🎭 PersonaReflect

Multi-Persona Self-Reflection Coach

🌟 Inspiration

People often tackle tough decisions from one perspective or lean on a single source of advice. But real progress needs multiple lenses— from compassion to logic to mindful presence.

Inspired by multi-perspective cognition, PersonaReflect gives users an internal board of advisors—four specialized AI personas—so reflection is deeper, more emotionally grounded, and action-oriented.

💡 What It Does

Users journal a dilemma (e.g., “I procrastinate on big projects”).
PersonaReflect spawns four complementary personas in parallel, then synthesizes their insights into a unified reflection and a personalized Action Plan.

Persona	Role	Focus
🧑‍🏫 Dr. Chen — CBT Agent	Cognitive-Behavioral Coach	Uncovers thinking patterns and core beliefs
💬 Maya — Empathetic Agent	Empathetic Friend	Emotional validation and supportive reframing
🧮 Alex — Rational Agent	Rational Analyst	Calm, structured, data-driven guidance
🧘 Sage — Mindfulness Agent	Mindfulness Mentor	Present-moment focus and non-judgmental awareness

Output: one concise response per persona → Unified Reflection → Action Plan with concrete next steps.

🏗️ How We Built It

Frontend → FastAPI Backend: Simple UI sends the user’s journal to the backend.
ADK Orchestrator: The “brain” that dispatches calls to the four agents in parallel.
Persona Agents: Each uses a tailored prompt grounded in its psychological framework (CBT, empathy, rational analysis, mindfulness).
Unified Response Module: Merges perspectives into a coherent summary.
Action Plan Generator: Turns insights into clear, time-aware steps and returns them to the UI.
Local Journal Store: Persists entries to enable future personalization.

⚙️ Challenges

Designing prompts so every persona stays distinct yet relevant.
Handling asynchronous LLM calls (4x) and two speech models within hackathon constraints.
Converting free-form reflections and mixed signals into consistent, actionable plans.

🏆 Accomplishments

Multimodal reflection: supports text and voice.
Speech fusion: Whisper ASR + SpeechBrain SER for emotion-aware guidance.
Emotion- & time-aware output: blends empathy with structure and scheduling.
Clean, extensible architecture: FastAPI + Google ADK + calendar integration.

🧠 What We Learned

The power of agent-based AI is role specialization. Instead of one “super AI,” a team of small, focused agents yields richer, more human-like guidance—especially with parallel orchestration for speed and depth.