Skip to content

06_examples.md

Latest commit

 

History

History
286 lines (212 loc) · 7.72 KB

06_examples.md

File metadata and controls

286 lines (212 loc) · 7.72 KB
Raw

📍 Real Examples — Watch the System Think

These are not made up. These show exactly how the system processes data step by step for real scenarios.

Example 1 — Mumbai on a Humid Summer Day

Scenario: It's April. Mumbai is hot and humid. You search "Mumbai."

Step 1 — Two Doctors Give Their Readings

Measurement WeatherAPI Says Open-Meteo Says
Temperature 33°C 31°C
Humidity 84% 82%
Wind 22 km/h 18 km/h
Cloud Cover 60% 55%
Precipitation 0 mm 0 mm

Step 2 — Fusion (Blending)

Temperature = (33 × 0.60) + (31 × 0.40) = 19.8 + 12.4 = 32.2°C ≈ 32°C
Wind        = (22 × 0.70) + (18 × 0.30) = 15.4 + 5.4  = 20.8 km/h
Cloud       = (60 × 0.50) + (55 × 0.50) = 30   + 27.5  = 57.5% ≈ 58%
Humidity    = 84% (WeatherAPI only — 100% weight)

Difference between sources: |33 - 31| = 2°C → Within 5°C threshold → both are used. ✅

Step 3 — What the Engines Calculate

conditionEngine:

  • Humidity = 84%, Cloud = 58%, Precip = 0mm, Visibility = 4km
  • Result: "Partly Cloudy" + ☁️

realFeelEngine:

  • Temp = 32°C, Humidity = 84%, Wind = 21 km/h
  • Heat index kicks in (humidity > 70%)
  • Result: Feels like 39°C (7 degrees hotter than actual — humidity traps heat)

aqiHandler:

  • PM2.5: 55 µg/m³, CO: 890 µg/m³
  • AQI resolved to: 112
  • Category: "Unhealthy for Sensitive Groups" 🟠

trendEngine:

  • Previous 3 readings: 30°C → 31°C → 32°C
  • Slope: +0.5°C per reading → ↑ Rising

predictionEngine:

  • Current: 32°C, Slope: +0.5°C/30min
  • Predicted (1 hr later): 33°C with "Partly Cloudy"

rainEngine:

  • Humidity: 84% → +20 points
  • Pressure trend: slightly falling → +10 points
  • Cloud: 58% → +10 points
  • Visibility: 4km (normal) → +0 points
  • Precipitation: 0mm → +0 points
  • Total: 40% rain probability

anomalyEngine:

  • Temp: 32°C (below 45°C threshold)
  • AQI: 112 (below 200 threshold)
  • Wind: 21 km/h (below 120 threshold)
  • Result: No anomalies detected ✅

insightEngine:

  • Rain = 40%, AQI = 112, feels like 39°C, UV = 8
  • Output:
    1. "Feels like 39°C due to high humidity — stay hydrated."
    2. "Air quality is moderately unhealthy — sensitive groups avoid prolonged outdoor activity."
    3. "Rain possible later in the day — consider an umbrella."

Final Dashboard Shows:

Mumbai — Partly Cloudy ☁️
32°C / 90°F  |  Feels like 39°C
↑ Rising  |  Confidence: Medium  |  Sources: 2

Rain Probability: 40%
AQI: 112  🟠  Unhealthy for Sensitive Groups
Predicted (1hr): 33°C Partly Cloudy

💡 Insights:
  → Feels like 39°C. Stay hydrated.
  → Air quality is moderately unhealthy.
  → Rain possible later. Take an umbrella.

Example 2 — Delhi in Winter Fog

Scenario: December. Delhi winters are famous for dense fog and terrible air quality.

What the APIs Return

Measurement WeatherAPI Open-Meteo
Temperature 9°C 10°C
Humidity 93% 91%
Visibility 0.3 km 0.4 km
Wind 4 km/h 3 km/h
Cloud 90% 85%
AQI (PM2.5) 280 µg/m³ —

Fusion Result

Temperature = (9 × 0.60) + (10 × 0.40) = 5.4 + 4.0 = 9.4°C ≈ 9°C
Visibility  = (0.3 × 0.50) + (0.4 × 0.50) = 0.15 + 0.20 = 0.35 km
Humidity    = 93%

Engine Results

conditionEngine:

  • Visibility = 0.35 km (< 0.5 km), Humidity = 93%
  • Rule: IF visibility < 0.5 AND humidity > 90% → "Dense Fog" 🌫️

realFeelEngine:

  • Temp = 9°C, Wind = 4 km/h (calm)
  • No significant wind chill at this speed
  • Result: Feels like 8°C

aqiHandler:

  • PM2.5: 280 → AQI = 245
  • Category: "Very Unhealthy" 🟣
  • Tip: "Everyone should avoid outdoor activity. Wear N95 if going out."

rainEngine:

  • Humidity = 93% → +35 points
  • Cloud = 87% → +20 points
  • Visibility = 0.35 km → +15 points
  • Precipitation = 0mm → +0 points
  • Total: 70% rain probability

anomalyEngine:

⚠️ Near-Zero Visibility: 0.35 km < 0.5 km threshold
   → Alert: "🟠 Dangerous Fog — Exercise extreme caution while driving"

⚠️ Dangerous AQI: 245 > 200 threshold
   → Alert: "🔴 Very Unhealthy Air — Everyone should avoid outdoor activity"

insightEngine outputs:

  1. "Visibility is critically low at 0.35 km — do not drive without fog lights."
  2. "Air quality is very unhealthy (AQI 245). Stay indoors and close windows."
  3. "70% chance of rain or drizzle — umbrella essential."

Final Dashboard Shows:

Delhi — Dense Fog 🌫️
9°C  |  Feels like 8°C
→ Stable  |  Confidence: High  |  Sources: 2

🔴 ALERT: Very Unhealthy Air (AQI 245)
🟠 ALERT: Dangerous Fog (Visibility: 0.35 km)

Rain Probability: 70%

💡 Insights:
  → Visibility is critically low. Do not drive.
  → Stay indoors. Close windows. AQI is very unhealthy.
  → Rain likely — take an umbrella.

This is something NO normal weather app shows you. They just say "9°C, Foggy."

Example 3 — London During a Pressure Drop (Storm Coming)

Scenario: Mid-afternoon. London. Pressure has been falling for the past 3 hours.

Trend Engine Readings

Entry 1 (past): pressure = 1015 mb
Entry 2:        pressure = 1012 mb
Entry 3:        pressure = 1009 mb
Entry 4 (now):  pressure = 1006 mb

The Trend Engine calculates the slope:

  • Drop of ~3 mb every 30 minutes
  • That's -6 mb per hour — a rapid pressure drop

Anomaly Engine Fires

Rule: IF pressure drops > 5 mb per hour → Alert
Actual drop: 6 mb/hr > 5 threshold → ANOMALY DETECTED

Alert: "🟠 Rapid Pressure Drop — Storm system approaching. Conditions may worsen rapidly."

Rain Engine

Pressure trend: −6 mb/hr → +25 points (maximum severity)
Humidity: 78% → +15 points
Cloud: 70% → +10 points
Precipitation: 0.4mm → +25 points
---------------------------------
Total: 75% 🌧️

Insight Engine

  1. "A storm system is approaching — barometric pressure dropped 6 mb in the last hour."
  2. "75% chance of rain within the next 2 hours. Bring an umbrella."
  3. "Conditions are volatile — expect rapid changes."

Example 4 — Two Sources Wildly Disagree (Outlier Handling)

Scenario: You search for a small city. WeatherAPI's ground station reads correctly. Open-Meteo's regional model is off.

WeatherAPI Open-Meteo
24°C 31°C

Fusion Engine Detects the Problem

Difference = |24 - 31| = 7°C
Threshold  = 5°C

7 > 5 → OUTLIER DETECTED
→ Ignoring Open-Meteo temperature
→ Using WeatherAPI only: 24°C

Confidence Engine Responds

Sources used: 1 (only WeatherAPI — Open-Meteo discarded as outlier)
Confidence: "Low" 🟠
Reason: "High disagreement between sources — secondary data discarded"

Dashboard shows:

🟠 Low Confidence · Sources: 1 (outlier removed) · Updated: 3:45 PM

The user now knows: "This data might not be perfect. Take it with a grain of salt."

Example 5 — Cache Hit Scenario

Scenario: 50 users search for "Kolkata" within 10 minutes.

User 1 at 3:00 PM:
  → Cache check: MISS
  → Server calls WeatherAPI + Open-Meteo (takes 650ms)
  → Processes 12 engines
  → Saves to cache: "fused:kolkata" at 3:00 PM
  → Sends response

Users 2–50 (3:01 PM to 3:09 PM):
  → Cache check: HIT (saved 2–9 minutes ago, < 10 min TTL)
  → Returns from memory instantly (takes ~2ms)
  → No API calls made

User 51 at 3:11 PM:
  → Cache check: MISS (10 minutes expired, data is stale)
  → Fetches fresh data again → 650ms
  → Saves new cache entry

Result:

  • 50 users served → only 2 API calls (instead of 100)
  • Users 2–50 experience <5 ms response time vs 650 ms

Next: 07_why_better.md — Why SkyGlass Beats Normal Weather Apps