MetricSpec starts with a small stable core: deterministic DuckDB fixture tests for business metric SQL, with clear diagnostics and CI-friendly output.
- Contract YAML schema.
- DuckDB fixture adapter for CSV, parquet, and SQL setup files.
- Read-only SQL safety guard.
- SQL-shape text checks.
- Expected-row comparison with numeric tolerance and value diffs.
- Human, JSON, and JUnit output.
metricspec init demo,validate,run, andexplain.- Documentation, governance, CI, build, and trusted publishing workflow.
- Add
metricspec --version. - Add a
metricspec doctorcommand for environment and package diagnostics. - Improve the generated demo test to assert created file contents.
- Add richer examples for tolerances, extra columns, and SQL-shape checks.
- Add release automation checks that fail when
CHANGELOG.mdmisses the version.
- dbt manifest or MetricFlow adapter discovery.
- Postgres read-only adapter for trusted development databases.
- BigQuery/Snowflake adapter prototypes behind explicit experimental flags.
Adapters should preserve the v1 rule: no hidden network calls, no telemetry, and no credential capture in contracts or diagnostics.
- More compact row-diff summaries for large result sets.
- Optional JSON schema for report output.
- GitHub Actions summary renderer.
- SQL parser-backed shape checks once a dependency is justified.
- Hosted dashboard.
- Semantic layer replacement.
- Always-on production monitoring.
- Secret management system.