Add insulin type to LoopInsights AI analysis

Passes the user's insulin type (e.g. Fiasp, Novolog) into the therapy
snapshot and AI prompts so the model can distinguish timing issues from
dosing issues based on pharmacokinetics.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: taylorpatterson-T1D

Loop/Models/LoopInsights/LoopInsights_Models.swift

@@ -720,6 +720,7 @@ struct LoopInsightsTherapySnapshot: Codable {
     let basalRateItems: [LoopInsightsScheduleItem]
     let insulinSensitivityItems: [LoopInsightsScheduleItem]
     let carbRatioItems: [LoopInsightsScheduleItem]
+    let insulinTypeName: String?
     let capturedAt: Date
 
     struct LoopInsightsScheduleItem: Codable, Identifiable {

Loop/Services/LoopInsights/LoopInsights_AIAnalysis.swift

@@ -179,6 +179,20 @@ final class LoopInsights_AIAnalysis {
         - CAFFEINE: Active caffeine >100mg can increase insulin resistance and glucose variability. \
           Factor caffeine timing into your assessment of glucose patterns, especially morning highs.
 
+        INSULIN TYPE CONTEXT — When insulin type data is provided:
+        - RAPID-ACTING (Novolog/Humalog/Apidra): Onset ~15 min, peak activity ~75 min, duration ~6 hrs. \
+          Standard absorption profile. Post-meal glucose should begin dropping within 60-90 min of bolus. \
+          Pre-bolusing 15-20 min before meals is effective. Corrections take 2-3 hrs to fully resolve.
+        - ULTRA-RAPID (Fiasp/Lyumjev): Onset ~2-5 min, peak activity ~55 min, duration ~6 hrs. \
+          Faster onset and earlier peak means: \
+          Post-meal spikes should be smaller — if spike is still large, CR is more likely the issue (not timing). \
+          Corrections resolve faster (~1.5-2 hrs) — if glucose stays high after correction, ISF is likely too high. \
+          Less tail stacking risk — basal adjustments can be slightly more aggressive per time block. \
+          Pre-bolusing is less critical — a large spike despite on-time bolusing strongly suggests weak CR.
+        - Use insulin type to distinguish TIMING issues from DOSING issues. A Novolog user with post-meal \
+          spikes that resolve by hour 3 may need more pre-bolus time, not a CR change. A Fiasp user with \
+          the same pattern likely needs a CR adjustment since Fiasp should already be active.
+
         RESPONSE FORMAT:
         Respond with valid JSON in this exact structure:
         {
@@ -259,6 +273,11 @@ final class LoopInsights_AIAnalysis {
             prompt += "- \(formatTime(item.startTime)): \(String(format: "%.1f", item.value)) g/U\n"
         }
 
+        if let insulinType = settings.insulinTypeName {
+            prompt += "\n### Insulin Type\n"
+            prompt += "- Currently using: \(insulinType)\n"
+        }
+
         prompt += "\n"
 
         // Glucose stats

Loop/Services/LoopInsights/LoopInsights_DataAggregator.swift

@@ -155,10 +155,13 @@ final class LoopInsights_DataAggregator {
             LoopInsightsTherapySnapshot.LoopInsightsScheduleItem(startTime: $0.startTime, value: $0.value)
         } ?? []
 
+        let insulinTypeName = settings.insulinType?.title
+
         return LoopInsightsTherapySnapshot(
             basalRateItems: basalItems,
             insulinSensitivityItems: isfItems,
             carbRatioItems: crItems,
+            insulinTypeName: insulinTypeName,
             capturedAt: Date()
         )
     }