Skills 2026-10-12 18:29 UTC
Dataverse-skillsv1.16.1
Version [v1.16.1] committed by arorashivam96. Update Category: Feature Addition & Enhancement. Commit details: fix: reduce raw urllib in dataverse-skills (SDK bulk delete/batching + fallback nudge) (#151). This update introduces new capabilities and module extensions.
CommitRaw commit message
fix: reduce raw urllib in dataverse-skills (SDK bulk delete/batching + fallback nudge) (#151)
## What
Reduce hand-rolled raw `urllib` traffic from the dataverse-skills plugin
in favor of the SDK.
## Why
Production WRS telemetry (2-day window): `dataverse-skills` urllib =
**542K requests** (the largest transport). **~87.5% is on operations the
SDK already covers** (CRUD + metadata reads); only ~1.7% is a genuine
SDK gap. Two root causes: (1) skills carried false "SDK can't" claims
that pushed agents to raw HTTP; (2) the raw helper
`scripts/auth.py:get_plugin_headers()` -- the single chokepoint that
stamps the `Python-urllib` UA -- was framed as a peer to the SDK, not a
fallback.
## Changes
**Skills -- correct false "SDK can't" claims:**
- dv-metadata: schema metadata is the single biggest raw-urllib driver.
SDK v1.1.0 (PR #206) adds SDK-native column constraints, closing all
three operations that previously forced urllib: (1) **create** with a
dict spec -- `add_columns(table, {"col": {"type": "money",
"min_value":..., "max_value":..., "precision":...}})` (keys:
`max_length`/`min_value`/`max_value`/`precision`/`format`/`required`/`display_name`);
(2) **update** in place -- `update_column` / `update_columns`; (3)
**typed read** -- `list_columns(..., typed=True)` / `get_column(...,
typed=True)` returns `MaxLength`/`MinValue`/`MaxValue` in one request,
no `@odata.type` cast. Replaced the raw `MoneyAttributeMetadata` POST
accordingly.
- dv-data: SDK *does* support bulk delete (`records.delete(table,
[ids])` -> BulkDelete) and batching (`client.batch`) -- removed the
false "unsupported" line, added a bulk-delete example. (Delete is the #1
urllib op: 110,958/2d.)
- dv-overview: "Web API for `$apply` aggregation" -> SDK `fetchxml()`;
auth line reframes `get_plugin_headers` as the fallback.
- dv-query (web-api-advanced): "SDK does not support $apply" -> leads
with SDK FetchXML; corrected the N:N comment to `builder().expand()` /
`records.list(expand=)`.
**Version gate -- require the SDK that backs the above:** dv-connect
preflight bumped `>=1.0.0` -> `>=1.1.0`; dv-data status `1.0.0` ->
`1.1.0`.
**Nudge (`scripts/auth.py`):** `get_plugin_headers` docstring reframed
as the raw **FALLBACK** (names `get_client()` as default; lists only
genuine gaps) + a one-time, non-fatal stderr hint (opt-out via
`DATAVERSE_SUPPRESS_FALLBACK_HINT=1`).
**Evals:** EVAL-AUTH-02 broadened to accept accurate fallback framing
(so authors needn't write false "SDK does not support" claims -- the
forcing that produced the stale claims); EVAL-AUTH-03 (new) guards the
fallback docstring against regression.
## Dependency
The dv-metadata examples require **SDK >= 1.1.0** (PR #206, released
2026-10-07). The plugin installs `PowerPlatform-Dataverse-Client` via
`pip install --upgrade` and the dv-connect preflight now enforces
`>=1.1.0`.
## Testing
- `static_checks` PASSED; `version_bump_check` PASSED (patch);
`test_auth` 59 tests OK (5 new nudge tests: hint fires once,
suppressible, call never broken, docstring framed).
- Behavioral A/B (300-record bulk delete): the bulk-delete fix turned a
**300-call single-delete loop** into **one BulkDelete**.
- Every SDK claim in the diff was verified against the v1.1.0 SDK source
(`records.delete` bulk, `client.batch`, `query.fetchxml`,
`builder().expand`, `records.list(expand=)`, `add_columns` dict spec,
`update_column`/`update_columns`, `list_columns`/`get_column` typed).
## Version
Patch 1.16.0 -> 1.16.1.
Requires 2 maintainer approvals
(`@microsoft/dataverse-skills-maintainers`).