Prepare release for verison 3.0.0-beta.22 (#2981)

* Updated changelogs ahead of release

* Potential fix for pull request finding

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

* Applied PR feedback

---------

Co-authored-by: Copilot Autofix powered by AI <175728472+Copilot@users.noreply.github.com>

Update create auth npmrc (#2925)

Migrate AppConfig tools to new tool design (#2871)

* Migrate AppConfig tools to new tool design

* Add CHANGELOG entry

Advisor: Add recommendation summary command and list filters (#2732)

* Add Advisor recommendation list filters and summary command

* Docs and changelog for Advisor recommendation filters and summary

* Register advisor_recommendation_summary in consolidated-tools.json

* Server-side KQL summarize; add --top to list; address PR review feedback

- Move summary aggregation into Resource Graph KQL (replaces client-side RecommendationAggregator)
- Add --top option to 'advisor recommendation list' (clamped 1-100)
- Rename 'recommendation' group-by value to 'recommendation-type'
- Drop 'resource' group-by value
- Remove redundant tolower() in KQL contains filters (contains is case-insensitive)
- Refine summary command Title/Description and --group-by description
- Update changelog, command docs, and e2e test prompts

* Address mentor feedback: full counts, Unknown-last, NL search routing

- Summary: drop server-side --top cap and AreResultsTruncated flag; TotalRecommendations now always reflects the full filtered population.
- Summary KQL: order by iff(key=='Unknown',1,0) asc, count_ desc, key asc so Unknown buckets sink to the tail regardless of count.
- Summary command: re-add --top as a post-aggregation display cap that slices Groups; Unknown is always preserved at the tail across slicing so uncategorized items remain visible.
- Tool descriptions hardened to steer the LLM: 'list' is for item retrieval only (capped at 100), 'summary' is the correct tool for any 'how many', 'top N', 'breakdown', 'distribution', or ranking question; LLM is told to render Unknown as a regular row, not a footnote.
- Search option description expanded with natural-language examples (e.g. 'related to Microsoft Foundry' -> --search Foundry) so the LLM extracts topical nouns instead of bypassing the filter.
- Updated unit tests and changelog accordingly.

* Fix trailing whitespace in RecommendationListCommandTests

* Address PR review: validate --group-by via Command.Validators

Move the group-by allowed-value check out of ExecuteAsync and into a System.CommandLine validator registered in RegisterOptions. Errors flow through BaseCommand.Validate and become BadRequest responses automatically. Preserves case-insensitive + trim-tolerant behavior covered by existing tests.

* Make advisor summary --group-by optional (default category); filter to active 'New' recommendations only

Prepare for 3.0.0-beta.19 release (#2903)

Prepare for 3.0.0-beta.19 release (#2903)

Prepare for Fabric MCP Server 1.1.0 release (#2878)

Bump version to 1.1.0, compile changelog entries into the 1.1.0 section, and sync the VS Code extension changelog.

Prepare 3.0.0-beta.18 release (#2867)

prepare for 3.18 beta release

Update a few tool areas to use single RecordedCommandTestsBase implem…

…entation (#2854)

* Move Communication to single recorded test class

* Move Marketplace to single recorded test class

* Move AKS to single recorded test class

* Move Functions to single recorded test class

* Move AppService to single recorded test class

* Fix linting

* Minor feedback

Replace NodeTool with UseNode, update to Node 24 (#2856)

* Replace NodeTool with UseNode, update to Node 24

* Prevent changes to eng/common in PRs