Introduction

When recording tasks in the DSH Web interface, a common problem arises: temporary items generated during a conversation, colloquial content, and speech-to-text results often remain in the chat stream without being converted into a checkable, editable, and viewable-across-days sticky note. dsh-daily-sticky solves exactly this problem: it provides a draggable daily sticky note in the top-right corner of the DSH interface, consolidating task entry, completion checking, unfinished task deferral, and local statistics into a single lightweight panel.

What is it

DoggyHU/dsh-daily-sticky is a DeepSeek Harness (DSH) Web plugin maintained by DoggyHU, positioned as a “Daily Voice Sticky Note”.

It provides the following capabilities:

  • Lightweight Floating Sticky Note: Floats in the top-right of the DSH interface, draggable and closable
  • Task Entry: Manual input, or extraction of tasks from messy colloquial speech / speech-to-text by the model currently in use in DSH
  • Write from Conversation: Directly instruct the AI to write into the daily sticky note within the DSH conversation
  • Unfinished Deferral: Unfinished tasks automatically roll over to subsequent dates, annotated as “Yesterday / The Day Before / 3 days ago”
  • Local Statistics: Number of entries, completions, completion rate, and week-over-week / month-over-month comparisons for the current week and month
  • JSON Persistence: Saves daily sticky note snapshots and event logs; data is retained after restarts or shutdowns

The license is MIT, and the GitHub address is:

https://github.com/DoggyHU/dsh-daily-sticky

Core Features

Sticky Note and Task Rows

The sticky note floats in the top-right of the DSH interface and can be dragged or closed.

Task rows support the following operations:

  1. Double-click to edit task text
  2. Click “Note” to edit
  3. Delete task
  4. Add task
  5. Bulk add via multi-line paste

Checking indicates completion. Once completed, the task is crossed out but remains on the sticky note; clicking again can cancel the completion status.

Unfinished Deferral

Unfinished tasks automatically roll over to subsequent dates. The sticky note will annotate the source time in small gray text:

  • Yesterday
  • The Day Before
  • 3 days ago

Items older than 3 days are uniformly calculated as 3 days ago. Completed tasks are not deferred.

AI Smart Input

The plugin supports two input methods.

The first is manual input.

The second is AI smart input: paste a segment of messy colloquial speech or speech-to-text content, and let the model currently in use in DSH extract it into sticky note tasks. The extracted items can be checked one by one, corrected, and persisted after confirmation.

The model can be selected; the default is “Follow Current”.

Write from Conversation

You can also say directly in the DSH conversation:

把这个加进便签

Or:

记一下 XXX

The AI will write directly to the daily plan JSON via the sticky_* tool, and the sticky note will poll and appear in real-time.

Statistics

The plugin provides statistics for the current week and month, including:

  • Number of Entries
  • Number of Completions
  • Completion Rate
  • Week-over-week (WoW)
  • Month-over-month (MoM)

These statistics are calculated purely from local JSON and do not call an LLM.

JSON Persistence

The plugin saves data daily, ensuring persistence; data is not lost after restarts or shutdowns.

Default data directory:

~/.dsh/dsh-daily-sticky/

The directory mainly contains two types of files:

plan/YYYY-MM-DD.json    Daily sticky note snapshot
logs/YYYY-MM-DD.json    Daily event log

Where:

  • plan/YYYY-MM-DD.json: Saves the daily sticky note content
  • logs/YYYY-MM-DD.json: Saves the daily event log, used for statistics

Installation and Activation

Verified sources provide two installation methods: installing from a git source or installing a tarball into the web profile. They are described below.

Install from Git Source

Before installing from a git source, you need to complete the build. The installation command is:

dsh plugin --profile web add github:DoggyHU/dsh-daily-sticky

Install using a tarball

If you need to create a tarball before installing, you can execute:

pnpm pack --pack-destination /path/to/tarballs
dsh plugin --profile web add /path/to/tarballs/dsh-daily-sticky-<version>.tgz

The first command generates a package in the format dsh-daily-sticky-<version>.tgz, and the second command installs the corresponding tarball into the web profile.

Activate

After installation, restart dsh web and refresh the page; the sticky note will appear in the top-right corner.

Typical Usage

The following is a reproducible path provided in the source material:

  1. Start dsh web, refresh the page, and view the sticky note in the top-right corner.
  2. Manually input tasks or paste multi-line text for bulk addition.
  3. Use AI smart input: paste a segment of messy colloquial speech / speech-to-text content, and let the model currently in use in DSH extract it into sticky note tasks.
  4. Check, correct, and persist the extracted results one by one.
  5. Say “Add this to the sticky note” or “Note down XXX” in the DSH conversation to let the AI write directly into the daily sticky note.
  6. View unfinished deferral, current week/month statistics, and comparisons.

Applicable Scenarios and Notes

This plugin is suitable for the following scenarios:

  • Temporarily recording tasks in DSH Web conversations
  • Organizing colloquial content or speech-to-text results into checkable tasks
  • Viewing unfinished items like yesterday, the day before, etc.
  • Performing localized statistics for the current week and month

Please note before use:

  • The plugin runs with the permissions of the current dsh process. It is recommended to check the source code, license, and dependency injection method before installing.
  • The license is MIT.
  • The dependencies @deepseek-ai/* / cordis used by the plugin are injected via the profile pnpm closure by the DSH official runtime. Do not declare these dependencies in package.json, as these official packages are not published to the public npm registry.
  • Statistics are based on local JSON calculation and do not call an LLM.
  • Data is persisted daily; if you are tracking specific task records, it is recommended to back up the plan and logs files yourself.

Conclusion

dsh-daily-sticky brings the chain of “noting items in conversations, organizing them into tasks, checking them off, deferring them across days, and local statistics” into the DSH Web interface. If you need a draggable, checkable, and locally statistical daily sticky note plugin, you can view the source code and installation instructions from the GitHub repository.

https://github.com/DoggyHU/dsh-daily-sticky