[Developer] #657 모델 파일명 .litertlm fix + 레거시 마이그레이션 + 에뮬레이터 dev 인프라

- ModelConfig.filename 기본값 gemma4_e2b_q4.bin → gemma-4-E2B-it.litertlm (#342 근본 원인)
- _migrateLegacyPath: 기존 .bin 설치본 rename + meta 갱신 (재다운로드 2.4GB 방지, 멱등)
- maxTokens 2048 유지 + 1024 회귀 방지 주석 (KV cache < prefill signature 시 추론 실패)
- LLM_BACKEND dart-define: 에뮬레이터 CPU 강제 (SwiftShader GPU 네이티브 SIGABRT 회피)
- GEMMA_MODEL_URL dart-define: 호스트 로컬 모델 서버 주입 (기본값 = HF URL 불변)
- debug 전용 cleartext manifest (10.0.2.2 모델 서버용, release 불변)
- 마이그레이션 테스트 4건 신규 (AC-2/3/4). 171 passed, analyze clean

2026-07-14 에뮬레이터 E2E 검증: 다운로드→로드→tool call 왕복 전 구간 성공.

Refs #657, #342

Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>

app/lib/data/ai/gemma_llm_service.dart

@@ -13,6 +13,20 @@ import 'llm_service.dart';
 /// local file path generally does not require the token.
 const String _hfToken = String.fromEnvironment('HF_TOKEN', defaultValue: '');
 
+/// Backend override for dev builds. The Android emulator advertises a
+/// software Vulkan adapter (SwiftShader) that SIGABRTs LiteRT-LM's GPU
+/// init natively — the Dart-level gpu→cpu fallback can't catch a native
+/// abort, so emulator runs must force CPU: `--dart-define=LLM_BACKEND=cpu`.
+/// Unset (default) keeps the SDK's gpu→cpu fallback for real devices.
+const String _backendOverride =
+    String.fromEnvironment('LLM_BACKEND', defaultValue: '');
+
+PreferredBackend? get _preferredBackend => switch (_backendOverride) {
+      'cpu' => PreferredBackend.cpu,
+      'gpu' => PreferredBackend.gpu,
+      _ => null,
+    };
+
 /// One-shot guard so [FlutterGemma.initialize] runs at most once per
 /// isolate. Re-init is unsupported by the underlying plugin.
 bool _initialized = false;
@@ -70,7 +84,15 @@ class GemmaLlmService implements LlmService {
       modelType: ModelType.gemma4,
       fileType: ModelFileType.litertlm,
     ).fromFile(modelPath).install();
-    final model = await FlutterGemma.getActiveModel(maxTokens: 2048);
+    // #342 root cause was the model *filename* (.bin — LiteRT-LM rejects it;
+    // must be .litertlm), NOT the KV cache size. maxTokens stays 2048: the
+    // Gemma 4 E2B compiled graph requires a cache ≥ its prefill signature —
+    // 1024 fails tensor allocation (DYNAMIC_UPDATE_SLICE prepare, verified
+    // on-emulator 2026-07-09).
+    final model = await FlutterGemma.getActiveModel(
+      maxTokens: 2048,
+      preferredBackend: _preferredBackend,
+    );
     _model = model;
     _loaded = true;
   }