Create account

DEV Community

Geets

Posted on Mar 1

GlowInspo: When AI Dresses My Monday Anxiety

#devchallenge #weekendchallenge #showdev #ai

DEV Weekend Challenge: Community

This is a submission for the DEV Weekend Challenge: Community

The Community

GlowInspo is built for women navigating high-context mornings.
Remote workers. Founders. Creators. Consultants.
Women who move between emotional states before 9AM.

7:03AM scenario:
Rain. Back-to-back meetings. Overwhelmed.
Instead of grabbing the first clean sweater, I want something that stabilizes me.

GlowInspo serves the the decision-fatigued morning crowd — women who want clothing to support their mental state, not compete with it.

It’s not about trends.
It’s about operational energy.
Just the 2-minutes reset!

Move quickly between emotional states
Use clothing as psychological support
Need fast, aligned decisions before work begins

What I Built

GlowInspo: 2-minute AI-powered emotional dressing reset

Check-in: "overwhelmed monday morning"
Energy: "Stabilize" (Soften/Stabilize/Brighten/Amplify)
Claude Ritual: Arrival → Reframe → Direction
Gallery: 900+ Pinterest outfits (energy + tone matched)
Memory: OOTD uploads → identity snapshots

Demo

Live Demo: https://glowinspo-cp4otqkvuwuzdukvy4nkvn.streamlit.app/?embed_options=dark_theme
Video: https://youtu.be/x679HuRIklU

Code

Geets1810 / Glowinspo

GlowInspo

A 2-Minute Styling Ritual for Women Who Dress With Intention

GlowInspo is an emotion-aware styling ritual built for women balancing busy workdays, creative lives, and internal emotional shifts Instead of dressing for trends, it helps you dress for the energy you want to bring into your day. This version was built for the DEV Challenge to serve women who want less decision fatigue and more emotional alignment.

💛 The Community

GlowInspo is designed for:

Women in fast-paced professional roles
Creatives navigating identity and visibility
Anyone who uses clothing as self-expression

Women who want to feel intentional, not overwhelmed Getting dressed shouldn’t feel like another task.

It can be a reset.

✨ How It Works

You check in: How are you arriving today? You choose how you want to show up:

Soften
Stabilize
Brighten
Amplify

GlowInspo generates emotionally aligned guidance.

You explore curated outfit suggestions If inspired, you upload your…

View on GitHub

How I Built It

GlowInspo is built as a lightweight, stateful Streamlit application designed for fast emotional interaction.

Architecture
The app runs as a single-page Streamlit interface using session state to manage:

Current recommendation set
Feedback state (Inspired / Not Quite / Self-Styled)
Emotional memory history
Controlled reshuffle logic

Session state prevents unnecessary recomputation and stabilizes UI behavior across interactions.

Data Model & Tagging
Outfit data is stored in a structured CSV dataset.
Each item contains:

energy_bucket (Soften, Stabilize, Brighten, Amplify)
tone_tags_v2 (multi-label descriptors such as structured, minimal, bold)
category
image_url

Tagging is multi-layered:

Primary filter → energy bucket
Secondary ranking → tone overlap score

Tone overlap is calculated dynamically:

filtered["match_score"] = filtered["tone_list"].apply(
    lambda tl: len(set(tl).intersection(set(tones)))
)

This ensures emotional alignment before aesthetic variation.
AI-Assisted Tagging

Tone tags were refined using a constrained prompt-based GenAI workflow.
Claude was used to:

Normalize tone vocabulary across 990 items
Ensure consistent multi-label tagging (2–3 tones per item)
Align items to one of four defined energy buckets

The model was restricted to a fixed tone taxonomy to prevent uncontrolled label drift.

This hybrid approach (AI-assisted tagging + structured constraints) allowed scalable tagging while maintaining emotional consistency.

Emotional Guidance Layer
Claude (Anthropic API) generates constrained structured output:

Arrival
Reframe
Direction
Energy
Tone tags

The prompt restricts output format and allowed tone vocabulary to maintain system consistency and prevent drift.

Controlled Recommendation Logic
Outfits are sampled with:

Category-aware diversification
Frozen suggestions until explicit reshuffle
Manual reset only via “Not Quite”

This avoids infinite scroll behavior and reduces cognitive overload.
Identity Memory
User-uploaded OOTDs are stored in session state with:

Mood
Energy
Composite identity label (e.g., “Tired + Amplify”)
Image Memory retrieval is energy-matched to surface prior successful identity states.

This prevents unnecessary reshuffling and reduces scroll fatigue.

Scaling & Future Extensions - Designing Human Centered AI Products

GlowInspo is currently session-based and lightweight by design, but the architecture can scale in several directions:

Persistent Storage

Move emotional memory and user-uploaded images to cloud storage (e.g., S3 or GCS)
Store structured mood + energy snapshots in a database (Postgres / Firestore)
Enable cross-session identity memory instead of session-only state

Richer Recommendation Engine