…load

perf(seeddata): replace MIN/MAX with ORDER BY LIMIT 1 for faster range queries
refactor(seeddata): split range query into two single-value queries
chore(seeddata): reduce query timeout from 2m to 30s

Allow callers to force a specific range column instead of using
per-model detection. This enables consistent range queries across
models when needed.

…nctions

This introduces functionality to sanitize IPv4 and IPv6 columns during seed data generation by hashing them with a shared salt, ensuring consistent anonymization across related data sets while preserving the original IP address type structure (IPv4 vs IPv6, including IPv4-mapped IPv6). This required adding salt generation, ClickHouse schema introspection (`DESCRIBE TABLE`), and dynamic SQL query construction to replace raw column selection with sanitization expressions in `lab_xatu_cbt_generate_seed_data` and `lab_xatu_cbt_generate_transformation` commands.

… transformation

This change updates the seed data generation and transformation commands to report which IP columns were sanitized during the process, improving user visibility into data masking operations.

feat: Sanitize IPs for parquet file uploads

…est generation

This change introduces predefined time range presets (e.g., "Last 5 minutes", "Last 1 hour") to simplify selecting the time window for seed data generation during transformation tests, along with accounting for ingestion lag when calculating the effective maximum time.

…ormation generation

feat: Defined ranges

…r-pt2

feat: add generate-transformation-test command for xatu-cbt

- Remove rangePresets, ingestionLagBuffer and manual range prompts
- Add --duration flag and new seeddata/discovery.go module
- Integrate Claude-based range strategy generation with fallback heuristic
- Validate data availability per model before generation
- Streamline UX: single duration prompt, AI summary, confirmation flow

…L filter analysis

- add support for entity/dimension tables (no time range) via intervalType
- read intermediate transformation SQL to extract WHERE clause filters
- extend discovery prompt to include correlation filters for dimension tables
- add FilterSQL and CorrelationFilter to TableRangeStrategy for precise filtering
- improve fallback discovery to handle entity models and missing ranges
- normalize YAML field names and fix unquoted datetime values in Claude responses
- extend QueryRowCount and GenerateOptions to accept additional SQL filters
- add S3 Cache-Control: no-cache header for fresh seed data downloads

feat: enhance seed-data discovery with dimension table support and SQL filter analysis

feat: replace manual range selection with AI-driven discovery

- add detection for common field name typos like "primaryrangeType"
- expand normalizeDiscoveryYAMLFields to handle PascalCase, snake_case
 and other variations Claude might output
- enhance discovery prompt with stricter formatting rules and examples
- include full YAML content in error messages for easier debugging
- log when YAML normalization actually changes field names