How are you running AnythingLLM?

Docker (remote machine)

What happened?

Summary

The voice-to-text feature is brilliant 👏, but currently, if there's any existing text in the prompt field (e.g. from using @agent or writing part of a query), the voice input completely overwrites it. Ideally, it should insert the transcribed text at the current cursor position, or append to the end, instead of replacing the whole field.

Steps to Reproduce

1. Type a partial prompt in the input box (e.g. @agent Summarise this:)
2. Click the mic to use voice input
3. Dictate a sentence
4. Observe: The entire box is replaced with only the dictated text

Expected Behavior

Voice input should respect the current prompt content and either:

• Insert the text at the current cursor location
  or
• Append it to the end of the existing input

Why it matters

Overwriting breaks workflows, especially when using structured or agent-based prompts.

Environment

• AnythingLLM Version: 1.8.4
• Browser: Chrome
• Platform: Docker

P.S. Love your work legends👍

Are there known steps to reproduce?

No response