Fix OpenClaw 10.2 installation and unify live model selection

[Suzhen Zheng (benniao85)]

Description

Fix OpenClaw 2026.10.2 installation compatibility and unify model configuration and selection across MicroClaw. Provider forms, default-model selection, and conversation-model selection share authenticated live catalogs instead of mixing static presets with saved-model subsets. Model-only changes no longer unconditionally restart the Gateway.

Includes the earlier installation fixes and is merged with main at f605761, retaining both model-startup assertions and upstream MXC lifecycle guards.

Related Issue

Related to #101 (closed): addresses remaining hardcoded model-list behavior and consistency across the three model selectors, rather than expanding the provider-family menu.

Changes

• Support OpenClaw 10.2 SQLite workers and captured local-model provider modules while retaining pinned identity/source checks and scoped sandbox access.
• Pin the compatible search plugin, preserve quoted Node paths, validate actual sandbox access, and avoid unnecessary database migration during same-version updates.
• Share bounded, authenticated, read-only provider discovery across all three model selectors. Isolate draft credentials and surface discovery errors without static fallback lists.
• Register newly selected models with revision-guarded Gateway updates and verify application before switching, preserving credentials, policies, defaults and conversation overrides.
• Keep known non-chat model families visible but disabled with explanations and enforce restrictions in the main process. Distinguish access rejection from invalid credentials and incompatible requests.
• Simplify provider cards with chevrons, adjacent API-key/refresh actions, key-before-model field order, no duplicate saved-model rows or headings, and independently persisted form selections.
• Use standard button activation and accessible, named real Element Plus popovers. Add opt-in non-activating startup/update behavior and distinguish installer captions from the desktop.
• Add regression and opt-in real-provider E2E tests and update English/Chinese documentation.

Type of Change

• Bug fix (non-breaking change that fixes an issue)
• New feature (non-breaking change that adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• CI/CD or build configuration change
• Refactoring (no functional changes)

Testing

• Tests pass locally (cd desktop && npm run test)
• Renderer tests pass locally (cd desktop/renderer && npm run test)
• Build succeeds (cd desktop && npm run build)
• Manual testing performed (describe below)

Validation after merging main:

Command	Result
cd desktop; npm test -- --reporter=dot	Passed: 1,614 Vitest tests; 14 environment-dependent skips; 5 WeChat staging tests passed
cd desktop\renderer; npm run test -- --reporter=dot	Passed: 535 tests
cd desktop; npm run build	Passed: TypeScript, Vue type checks and Vite build
python -m unittest discover -s tests -p test_windows_setup_upgrade.py -q	Passed: 132 tests; files unchanged by the main merge
python -m unittest discover -s tests -p test_webview_bridge.py -q	Passed: 37 tests; files unchanged by the main merge
git diff --check / git diff --cached --check	Passed

Existing Vite large-chunk warnings and renderer shallow-stub warnings remain.

Additional feature-build validation on 0e89412, before merging main:

• powershell -ExecutionPolicy Bypass -File .\build.ps1 — passed; generated desktop, portable ZIP, installer ZIP and full Setup EXE.
• cd desktop; npx vitest run src\provider-model-registration.integration.test.ts --reporter=verbose, with MICROCLAW_TEST_OPENCLAW_ENTRY set — passed against isolated OpenClaw 10.2; no inference or Gateway restart during registration/selection.
• cd desktop; npx vitest run src\provider-model-live.integration.test.ts --reporter=verbose, with MICROCLAW_TEST_OPENCLAW_ENTRY and MICROCLAW_TEST_LIVE_QWEN_CONFIG explicitly set — passed with real Qwen Max/Flash replies, preserved default/session scope, unchanged source configuration and no restart. This optional test may consume quota and is skipped by default.
• Real Windows transactional installation committed; installed archive matched the package; HTTP health and authenticated configuration RPC checks passed.
• Actual installed-client E2E covered matching 176-ID catalogs, Max/Flash replies, access-denied recovery, incompatible-model rejection, credential gating, refresh states, model-only Save, reopening and persistence across restart.
• Verified default changes preserve existing session overrides; restored original default/form preferences, retained AppContainer, and did not change billing settings.
• Verified one desktop main process, non-activating update/startup and cleanup of debugging listeners. Final background verification used app-local actions without keyboard/mouse takeover.

These checks do not certify every returned model's account entitlement or all input/tool capabilities. MiniMax inference was not exercised without an authorized MiniMax credential. Local installer artifacts are unsigned.

Checklist

• My code follows the coding guidelines of this project
• I have performed a self-review of my code
• I have commented my code where necessary
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or my feature works
• New and existing unit tests pass locally with my changes