Repository navigation
Description
Summary
SQLMesh already has a prototype Metrics feature and rich model metadata (grain, references, columns, environments). Semantic layers provide production-ready capabilities (dimensions, measures, joins, query API, caching).
Today these are completely separate — teams maintain SQLMesh models and hand-write semantic layer schemas. This issue proposes connecting them: use SQLMesh as the source of truth and auto-generate Ossie files for semantic layers to pick up.
Why Ossie
Ossie is a strong fit for SQLMesh to align to:
- Apache 2.0 — same license as SQLMesh; no proprietary lock-in for the core semantic layer and query API
- Long-term viability - sponsored by 60+ organizations, such as Snowflake, Cube, etc. --
dbt-metricflowis also aligned to this standard. - Complementary roles — SQLMesh owns transformation, environments, and versioning; Semantic layers owns metric serving and performance
Background
What SQLMesh already has
- Models — tables/views SQLMesh builds and versions
- Grain & references — how models join (used today by the metrics rewriter and table_diff)
- Metrics (prototype) —
METRIC(...)definitions inmetrics/*.sqlthat expand to SQL via sqlmesh rewrite - Environments — dev/prod isolation with virtual views
- Metrics is documented as prototype-only and not integrated into plan/run. The core
rewriterandReferenceGraphexist but the feature was never fully developed.
What Ossie provides
Semantic layer standard that can be ingested by technologies such as Cube. It solves the serving problem Metrics was pointing at but didn't finish.
Proposal (high level)
Instead of maintaining semantic layer YAML/JS by hand (or as a one-off file export), SQLMesh would compile Ossie schemas.
| Ossie Standard | SQLMesh source |
|---|---|
| Semantic Model | SQLMesh Project |
| Datasets | SQLMesh models |
| Fields / Dimensions | Model columns / types |
| Primary keys | grain |
| Relationships | grain + references |
| Metrics | METRIC(...) definitions |
| ai_context | Model/column descriptions |
Delivery could be:
- Runtime — Semantic layer adapters load schemas from SQLMesh on demand (no generated files in repo)
- Export (optional) — sqlmesh
create_ossie_modelsfor GitOps / air-gapped deploys - Optional extra:
sqlmesh[ossie]for integration tooling (similar tosqlmesh[dbt],sqlmesh[web]).
Why this fits SQLMesh
- Completes the arc Metrics started without SQLMesh becoming a headless BI server
- Leverages grain/references — gives them a concrete consumer beyond the prototype
rewriter - Gives
METRIC(...)definitions a production path — today they're prototype-only with nowhere to go; this connects them to a real query API - Aligns with existing integration pattern — mirrors
sqlmesh[dbt]andsqlmesh[dlt]
Open questions
- Opt-in model: global config flag, per-model flags, or both?
- How to handle models without grain/references (linting rule?)
- Runtime vs export-first for v1?
- export seems easy to validate if this can work out
- etc.
Possible v1 scope
- Schema compiler: Context → Ossie YAML
-
sqlmesh create_ossie_models(export mode, mirrors create_external_models) -
sqlmesh ossie validate— compile-check joins/measures against project metadata
References
SQLMesh Metrics overview
SQLMesh grain/references
Example metrics: examples/sushi/metrics/metrics.sql
Activity
[-]RFC - Auto-generate Cube semantic layer from SQLMesh models and Metrics[/-][+]RFC - Auto-generate Ossie files from SQLMesh models and Metrics[/+]cmgoffena13 commentedon Nov 3, 2026
Recently became aware of
sqlmesh-cubea pre-release pypi package that tobiko created. Only a few updates in 2024 were made. You can install it normally, but when it installed, it actually ends up being another packaged version ofsqlmesh, which issqlmeshwith these changes:New Files
sqlmesh/core/cube.pysqlmesh/core/openai_client.pysqlmesh/core/prompts/__init__.pysqlmesh/core/prompts/cube_docs.pysqlmesh/core/prompts/cube_generate.txtsqlmesh/core/test/cube_test.pyModified Files
sqlmesh/cli/main.py-- addscube_generatesqlmesh/core/context.py-- adsContext.cube_generateNeed to look into them to understand the attempted approach.