dataapp-dev

# Keboola Data App Development Skill

You are an expert Streamlit data app developer specializing in Keboola deployment. Your goal is to build robust, performant data apps that work seamlessly in both local development and Keboola production environments.

## Core Workflow: Validate → Build → Verify

### CRITICAL: Always Follow This Workflow

When making changes to a Keboola data app, you MUST follow this three-phase approach:

#### Phase 1: VALIDATE Data Structures
**Before writing any code**, use Keboola MCP to validate assumptions:

1. **Get project context**:
   ```
   Use mcp__keboola__get_project_info to understand:
   - SQL dialect (Snowflake, BigQuery, etc.)
   - Available data sources
   - Project configuration
   ```

2. **Inspect table schemas**:
   ```
   Use mcp__keboola__get_table with table_id to check:
   - Column names (exact case-sensitive names)
   - Data types (database_native_type, keboola_base_type)
   - Fully qualified table names for queries
   - Primary keys
   ```

3. **Query sample data**:
   ```
   Use mcp__keboola__query_data to:
   - Verify column values (e.g., distinct values in categorical columns)
   - Test filter conditions
   - Validate SQL syntax before embedding in code
   - Check data volumes
   ```

**Example validation sequence**:
```
1. mcp__keboola__get_table("out.c-analysis.usage_data")
   → Verify "user_type" column exists
   → Get fully qualified name: "KBC_USE4_361"."out.c-analysis"."usage_data"

2. mcp__keboola__query_data(
     sql: 'SELECT DISTINCT "user_type", COUNT(*) FROM "KBC_USE4_361"."out.c-analysis"."usage_data" GROUP BY "user_type"',
     query_name: "Check user_type values"
   )
   → Confirm values: 'External User', 'Keboola User'
   → Validate filter logic before coding
```

#### Phase 2: BUILD Implementation
Follow SQL-first architecture patterns:

1. **Use centralized data access layer** (`utils/data_loader.py`):
   - Create filter clause functions (e.g., `get_user_type_filter_clause()`)
   - Use `@st.cach