Socrates-Voice

Socrates Voice is a voice-first Cognitive Behavioral Therapy (CBT) companion that makes evidence-based mental health support accessible through natural, real-time conversation. ## The Problem Traditional CBT apps are text-heavy and break the therapeutic flow. People in emotional distress often struggle to articulate their thoughts by typing, and access to therapists is limited by cost, stigma, and availability. ## The Solution Socrates Voice guides users through the 3-step CBT model — Catch, Challenge, Replace — entirely by voice. The user simply speaks; the system listens, understands, and responds with empathy. ## How It Works 1. Catch — AssemblyAI's real-time voice agent transcribes speech with low latency and active turn detection. Gemini identifies Automatic Negative Thoughts (ANTs) from the transcript. 2. Challenge — The agent poses analytical Socratic questions, guiding the user to evaluate evidence and recognize cognitive distortions while preserving multi-turn session history for continuity. 3. Replace — The user is guided to formulate a balanced, constructive, and objective alternative perspective. ## Safety First — MIND-SAFE Protocol Every transcript is monitored in real time. Crisis keywords trigger an immediate session pause and present international helpline information (988 Lifeline, 911/112). A user in crisis is never left without guidance. ## Architecture Microservice-based design: Vite + TypeScript frontend, Fastify + Drizzle ORM gateway backed by PostgreSQL, and a FastAPI Python service handling CBT logic and Google Gemini integration. Multi-container orchestration via Docker Compose. ## Why AssemblyAI The voice agent provides the real-time turn-taking backbone that makes conversational therapy feel natural rather than scripted — essential for building therapeutic rapport.