[Paul Lizer (paullizer)]

Summary

• Operators should not need vendor-specific URL suffixes to add Microsoft models. One native Foundry resource/project connection now resolves the selected OpenAI, Grok, MAI, or Anthropic model to its supported inference API without changing the saved discovery endpoint.
• Built-in model selections preconfigure canonical model identity and native underlying model names. Automatic exact name matching remains available; deployment names, Custom gateway aliases, explicit overrides, publication choices, and global defaults stay intact.
• Add the verified MAI-Thinking-1 preview profile, native audience/key handling, generation limits, streaming and function-tool compatibility. A shared Responses preference applies only where the native model contract supports it; MAI keeps Chat Completions automatically.
• Carry the effective per-model API through sync clients, SK/local agents, workflow/capture metadata, and tabular budgets. Automatic URL routing is restricted to native public Foundry authorities; Custom/APIM URLs and authentication remain authoritative. Update application version to 0.261.410 and document the configuration.

Linked issue

N/A. The user opted to continue without an associated issue.

Release Notes & Latest Features

• New Feature
• Bug Fix
• UI Enhancement
• Breaking Change
• Internal only

Is this visible to end users?

• Yes
• No

Is this admin-facing (Admin Settings, governance, deployment, config)?

• Yes
• No

Should this become a Latest Feature card?

• Yes
• No
• Already added

Screenshot needed for the card?

• Yes
• No
• Attached

Version bump

• application/single_app/config.py VERSION third segment bumped, or not needed because this is docs-only
• deployers/version.txt bumped, or not needed because deployers/ was not changed

Application: latest base 0.261.409 -> 0.261.410. The initial feature draft was 0.261.407; the pending release and its documentation/test headers are now aligned with the latest base. No deployer files changed.

Testing / validation

• Passed after merging the latest base through b0a358646: 268 tests and 755 subtests. The following targeted suites include real SDK request construction, mixed-model native connections, streaming/tool continuation, normal and optimized Python, capture/workflow metadata, and tabular budgets. All inference HTTP is mocked; no production deployment or paid inference was performed.

python -m pytest -q --disable-warnings .\functional_tests\test_microsoft_mai_model_connections.py .\functional_tests\test_model_inference_api_versions.py .\functional_tests\test_model_responses_selection.py .\functional_tests\test_model_catalog_token_evidence.py .\functional_tests\test_model_catalog_profiles.py .\functional_tests\test_model_capability_catalog_resolution.py .\functional_tests\test_model_endpoint_protocol_inference.py .\functional_tests\test_model_endpoint_provider_registry.py .\functional_tests\test_custom_model_endpoint_diagnostics.py .\functional_tests\test_responses_model_tool_execution.py

• python -m pytest -q --disable-warnings .\functional_tests\test_tabular_model_token_budgets.py .\functional_tests\test_excel_native_inspection_chat_evidence.py .\functional_tests\test_tabular_table_aware_analytics_replay.py - passed, 63 tests, covering the overlapping base Excel/table-aware changes and model budgets.
• node .\functional_tests\test_v2_model_connections_logic.mjs - passed, 43 checks.
• python -m pytest -q --disable-warnings .\functional_tests\test_v2_model_catalog_workbench.py - passed, 3 tests, including 12 bundled TypeScript checks. The browser-origin fixture was completed for the existing API client.
• npm --prefix .\application\v2_ui run typecheck -- --pretty false - passed.
• npm --prefix .\application\v2_ui run build - passed; existing large-bundle warning remains.
• python .\functional_tests\route_tests\test_route_blueprint_policy_inventory.py - passed, 15 checks.
• python .\functional_tests\route_tests\test_route_unauthenticated_policy_contract.py - passed, 10 checks.
• python .\functional_tests\route_tests\test_route_policy_test_coverage.py - passed, 3 checks.
• python .\scripts\build_docs_inventory.py - passed before the base merge, regenerated inventory had no content changes.
• python .\functional_tests\test_docs_app_surface_coverage.py - passed after the base merge, 7 checks.
• python .\functional_tests\test_docs_site_quality.py - passed after the base merge, 6 checks.
• python -m pytest --collect-only -q .\ui_tests\test_microsoft_mai_ai_connections.py - 4 tests collected before the base merge. Browser execution was intentionally not performed; desktop/mobile native saves, automatic matching/presets, and Custom alias preservation are covered by the authored fixtures.
• python .\scripts\check_xss_sinks.py --full-file .\application\v2_ui\src\components\admin\ModelConnectionsManager.tsx .\application\v2_ui\src\components\admin\ModelCatalogManager.tsx .\application\v2_ui\src\lib\modelConnections.ts .\application\v2_ui\src\lib\customModelConnections.ts .\application\v2_ui\src\lib\modelInference.ts .\application\v2_ui\src\lib\modelCatalog.ts - passed before the base merge, 6 files; those feature implementations were unchanged by conflict resolution.
• git diff --check and git diff --cached --check - passed.

Historical baseline limitations: Before the latest-base integration, additional legacy reasoning/staging fixtures had 42 failures from missing extracted AST dependencies (apply_model_profile, should_include_current_turn_images) and an outdated staging callback. The same failures were reproduced against pre-feature commit 98e01345f. The base update repairs the staging fixture, and all current tabular model-budget tests now pass in the 63-test overlap suite above. Older unrelated reasoning/Custom/image/embedding fixtures are not represented as passing checks or independently repaired here.

Live model access, RBAC, regional availability, quotas, and gateway readiness remain deployment-specific checks. Custom connections do not acquire the application's managed-identity token automatically, and the optional encrypted MAI reasoning envelope is not requested.

Documentation

• Release notes updated, or not needed
• Feature documentation updated, or not needed
• Fix documentation updated, or not needed

Updated the AI Connections guide, AI Models and Model Catalog references, release notes, and docs/explanation/fixes/MICROSOFT_MAI_MODEL_CONNECTIONS_FIX.md with the automatic native flow and explicit gateway boundary. Conflict resolution preserves the base Excel release entries and both sets of tabular test metadata.

Security checklist

• New Flask routes include @swagger_route(security=get_auth_security())
• Settings sent to non-admin frontends use sanitize_settings_for_user()
• Browser JavaScript is served from local SimpleChat static assets only; no CDN-hosted JS
• No secrets, keys, connection strings, or local-only artifacts are included

No routes or frontend settings-exposure boundaries were added by this feature. Existing authentication/authorization policies are verified by the route-policy checks.