## Summary
- replace the former three-provider model with two pre-login choices:
- **ABox**: authenticate through the vendor ABSDK on the current local network
- **Local experience**: establish a production simulation session without credentials, network access, or ABSDK calls
- remove the compatible-backend product/server scope and safely ignore legacy persisted compatible-provider values
- bind one process-wide `AbsdkBridge` for ABox session, socket, temperature/humidity, and infrared operations
- add production `:runtime:simulation` implementations and route device operations by the active provider
- retain provider-scoped username preferences, Keystore-encrypted automatic-login credentials, token recovery, logout, and session isolation
- add regression coverage for authentication gating, profile/session alignment, provider switching cleanup, routing, and simulation behavior
- align requirement, plan, architecture, and protocol documentation with LAN ABSDK access
## Verification
```text
./gradlew check testDebugUnitTest test
BUILD SUCCESSFUL (220 actionable tasks)
./gradlew :android:app:assembleDebug :android:app:assembleRelease
BUILD SUCCESSFUL (217 actionable tasks)
./gradlew :android:app:compileDebugAndroidTestKotlin
BUILD SUCCESSFUL
git diff --check
passed
```
Emulator smoke validation:
- provider selection displays ABox and Local experience only
- Local experience enters the authenticated route without credentials
- Local experience also enters while emulator Wi-Fi/data are disabled
- Local Sign out and Switch provider return to provider selection
- ABox opens the credential form; empty submission stays on login with the required-fields error
## Known constraints and accepted risks
- The vendor JAR uses cleartext HTTP and hardcodes `a-box.com.cn`; Debug and Release intentionally share the same ABox capability under the approved no-Release-exception decision. `@CleartextAbsdkTransport`, `usesCleartextTraffic`, and the architecture/protocol documents explicitly record this risk. This does not provide confidentiality or integrity.
- Real authenticated ABox and read-only hardware smoke validation is deferred because the gateway is currently unavailable locally. This PR does not claim that current-environment validation completed.
- The mapping from the vendor JAR's hardcoded hostname to the LAN gateway remains an external network/deployment prerequisite. The ABSDK JAR and bridge path remain identical to the previously hardware-validated `feat/moler-infrared-control` branch.
- `FakeDeviceCatalog` remains a temporary debug-only placeholder until the device-catalog module provides the production ABox catalog. ABox catalog requests are not routed to Local experience simulation.
Reviewed-on: #4
Co-authored-by: KremeCN <3265494849@qq.com>
## Summary
- add a pure Kotlin/JVM anti-corruption layer for ABSDK 0630
- serialize blocking SDK calls and enforce one provider-scoped session
- map `ABRet`, raw maps, SDK exceptions, and error codes into typed domain results
- implement session, socket, temperature/humidity, and infrared gateways
- document bytecode-confirmed protocol facts, SDK limitations, and unresolved behavior
- align JVM module bytecode with the Android Java 17 target
## Scope boundaries
- the bridge is not wired into the Android app in this PR; debug continues using Fakes
- `setHostInfo` is intentionally not called or implemented as a provider flow
- logout clears only the bridge-local session because the JAR exposes no logout or token setter
- no real credentials, network calls, or hardware were used in tests
- `docs/requirement_spec.md` and `docs/project_plan.md` are unchanged
## Verification
```text
gradlew.bat --stop
gradlew.bat clean :absdk-bridge:test :domain:test :testing:fakes:test :android:app:lintDebug :android:app:assembleDebug :android:app:assembleRelease build
```
Result: `BUILD SUCCESSFUL`
The public JAR API was also rechecked with `javap`, including the actual `getThStatus(String)` method, absence of a public logout API, `ABRet.getDicDatas()`, `IrInfo.getStatus()`, and `downloadIrCodes(List<IrInfo>)`.
Reviewed-on: #3
Co-authored-by: KremeCN <3265494849@qq.com>
KremeCN
Moler