Reflecta | Devpost

Inspiration

Most people don’t talk about how they’re doing — even to themselves.
But they do write.

The inspiration for Reflecta came from noticing how journaling is often treated as a one-off coping mechanism instead of a long-term mirror. Pages get written, closed, and forgotten. The emotion disappears into text.

We asked a simple question:

What if journaling could show you who you’re becoming over time?

Not as therapy.
Not as diagnosis.
Just as reflection.

What it does

Reflecta is a privacy-first identity reflection app.

Users write daily journal entries. From that writing, Reflecta:

Extracts emotional signals (sentiment, stress, intensity, energy)
Converts them into a daily Mental Health / Identity Score (MLH) on a 0–100 scale
Visualizes those scores over time as a calm, non-judgmental timeline
Highlights Identity Shifts — moments where something clearly changed
Generates a short AI-written reflection written in the user’s own tone

There are no diagnoses, no labels, and no advice.

Reflecta is not a therapist.
It’s a mirror.

How we built it

The system follows a simple but intentional pipeline:

Write → Analyze → Score → Visualize → Reflect

Core Stack

Next.js (App Router) for the frontend and routing
Tailwind CSS for a calm, minimal UI
Supabase for authentication and Postgres storage
Google Gemini API for emotional signal extraction and reflection generation
Custom MLH Algorithm (rule-based + weighted factors)
Recharts for visualizing identity shifts over time

We deliberately separated:

Scoring logic (deterministic, transparent)
AI reflection (expressive, human, non-clinical)

This allowed us to avoid black-box scoring while still using AI for narrative insight.

Challenges we ran into

Technical Challenges

Separating “analysis” from “advice”
Early AI outputs sounded too much like therapy. We had to aggressively constrain prompts to ensure the model acted as a reflective narrator — not a mental health professional.
Designing a meaningful score
Reducing complex emotion into a single number felt dangerous. We iterated multiple times to ensure the MLH score represented change, not judgment.

Human & Team Challenges

This project wasn’t just technically hard — it was emotionally uncomfortable.

We were building something that forces introspection. That meant:

Debating ethics late into the night
Questioning whether we were crossing a line
Scrapping features that felt impressive but wrong

At one point, we almost pivoted away entirely because we didn’t want to unintentionally harm users. The breakthrough came when we reframed the app:

Reflecta does not tell you who you are — it shows you who you were.

That clarity aligned the entire team.

Accomplishments that we're proud of

Building an AI-powered product that explicitly avoids medical framing
Designing a scoring system that is interpretable and user-respecting
Creating reflections that feel personal without being invasive
Maintaining a strong ethical stance under hackathon pressure
Shipping a product that feels calm, not noisy — rare in AI apps ## What we learned
AI doesn’t need to advise to be helpful
Numbers can be reflective if framed carefully
Ethics is not a feature — it’s a design constraint
Emotional products require slower, more deliberate decisions
The hardest problems aren’t technical — they’re conceptual

What's next for Reflecta

Next, we want to expand Reflecta into a deeper identity system:
Weekly and monthly identity summaries
Optional voice & video reflections generated from past entries
Stronger control over how data is interpreted and visualized
Long-term pattern detection without surveillance
Even clearer consent and transparency tools