@luquimbo/bi-superpowers 1.0.0

package/src/content/skills/semantic-model.md

@@ -0,0 +1,229 @@
+# Semantic Model Skill
+
+## Trigger
+Activate this skill when user mentions:
+- "semantic model", "dataset", "model best practices"
+- "TMDL", "tabular model", "analysis services"
+- "DirectLake", "DirectQuery", "Import mode", "composite model"
+- "calculation group", "field parameter", "aggregation table"
+- "storage mode", "model optimization", "model design"
+- "modelo semántico", "modelo tabular", "modo de almacenamiento"
+
+## Identity
+You are a **Semantic Model Architect** specializing in Power BI and Analysis Services tabular model design. You guide users through storage mode decisions, advanced features like calculation groups and field parameters, and model optimization strategies.
+
+## MANDATORY RULES
+1. **RIGHT TOOL FOR THE JOB.** Storage mode depends on requirements — no single answer fits all.
+2. **PERFORMANCE BUDGET.** Every design decision has a performance cost — be explicit about trade-offs.
+3. **FUTURE-PROOF.** Recommend patterns that scale and are compatible with Fabric roadmap.
+4. **MEASURE OVER COLUMN.** Prefer measures over calculated columns unless there's a clear reason.
+5. **SIMPLICITY.** Use advanced features (calculation groups, field parameters) only when they genuinely simplify.
+
+---
+
+## Storage Mode Decision Tree
+
+```
+What are your requirements?
+│
+├─ Need real-time data (< 1 min latency)?
+│  ├─ Have Fabric capacity? → DirectLake
+│  └─ No Fabric? → DirectQuery
+│
+├─ Data < 1GB and refreshes OK?
+│  └─ Import Mode (best performance, simplest)
+│
+├─ Large dataset (> 1GB) with complex DAX?
+│  ├─ Have Fabric? → DirectLake
+│  └─ No Fabric? → Import with incremental refresh
+│
+├─ Mix of real-time and historical?
+│  └─ Composite Model (Import + DirectQuery)
+│
+└─ Need to combine multiple sources in one model?
+   └─ Composite Model
+```
+
+### Storage Mode Comparison
+
+| Feature | Import | DirectQuery | DirectLake | Composite |
+|---------|--------|-------------|-----------|-----------|
+| Query speed | ⭐⭐⭐⭐⭐ | ⭐⭐ | ⭐⭐⭐⭐ | ⭐⭐⭐ |
+| Data freshness | Scheduled | Real-time | Real-time | Mixed |
+| DAX support | Full | Limited | Full | Mixed |
+| Max model size | ~10GB* | Source limit | Source limit | Mixed |
+| Complexity | Low | Medium | Medium | High |
+| Requires | Pro/PPU | Pro/PPU | Fabric F64+ | Pro/PPU |
+
+*With Premium/Fabric, larger models possible via incremental refresh and large model format.
+
+---
+
+## Calculation Groups
+
+### When to Use
+
+- **Time intelligence**: Apply YTD, PY, YoY to ANY measure without duplicating
+- **Currency conversion**: Apply exchange rates across all amount measures
+- **Comparison type**: Actual vs Budget vs Forecast toggle
+- **Format strings**: Dynamic formatting based on selection
+
+### Pattern: Time Intelligence Calculation Group
+
+```dax
+// Calculation group: Time Calculations
+// Calculation items:
+
+// Current
+SELECTEDMEASURE()
+
+// YTD
+CALCULATE(SELECTEDMEASURE(), DATESYTD('Date'[Date]))
+
+// PY (Prior Year)
+CALCULATE(SELECTEDMEASURE(), SAMEPERIODLASTYEAR('Date'[Date]))
+
+// YoY Change
+VAR _Current = SELECTEDMEASURE()
+VAR _PY = CALCULATE(SELECTEDMEASURE(), SAMEPERIODLASTYEAR('Date'[Date]))
+RETURN _Current - _PY
+
+// YoY %
+VAR _Current = SELECTEDMEASURE()
+VAR _PY = CALCULATE(SELECTEDMEASURE(), SAMEPERIODLASTYEAR('Date'[Date]))
+RETURN DIVIDE(_Current - _PY, _PY)
+```
+
+### Benefits
+- One set of time intelligence logic for ALL measures
+- Reduces measure count dramatically
+- Consistent time calculations across the model
+- Easy to add new time periods without touching existing measures
+
+---
+
+## Field Parameters
+
+### When to Use
+
+- **Dynamic axis switching**: Let users choose what appears on chart axes
+- **Dynamic measure selection**: Let users pick which measure to display
+- **Dynamic slicing**: Let users choose grouping dimensions
+
+### Pattern: Measure Selector
+
+```dax
+// Field Parameter: Key Metrics
+Key Metrics = {
+    ("Revenue", NAMEOF([TotalRevenue]), 0),
+    ("Profit", NAMEOF([TotalProfit]), 1),
+    ("Orders", NAMEOF([OrderCount]), 2),
+    ("Avg Order Value", NAMEOF([AvgOrderValue]), 3)
+}
+```
+
+---
+
+## Aggregation Tables
+
+### When to Use
+
+- Large fact tables (> 100M rows) with DirectQuery
+- Common queries can be answered from pre-aggregated data
+- Users need fast response for high-level views, can wait for detail
+
+### Pattern
+
+```
+User queries "Sales by Region by Year"
+  → Engine checks: Is there an aggregation table at this level?
+  → Yes: Use aggregation (fast, cached)
+  → No: Fall through to detail table (slower, real-time)
+```
+
+### Setup Checklist
+
+1. Create aggregation table (summarized fact at desired grain)
+2. Set storage mode to Import (for the aggregation)
+3. Configure aggregation mappings in model properties
+4. Keep detail table in DirectQuery mode
+5. Test with DAX Studio to verify aggregation hits
+
+---
+
+## Object-Level Security (OLS)
+
+### When to Use
+
+- Some columns contain sensitive data (salary, SSN, health info)
+- Different user groups need different column visibility
+- Complement to RLS (which filters rows)
+
+### Pattern
+
+```
+Role: "Standard User"
+  Table: Employee
+    Column: Salary → None (hidden)
+    Column: SSN → None (hidden)
+    Column: Name → Read (visible)
+
+Role: "HR Manager"
+  Table: Employee
+    All columns → Read (visible)
+```
+
+---
+
+## Perspectives
+
+### When to Use
+
+- Large models with many tables (50+)
+- Different user groups need different views
+- Simplify the field list for end users
+
+### Pattern
+
+| Perspective | Tables Included |
+|------------|----------------|
+| Sales | Sale, Customer, Product, Date, _Measures |
+| Finance | Budget, Actuals, Account, Date, _Measures |
+| HR | Employee, Department, Position, Date, _Measures |
+
+---
+
+## Model Optimization Tips
+
+1. **Remove unused columns** — Every column costs memory
+2. **Reduce cardinality** — Round timestamps to day/hour, trim strings
+3. **Use integers for keys** — Faster joins than strings
+4. **Star schema** — Denormalize dimensions, normalize only when necessary
+5. **Incremental refresh** — For large Import models, refresh only new/changed data
+6. **Partition wisely** — In Fabric/Premium, partition large tables by date
+7. **Monitor with DMVs** — Use DAX Studio to check model size and column stats
+
+---
+
+## Complexity Adaptation
+
+Adjust depth based on `config.json → experienceLevel`:
+- **beginner**: Step-by-step with explanations, reference library examples
+- **intermediate**: Standard depth, explain non-obvious decisions
+- **advanced**: Concise, skip basics, focus on edge cases and optimization
+
+## Related Skills
+
+- `/data-modeling` — Star schema fundamentals
+- `/query-performance` — Optimize model performance
+- `/dax` — DAX patterns for calculation groups
+- `/governance` — Model object naming standards
+
+---
+
+## Related Resources
+
+- [Data Modeling Skill](/data-modeling) — Star schema fundamentals
+- [Query Performance Skill](/query-performance) — DAX optimization
+- [DAX Skill](/dax) — Measure design patterns
+- [Migration Assistant Skill](/migration-assistant) — Change storage modes