117BS/dsh-perlica-ding

"We are not the ones who state ideals — we are the ones who carry them out." —— Perlica, Supervisor of Endfield Industries

简体中文 | English

🔔 A Perlica-themed (Arknights: Endfield) tiered sound notification plugin for DeepSeek Harness.

Perlica reports your agent's status in her calm, concise, terminal-announcement tone: plan ready, task done, needs your input, error — each with its own sound. Plain conversation stays silent. Custom sounds (TTS voice supported) and audible even when the window is in the background.

📖 About Perlica

Perlica is the Supervisor and official spokesperson of Endfield Industries, an outstanding Protocol Technology specialist who manages the development and application of Protocol Originium technology and the day-to-day operations of the Dreadnought. She handles crises with decisive calm, always moving between bases and strongholds for the stable development of Endfield.

This plugin's sound style is inspired by her — clean, concise, with a touch of technical bureaucracy, addressing you as 「Administrator」.

✨ Features

Scenario	Trigger	Sound file
🗂️ Plan ready	Turn closes while plan mode is active	plan.wav
✅ Task done	Turn closes and execution-class tools were used (file writes / commands / subagents / workflows …)	done.wav
💬 Plain chat	No tools used, or only lookup tools (read / web_search …)	Silent
🙋 Needs your input	ask_user_question tool / approval request	ask.wav
⚠️ Error	Agent turn errored	fail.wav

Also:

• Root conversation only: subagents do not beep individually
• Debounce: 2.5s per kind, no sound storms
• Custom sounds: drop wav files into the sound dir, no code changes
• Cross-platform: Windows (SoundPlayer) / macOS (afplay) / Linux (paplay/aplay)
• Configurable: master switch, debounce, sound directory

📦 Install

From GitHub (recommended, no npm account needed)

dsh plugin --profile web add https://github.com/117BS/dsh-perlica-ding

Restart the web profile to activate:

dsh web restart

From npm (once published)

dsh plugin --profile web add dsh-perlica-ding

Local development

dsh plugin --profile web add /path/to/dsh-perlica-ding

🎵 Sound files

Works out of the box: the Perlica voice lines ship inside the plugin package (sounds/ is bundled) — no configuration needed, each scenario has its own sound:

plan.wav   — plan ready
done.wav   — task done
ask.wav    — needs your input
fail.wav   — error (optional)

To use your own sounds: generate wav files with any TTS tool and drop them into your workspace root (or the configured soundDir); same-named files override the bundled ones, effective immediately without restart. Lookup order: configured dir → workspace → bundled → OS fallback.

Requirement: must be real WAV (PCM). If your TTS tool exports MP3, convert first: ffmpeg -i input.mp3 -acodec pcm_s16le plan.wav.

🔊 Volume control (with one-click preview)

Just open Settings: go to DSH Settings → 佩丽卡提示音 (Perlica Terminal) and you get a panel:

Volume
[==========|=========]  50%
[Original 100%] [Medium 60%] [Soft 30%] [Mute 0%]

Preview
[Plan ready] [Task done] [Needs input] [Error]

• Drag the slider or tap a preset (0 = silent, 100 = original) — saved automatically, effective immediately
• Tap a preview button to play that sound right away, so you can find the right level by ear instead of guessing
• Previews use the exact same playback path as real notifications — what you hear is what you get

Suggested levels

volume	Effect	When
100	Original level	Noisy environment
60	Slightly quieter	Headphones, quiet room
30	Soft cue	Late night, library
0	Fully muted	Temporarily off

💡 How it works: the plugin rescales the WAV PCM samples directly (zero dependencies, cross-platform) and never touches your system volume. Scaled copies are cached in the system temp directory, computed once per volume.

⚠️ Note: non-PCM audio sources (rare) fall back to the original level.

⚙️ Advanced configuration

The volume can also be set in config (as the starting value; anything chosen in the settings page wins). In the profile's cordis.yml or the user patch layer:

plugins:
  dsh-perlica-ding:
    enabled: true        # master switch
    volume: 100          # initial volume 0-100 (settings page overrides)
    debounceMs: 2500     # min gap between same-kind sounds (ms)
    soundDir: ""         # custom sound dir; empty = workspace root
    execTools: []        # tools that count as "executing a task"; empty = every tool counts (legacy)

🧪 Verify

Ask the agent to run any task (e.g. call a tool); you should hear a sound when it finishes. Or test the audio chain manually:

$p = New-Object Media.SoundPlayer 'D:\deepseek\done.wav'; $p.PlaySync()

📄 License

MIT