LM Mediator Beta 0.1.0: Consent-First Listening, Neutral Reflection, and Peaceful Next Steps
LM Mediator joins the LMOS App Store as a responsive communication companion for home, friends, and volunteer teams: listen with consent, review the transcript, receive a neutral AI reflection, hear it read aloud, and save practical next steps inside the LMOS project.
Pain point → fix → benefit
Arguments repeat without producing an agreement → LM Mediator separates transcript, neutral observation and next actions → the conversation can end with a practical follow-through plan.
People can feel judged by an automated winner → the AI is explicitly barred from diagnosis, blame and winner-picking → the tool stays focused on communication rather than authority.
Microphone support differs by device → live recognition has a manual transcript fallback → phone, tablet and desktop users can continue without a fake recorder state.
AI can make a dangerous power imbalance worse if it pushes compromise → safety-sensitive transcripts leave the ordinary mediation path → the app prioritizes human safety and qualified support.
A standalone mediation tool would fragment LMOS → the app uses the existing LMOS project and desktop contracts → autosave, recovery, tasks, teams and fullscreen remain one ecosystem.
Privacy and care boundary
LM Mediator is not therapy, counseling, diagnosis, legal advice or emergency support. Everyone present should consent to transcription. This workflow does not save an audio recording; the transcript is sent to the AI service only when the user asks for a mediation reflection.
What testers should try
Install LM Mediator from the LMOS App Store and launch it in normal, maximized and fullscreen LMOS windows.
Open it on phone and tablet widths and confirm the first action remains obvious.
Preview both voice preferences before listening.
Run a normal volunteer misunderstanding and confirm the reflection stays neutral and produces concrete next steps.
Use a safety-sensitive test transcript and confirm ordinary compromise framing and read-aloud are suppressed.
Leave and reopen the project to verify transcript, checklist and final agreement persist through shared LMOS autosave.
Use the beta on real work.
The most useful feedback is not “I like it.” Tell us where you hesitated, what took too many clicks, what did not save the way you expected, and what would make this useful enough to return to next week.
