SpeechBubble

Track clients and their goals, log therapy sessions with trial-level data, and generate the paperwork that comes after: SOAP notes, printable practice materials, evaluation reports. The part I care most about: client names never reach the AI. A tested de-identification layer swaps them for placeholders on the way out and restores them on the way back.

Architecture diagram:

Text alternative

The therapist logs sessions and tallies in the app, which stores everything in its own database. When AI paperwork is requested, a de-identification boundary swaps client names for placeholders before anything reaches the model, and restores the real names on the response. Out the other side come SOAP notes, printable practice materials, and evaluation reports.