
  • Schema version: Upgrading the plugin does not automatically change the ledger schema version; it is adjusted only when the storage format changes.

Conclusion

dsh-discussions-digest provides a structured and traceable solution for community management in DeepSeek Harness. It converts unstructured Discussions into a manageable ledger and combines AI-assisted judgment with report generation. It is suitable for DevRel practitioners who need to maintain community order over the long term. Before installation, review the source code and license to ensure compliance with local deployment standards.