← Back to home

Wadrobe – AI Powered Personal Stylist

A playful Airflow-powered assistant that checks today's weather, calls an LLM, and sends you an outfit suggestion before you leave the house — turning ETL skills into a tiny lifestyle product.

Fun side project · 2025Airflow · LLMs · Weather APIView project on GitHub
Situation
Deciding what to wear every morning is a tiny but persistent decision tax. Most weather apps show temperature, not "what should I actually wear?"
Task
Build a fun, low-friction assistant that turns weather data into concrete outfit suggestions — using the same stack normally used for "serious" data pipelines.
Action
Implemented an Airflow DAG that fetches weather data from a public API, feeds it to an LLM (Ollama), and generates a short, weather-aware outfit recommendation. The result is sent as a simple push notification via Pushover.
Result
Users get a daily, context-aware nudge: "here's what to wear" for sunny, rainy, or cold days — showing how data engineering skills can power delightful everyday experiences, even through a plain text alert.

High-level architecture

Wadrobe is an Airflow DAG glued to a weather API, an LLM prompt, and a notification channel — effectively a tiny decision-support system for your wardrobe. In the real project, the final output is a concise Pushover alert sent to your phone.

1Step 1 – Airflow fetches weather

A scheduled DAG calls a public weather API for your city and day.

Airflow scheduler & DAG
  • Runs on a morning schedule (e.g. 7:00 AM local time).
  • Calls a weather API with your location and desired forecast window.
  • Passes temperature, conditions, and time of day forward.
LLM outfit generator
  • Sends a compact JSON of weather data to an LLM via Ollama / OpenAI.
  • Prompt includes style preferences and "avoid crazy suggestions" guardrails.
  • Returns a short summary + bullet list of outfit components.
Notification delivery
  • Formats the suggestion into a concise Pushover message (title + body).
  • Sends a single alert to your device — no UI, just a helpful ping.
  • Optionally logs suggestions to a table for later analysis (seasonality, preferences).

Demo: how outfits change with the weather

In the actual implementation, Wadrobe sends a simple Pushover alert like: "Today is rainy and 16°C · here's what to wear". The animated card below is a visual mock of that message — a portfolio-friendly way to show how the suggestion logic behaves for sunny, rainy, and cold conditions.

(Auto-rotating every few seconds)
☀️Sunny day
24°C · clear sky
Pushover-style alert

Light, breathable layers with sun protection.

  • Lightweight cotton t-shirt
  • Chino shorts or breathable jeans
  • Sneakers or loafers
  • Sunglasses + light cap
This card represents the text content sent via a Pushover notification.LLM-powered suggestion

Tech stack

Apache AirflowPythonWeather APILLMs (Ollama / OpenAI)Pushover notifications

Wadrobe is intentionally small, but it demonstrates something important: once you have orchestration, APIs, and LLMs wired up, you can build tiny, delightful assistants for everyday life — even if the final UX is just a clean, well-timed push notification.

Explore Wadrobe on GitHub