[ENHANCEMENT] Audio Feedback for Task Completion and Replies

[Shaerif]

Feature Description

Add or fix audio feedback for task completion and replies. The system already supports sound alerts for completion, waiting for action, and abnormal stop. Each has its own sound file and play option, with volume at maximum. Extend this behavior to include a sound for successful task completion and for replies. Use concise, distinct tones so users can tell them apart without looking at the screen.

Current Behavior

The audio feedback feature does not play sounds after a task completes or when a reply arrives. Users do not hear any audio alert even when notification settings are enabled, and the volume is set to maximum. This causes missed updates when a task needs attention or has finished running.

Steps to Reproduce

Enable sound alerts in notification settings
Run a task and wait for completion
Wait for a reply or system response
Observe that no sound plays

Expected Result

A short, distinct sound should play when a task completes or a reply arrives.

Actual Result

No audio plays. The user must check the interface manually to see task status or replies.

Use Case

When a task completes, a short success sound plays. When a reply is ready, a different sound plays. The user hears the alert instantly, even when the system window is in the background. This helps identify updates quickly and reduce missed notifications.

What "Replies" Means

"Replies" refers to any response from the LLM (Large Language Model) or the extension that requires user attention. This includes:

• LLM Responses: When the AI model generates a response after processing a user request
• Extension Messages: Notifications or status updates from the Roo Code extension itself
• User Interaction Requests: When the system needs user input, approval, or clarification to continue (e.g., auto-approve prompts, clarification questions, or permission requests)

Each of these scenarios should have its own distinct audio cue so users can immediately recognize what type of attention is needed without switching to the VS Code window.

Impact

Users miss task completion or reply notifications, which delays response time. This affects work that depends on fast task monitoring and automation tracking.

Additional Context

Audio feedback makes it easier for users to stay attentive while multitasking. It supports accessibility for users who depend on audio signals or work in environments where visual alerts are missed. The current sound notification system already handles several states, so adding this extension will require minor configuration changes. This feature improves reaction time and helps users keep track of system output efficiently.

Suggested Implementation

Please check the audio trigger event for task completion and reply states. Confirm the correct sound files are linked and that playback starts when the status updates.

Custom Sound Selection

Allow users to choose custom sounds for each notification type. This would include:

• Task Completion Sound: User-selectable audio file for when a task finishes successfully
• LLM Reply Sound: Custom sound for when the AI generates a response
• User Interaction Request Sound: Distinct audio for when user input is required
• Volume Control Per Sound: Individual volume sliders for each notification type
• Sound Preview: Play button to preview each sound before saving settings

Users should be able to select from a set of built-in sounds or upload their own audio files (e.g., MP3, WAV). This personalization ensures users can choose sounds that are most noticeable and pleasant for their specific work environment.

[roomote-v0]

I've analyzed the existing audio feedback system in the codebase and have some findings and questions.

Current state of audio in the codebase:

The system already has 3 sound files (celebration.wav, notification.wav, progress_loop.wav) and the following triggers in ChatView.tsx:

Sound	Trigger	Description
celebration	completion_result ask	Plays when a task completes (only if not partial and message queue is empty)
progress_loop	api_req_failed / mistake_limit_reached asks	Plays on API failures or mistake limit
notification	interactionRequired message	Plays when user interaction is needed

So task completion already has a sound (celebration). The soundEnabled setting and volume controls are also fully implemented in Settings > Notifications.

What's missing (likely the issue):

The followup ask type (when the AI replies with a question or response that needs user attention) does not trigger any sound. This is likely the "reply" scenario described in the issue -- the user doesn't hear anything when the AI responds with a followup.

Clarifying questions:

1. "Reply" definition: When you say "a reply arrives," do you mean the AI's followup question/response (the followup ask type)? Or do you mean something else, like any say message from the AI?

2. Task completion already works: The celebration sound already plays on task completion. If you're not hearing it, could this be a browser/webview audio playback issue rather than a missing trigger? Or are you referring to a specific completion scenario where it doesn't fire?

3. Scope: Should we add a notification sound to ALL tool approval requests (tool ask type), or just to followup asks? Currently, tool approvals (file edits, commands, etc.) are silent unless auto-approved, which could also be considered a "the task needs your attention" moment.

Proposed plan (pending answers above):

1. Add playSound("notification") to the followup ask handler in ChatView.tsx (line ~298) so users hear a sound when the AI replies with a question
2. Optionally add sound to other ask types that require user attention (e.g., tool, command, use_mcp_server) -- pending clarification on scope
3. Add/update tests in ChatView.notification-sound.spec.tsx

Please mention @roomote in a comment saying that you approve this plan (or with answers to the questions above), and I'll proceed with the implementation.

[Shaerif]

Hi @roomote-v0,

Thanks for the detailed analysis! Let me clarify a few points:

1. Definition of "Reply"

By "reply," I specifically mean the AI's follow-up questions or responses (the followup ask type)
essentially anything that requires user attention and breaks the flow of autonomous execution. This includes:

• When the AI needs clarification on a task
• When the AI asks a follow-up question to proceed
• Any situation where the system stops and waits for user input

These should trigger a distinct notification sound that's different from the task completion sound, so users can immediately tell the difference between "task finished" and "AI needs your attention."

2. Task Completion Sound

You're right that the celebration sound already plays on task completion — I can confirm it does work. However, the main issue is that there's currently no way for users to customize which sound plays for each event type. Users are stuck with the default sounds and cannot:

• Choose different sounds for different notification types
• Upload their own custom audio files
• Adjust volume per notification type
• Preview sounds before selecting them

3. Suggestions Based on Your Analysis

Your proposed plan looks great! To summarize what would address this feature request:

• ✅ Add playSound("notification") to the followup ask handler (as you suggested at line ~298)

• ✅ Consider adding sounds to other ask types that require user attention (e.g., tool, command, use_mcp_server) — this aligns with the "User Interaction Requests" concept in the issue

• ✅ Implement the Custom Sound Selection feature allowing users to:

  • Select custom audio files per notification type
  • Upload their own MP3/WAV files
  • Have individual volume controls per sound
  • Preview sounds in settings

Would love to see this implemented! Happy to test or provide more feedback if needed.

[roomote-v0]

I've seen the comment and will explore a potential fix. The followup ask type currently only gets a notification sound via the delayed interactionRequired handler (2-second timeout), but not immediately when the follow-up question appears. I'll add immediate notification sound feedback for followup asks so users get audio cues right when the AI needs their attention. I'll open a PR to share progress.

[roomote-v0]

Opened a draft PR with my attempt at addressing this: #12048

The PR adds an immediate notification sound when the AI asks a follow-up question (the followup ask type), so users get audio feedback right when the AI needs their attention instead of waiting for the 2-second interactionRequired timeout. Includes tests for both the non-partial and partial (streaming) cases. Feedback welcome!

[Shaerif]

You still cannot use a custom sound for each action that requires user interaction.