feat: auto-apply Snowflake ownership object tags on publish

README.md

@@ -3,6 +3,7 @@
 ## Metaflow API Docs
 
 - [BatchInferencePipeline](docs/metaflow/batch_inference_pipeline.md)
+- [create_ownership_registry_view](docs/metaflow/create_ownership_registry_view.md)
 - [make_pydantic_parser_fn](docs/metaflow/make_pydantic_parser_fn.md)
 - [publish](docs/metaflow/publish.md)
 - [publish_pandas](docs/metaflow/publish_pandas.md)

docs/metaflow/create_ownership_registry_view.md

@@ -0,0 +1,46 @@
+# `create_ownership_registry_view`
+
+Source: `ds_platform_utils.metaflow.registry.create_ownership_registry_view`
+
+Creates (or replaces) the central **table-ownership registry view**,
+`PATTERN_DB.DATA_SCIENCE.TABLE_OWNERSHIP_REGISTRY`. The view pivots the object tags
+applied by [`publish`](publish.md) / [`publish_pandas`](publish_pandas.md) into one row
+per table, exposing `owner`, `team`, `domain`, `project`, `status`, `sla` and `contact`.
+
+This is a one-time admin helper.
+
+## Signature
+
+```python
+create_ownership_registry_view(conn: SnowflakeConnection | None = None) -> str
+```
+
+| Parameter | Type                          | Required | Description                                                              |
+| --------- | ----------------------------- | -------: | ------------------------------------------------------------------------ |
+| `conn`    | `SnowflakeConnection \| None` |       No | Open Snowflake connection. If omitted, one is created via `get_snowflake_connection()`. |
+
+**Returns:** the executed `CREATE OR REPLACE VIEW` SQL string.
+
+## Usage
+
+```python
+from ds_platform_utils.metaflow import create_ownership_registry_view
+
+create_ownership_registry_view()
+```
+
+Then query it:
+
+```sql
+SELECT * FROM PATTERN_DB.DATA_SCIENCE.TABLE_OWNERSHIP_REGISTRY
+ORDER BY team, table_name;
+```
+
+## Notes
+
+- **No refresh needed.** A view is not materialized — it re-runs its query on every read,
+  so it is always live.
+- **~2h lag.** The view reads `SNOWFLAKE.ACCOUNT_USAGE.TAG_REFERENCES`, which itself lags
+  up to ~2 hours. For the current value of a single table's tag, use
+  `SYSTEM$GET_TAG('PATTERN_DB.DATA_SCIENCE.TABLE_OWNER', '<table>', 'table')` instead.
+- **Adoption-based.** Only tables that have at least one ownership tag appear in the view.

docs/metaflow/publish.md

@@ -14,6 +14,7 @@ publish(
     ctx: dict[str, Any] | None = None,
     warehouse: Literal["XS", "MED", "XL"] = None,
     use_utc: bool = True,
+    tags: dict[str, str] | None = None,
 ) -> None
 ```
 
@@ -22,6 +23,8 @@ publish(
 - Reads SQL from a string or `.sql` path.
 - Runs write/audit/publish operations through Snowflake.
 - Adds operation details and table links to the Metaflow card when available.
+- **Automatically applies ownership object tags to production tables** (see
+  [Ownership tags](#ownership-tags) below).
 
 ## Parameters
 
@@ -33,6 +36,7 @@ publish(
 | `ctx`        | `dict[str, Any] \| None`             |       No | Optional template substitution context for SQL operations.                                                    |
 | `warehouse`  | `Literal["XS", "MED", "XL"] \| None` |       No | Snowflake warehouse override for this operation. Supports `XS`/`MED`/`XL` shortcuts or a full warehouse name. |
 | `use_utc`    | `bool`                               |       No | If `True`, uses UTC timezone for Snowflake session.                                                           |
+| `tags`       | `dict[str, str] \| None`             |       No | Overrides for the ownership object tags applied to the published table. See [Ownership tags](#ownership-tags).|
 
 **Returns:** `None`
 
@@ -47,3 +51,58 @@ publish(
     audits=["SELECT COUNT(*) > 0 FROM PATTERN_DB.{{schema}}.{{table_name}}"],
 )
 ```
+
+## Ownership tags
+
+When publishing to **production**, `publish()` automatically applies the table-ownership
+object tags from the table-ownership RFC. The seven tags are:
+
+| Tag             | Source                                                  | Always set?     |
+| --------------- | ------------------------------------------------------- | --------------- |
+| `TABLE_OWNER`   | `ds.owner` flow tag, else owning-team alias derived from the domain (`ds-<domain>-team`), else `unknown` | yes |
+| `TABLE_TEAM`    | `data-science`                                          | yes             |
+| `TABLE_DOMAIN`  | `ds.domain` Metaflow tag, else `unknown`                | yes             |
+| `TABLE_PROJECT` | `ds.project` Metaflow tag, else `unknown`               | yes             |
+| `TABLE_STATUS`  | `active` (override allows `active`/`development`/`testing`/`deprecated`/`archived`/`retired`) | yes |
+| `TABLE_SLA`     | override only (`streaming`/`realtime`/`hourly`/`daily`/`weekly`/`monthly`/`quarterly`/`ad_hoc`/`on_demand`) | only if given |
+| `TABLE_CONTACT` | override only (Slack channel or email)                  | only if given   |
+
+> **`TABLE_OWNER` is not the run user.** Owner is resolved by priority:
+> (1) an explicit `tags={"owner": ...}` override, else
+> (2) the **`ds.owner`** Metaflow flow tag (set in CI alongside `ds.domain`/`ds.project`), else
+> (3) the owning-team alias `ds-<domain>-team` when the domain is known (e.g. domain
+> `advertising` → `ds-advertising-team`), else (4) `unknown`. We don't use `current.username`,
+> because on deployed/argo runs it resolves to a service identity (`argo-workflows`) rather
+> than a person. Set `ds.owner` on the flow for a per-flow owner, or pass `tags={"owner": ...}`
+> per call.
+
+> **`TABLE_DOMAIN` / `TABLE_PROJECT` depend on flow tags.** These are read from the
+> `ds.domain` / `ds.project` Metaflow tags. If a flow runs without them, the value falls
+> back to the literal string `unknown` and a warning is printed (the same warning used
+> for select.dev cost tracking). Make sure your flow carries `--tag "ds.domain:..."` and
+> `--tag "ds.project:..."` — these are applied automatically in CI and the standard `poe`
+> run commands in the monorepo — or pass `tags={"domain": ..., "project": ...}` explicitly.
+> Note: because owner is derived from the domain, a missing domain also means
+> `TABLE_OWNER` falls back to `unknown`.
+
+Pass `tags=` to override any value. Keys may be `owner`/`team`/`domain`/`project`/
+`status`/`sla`/`contact` (optionally `TABLE_`-prefixed):
+
+```python
+publish(
+    table_name="OUT_OF_STOCK_ADS",
+    query="sql/create_training_data.sql",
+    tags={"sla": "daily", "contact": "#ds-recsys", "status": "active"},
+)
+```
+
+Notes:
+
+- Tags are applied **only to production tables** (`DATA_SCIENCE`). Non-prod
+  (`DATA_SCIENCE_STAGE`) runs apply no tags. The publishing role needs `APPLY` on the tags.
+- The tag *definitions* must first be created once by a Snowflake admin in `DATA_SCIENCE`
+  (the RFC `CREATE TAG` setup). Until then, tagging is **skipped with a warning** — the publish
+  still succeeds.
+- Invalid `status`/`sla` values raise `ValueError` before any data is written.
+- Tagged tables surface in the `TABLE_OWNERSHIP_REGISTRY` view (see
+  `create_ownership_registry_view`).

docs/metaflow/publish_pandas.md

@@ -22,6 +22,7 @@ publish_pandas(
     use_utc: bool = True,
     use_s3_stage: bool = False,
     table_definition: list[tuple[str, str]] | None = None,
+    tags: dict[str, str] | None = None,
 ) -> None
 ```
 
@@ -30,6 +31,8 @@ publish_pandas(
 - Validates DataFrame input.
 - Writes directly via `write_pandas` or via S3 stage flow for large data.
 - Adds a Snowflake table URL to Metaflow card output.
+- **Automatically applies ownership object tags to production tables** (see
+  [Ownership tags](#ownership-tags) below).
 
 ## Parameters
 
@@ -49,9 +52,34 @@ publish_pandas(
 | `use_utc`           | `bool`                          |       No | If `True`, uses UTC timezone for Snowflake session.                                                           |
 | `use_s3_stage`      | `bool`                          |       No | If `True`, publishes via S3 stage flow; otherwise uses direct `write_pandas`.                                 |
 | `table_definition`  | `list[tuple[str, str]] \| None` |       No | Optional Snowflake table schema; used by S3 stage flow when table creation is needed.                         |
+| `tags`              | `dict[str, str] \| None`        |       No | Overrides for the ownership object tags applied to the published table. See [Ownership tags](#ownership-tags).|
 
 **Returns:** `None`
 
+## Ownership tags
+
+When publishing to **production**, `publish_pandas()` automatically applies the same
+seven table-ownership object tags as [`publish`](publish.md#ownership-tags):
+`TABLE_OWNER`, `TABLE_TEAM`, `TABLE_DOMAIN`, `TABLE_PROJECT`, `TABLE_STATUS` and
+(when provided via `tags=`) `TABLE_SLA` / `TABLE_CONTACT`.
+
+```python
+publish_pandas(
+    table_name="MY_TABLE",
+    df=df,
+    tags={"sla": "daily", "contact": "#ds-recsys"},
+)
+```
+
+- Tags are applied **only to production tables** (`DATA_SCIENCE`); non-prod runs apply none.
+- `TABLE_DOMAIN` / `TABLE_PROJECT` come from the `ds.domain` / `ds.project` Metaflow tags;
+  if a flow runs without them they fall back to the literal `unknown` and a warning is
+  printed. Ensure the flow carries those tags (automatic in CI / standard `poe` commands)
+  or pass `tags={"domain": ..., "project": ...}`. See [`publish`](publish.md#ownership-tags).
+- Tag *definitions* must first be created by a Snowflake admin (RFC `CREATE TAG` setup);
+  until then tagging is **skipped with a warning** and the publish still succeeds.
+- Invalid `status`/`sla` values raise `ValueError` before any data is written.
+
 ## Limitations
 
 - When `use_s3_stage=True`, some column data types may not map exactly as expected between pandas/parquet and Snowflake.

pyproject.toml

@@ -1,11 +1,12 @@
 [project]
 name = "ds-platform-utils"
-version = "0.4.2"
+version = "0.5.0"
 description = "Utility library for Pattern Data Science."
 readme = "README.md"
 authors = [
     { name = "Amit Vikram Raj", email = "amit.raj@pattern.com" },
-    { name = "Eric Riddoch", email = "eric.riddoch@pattern.com" }
+    { name = "Eric Riddoch", email = "eric.riddoch@pattern.com" },
+    { name = "Vinay Shende", email = "vinay.shende@pattern.com" }
 ]
 # requires-python = ">=3.7"
 dependencies = [