Seasonal Allergy Radar

Overview

Hay fever suffers through the same 6-week misery every year, then repeats it — because nothing connects three facts they already have: when their symptoms spike, what pollinates when in their region, and that antihistamines work best started 1–2 weeks BEFORE the season.

This skill connects them:

1. Symptom log — 10-second daily entries: overall score 0–5 + affected systems (eyes/nose/lungs/skin). No app account, no cloud — a JSON file.
2. Regional pollen calendar — month-by-month tree/grass/weed risk for your climate zone (temperate-Northern, Mediterranean, continental, subtropical, coastal-low-pollen). Curaated from aerobiology network patterns (see references).
3. Correlation engine — aligns your log against the calendar, tests lags 0–7 days, and reports which pollen class(s) your symptoms track with correlation coefficient + your personal onset/peak.
4. 14-day personal forecast — calendar × your fitted severity curve → expected symptom score per day, with "start/step-up medication" flags.
5. Pre-medication planner — antihistamine start dates (season minus 10–14 days), nasal steroid ramp (they need days-weeks to peak effect), ophthalmologist-grade eye-drop timing, and an allergist-visit brief from your data.

When to Use

• "Every April I'm wrecked — is it tree or grass?" → correlation + calendar
• Planning a move or a trip during your allergy season
• "When should I start my antihistamine this year?" → pre-med plan
• Deciding if OTC management is enough or symptoms warrant an allergist
• Multi-year: "is it getting worse?" → season-over-season severity trend

Don't use for: asthma management, food allergy, anaphylaxis history, or substituting an allergist's testing (skin prick/IgE) — this is a pattern instrument, not a diagnostic one. Red-flag symptoms (wheezing, chest tightness) → doctor, not a log.

Quick Start

# 1. Print a month of blank log entries (paper or notes-app friendly)
python3 scripts/allergy.py log init --from 2027-03-01 --days 31 --out my_log.json

# 2. Record a day (takes 10 seconds)
python3 scripts/allergy.py log add --log my_log.json --date 2027-03-14 \
    --score 4 --eyes --nose

# 3. What's in the air this month where I live?
python3 scripts/allergy.py calendar --zone temperate-northern

# 4. Correlate my log against the calendar (which pollen is me?)
python3 scripts/allergy.py correlate --log my_log.json --zone temperate-northern

# 5. My personal 14-day forecast + med flags
python3 scripts/allergy.py forecast --log my_log.json --zone temperate-northern \
    --from 2027-04-01

# 6. Pre-medication plan for my main season
python3 scripts/allergy.py plan --zone temperate-northern --target tree \
    --year 2027 --antihistamine cetirizine --nasal-steroid fluticasone

Symptom Log Format

One JSON file, one entry per day:

[
  {"date": "2027-03-14", "score": 4, "systems": ["eyes", "nose"], "note": "driving with windows open"}
]

• score 0 (nothing) to 5 (incapacitating)
• systems any of eyes nose lungs skin
• Best data: log EVERY day through the season, even good days — correlation needs the zeros as much as the spikes.

Pollen Calendar Zones

(Calendars are 3-week windows around these peaks; script holds the month-level table.)

How Correlation Works

For each pollen class (tree/grass/weed):

1. Build the calendar's daily risk curve for your zone (0–3 scale).
2. For lag L in 0–7 days: Pearson r between your score(t) and pollen(t−L).
3. Best |r| wins; report r, lag, your days-worst percentile, inferred onset and peak dates from your own entries.
4. If no class clears |r| ≥ 0.35, verdict = "no clear pollen driver — consider dust-mold-indoor factors or an allergist."

The lag matters clinically: symptoms can trail pollen by 1–3 days (priming effect), and antihistamines taken after onset fight a losing battle — receptor saturation works best pre-exposure.

Common Pitfalls

1. Logging only bad days. Correlation needs baseline zeros; a log of only flare days correlates with everything.
2. Starting meds when symptoms start. Second-generation antihistamines reach steady effect in 1–3 days but work best on unstained mast cells; nasal steroids need 1–2 weeks to plateau. The plan command back-dates starts from the season onset accordingly.
3. Wind vs insect pollinated. Showy flowers are rarely your problem; birch/grass/ragweed are wind-pollinated and produced in millions of grains per day. The calendar is built from wind-pollinators.
4. Rain resets. Pollen drops during rain, spikes after (fragrance + burst grasses). A single rainy week can look like "cure" — it's weather, not tolerance.
5. Cross-reactivity confusions. Birch-apple, mugwort-celery oral allergy syndrome: your "sudden fruit allergy" each April may be the tree pollen priming your mouth. Correlate before eliminating foods.

Verification Checklist

• Log has ≥ 21 days with honest zeros included
• Zone matches actual climate (not vacation spots in the log period)
• Correlation |r| ≥ 0.35 before trusting any trigger verdict
• Med start dates back-dated before calendar onset
• Red-flag symptoms (wheezing/chest) routed to a doctor, not the log

One-Shot Recipes

"What am I allergic to?" (one season of data):

python3 scripts/allergy.py log init --from 2027-03-01 --days 90 --out s.json
# ...log daily...
python3 scripts/allergy.py correlate --log s.json --zone temperate-northern

This year's pre-med calendar (birch-sensitive):

plan --zone temperate-northern --target tree --year 2027 \
     --antihistamine cetirizine --nasal-steroid fluticasone

Trip planning — "April in Lyon vs Lisbon":

calendar --zone mediterranean
forecast --log my_log.json --zone mediterranean --from 2027-04-05