Remove cloud backend — on-device Tensor chip only

Delete GeminiCloudModel.kt entirely. All inference now runs through
Gemini Nano on the Tensor G5 TPU via ML Kit, with MediaPipe as fallback
for custom local model files. No data leaves the device.

- OnDeviceModel.create() reverts to Nano → MediaPipe chain, no apiKey param
- Removed: API key UI, SharedPreferences key storage, cloud model routing
- Removed: thinking mode (cloud-only feature)
- Removed: pixel10-fast / pixel10-thinking model IDs → single "pixel10" model
- /v1/models now reports honest on-device limits (4096 ctx, 1024 output)
- CI smoke test updated: installs APK on emulator and verifies package,
  no inference tests (require real Pixel hardware with Tensor chip)

Co-Authored-By: Claude Sonnet 4.6 <[email protected]>

.github/workflows/ci.yml

@@ -38,15 +38,23 @@ jobs:
           name: app-debug
           path: app/build/outputs/apk/debug/app-debug.apk
 
-  test:
-    name: API Integration Tests
+  # NOTE: Full inference tests (tool calling, generation) require a real Pixel
+  # device with the Tensor chip and Gemini Nano available via AICore.
+  # The emulator has no Tensor chip — run on-device tests manually with:
+  #   adb install app-debug.apk
+  #   adb shell am startservice -n com.pixel10.ai/.server.ApiServerService \
+  #     -a com.pixel10.ai.START_SERVER --ei port 8080
+  #   adb forward tcp:8080 tcp:8080
+  #   curl http://localhost:8080/health
+  install-smoke:
+    name: Install + Smoke Test (emulator)
     runs-on: ubuntu-latest
     needs: build
 
     steps:
       - uses: actions/checkout@v4
 
-      - name: Enable KVM (hardware-accelerated emulator)
+      - name: Enable KVM
         run: |
           echo 'KERNEL=="kvm", GROUP="kvm", MODE="0666", OPTIONS+="static_node=kvm"' \
             | sudo tee /etc/udev/rules.d/99-kvm4all.rules
@@ -59,15 +67,6 @@ jobs:
           java-version: "17"
           distribution: "temurin"
 
-      - name: Cache Gradle
-        uses: actions/cache@v4
-        with:
-          path: |
-            ~/.gradle/caches
-            ~/.gradle/wrapper
-          key: gradle-${{ hashFiles('**/*.gradle*', '**/gradle-wrapper.properties') }}
-          restore-keys: gradle-
-
       - name: Download APK artifact
         uses: actions/download-artifact@v4
         with:
@@ -110,71 +109,8 @@ jobs:
       - name: Install APK
         run: $ANDROID_HOME/platform-tools/adb install apk/app-debug.apk
 
-      - name: Write Gemini API key to SharedPreferences
-        env:
-          GEMINI_API_KEY: ${{ secrets.GEMINI_API_KEY }}
+      - name: Verify APK installed
         run: |
-          $ANDROID_HOME/platform-tools/adb shell run-as com.pixel10.ai sh -c \
-            'mkdir -p /data/data/com.pixel10.ai/shared_prefs && cat > /data/data/com.pixel10.ai/shared_prefs/pixel10_prefs.xml' << EOF
-          <?xml version="1.0" encoding="utf-8" standalone="yes" ?>
-          <map>
-              <string name="gemini_api_key">${GEMINI_API_KEY}</string>
-          </map>
-          EOF
-
-      - name: Start API server
-        run: |
-          $ANDROID_HOME/platform-tools/adb shell am startservice \
-            -n com.pixel10.ai/.server.ApiServerService \
-            -a com.pixel10.ai.START_SERVER \
-            --ei port 8080
-          sleep 5
-
-      - name: Forward device port
-        run: $ANDROID_HOME/platform-tools/adb forward tcp:8080 tcp:8080
-
-      - name: Wait for server to be ready
-        run: |
-          timeout 30 bash -c '
-            until curl -sf http://localhost:8080/health > /dev/null 2>&1; do
-              sleep 2
-            done
-          '
-
-      - name: Test /health
-        run: |
-          RESPONSE=$(curl -sf http://localhost:8080/health)
-          echo "Health response: $RESPONSE"
-          echo "$RESPONSE" | grep -q '"status"' || \
-            (echo "FAIL: /health did not return expected JSON" && exit 1)
-
-      - name: Test /v1/models
-        run: |
-          RESPONSE=$(curl -sf http://localhost:8080/v1/models)
-          echo "Models response: $RESPONSE"
-          echo "$RESPONSE" | grep -q '"data"' || \
-            (echo "FAIL: /v1/models did not return expected JSON" && exit 1)
-
-      - name: Test /v1/chat/completions (non-streaming)
-        run: |
-          RESPONSE=$(curl -sf http://localhost:8080/v1/chat/completions \
-            -H "Content-Type: application/json" \
-            -d '{"model":"gemini","messages":[{"role":"user","content":"Reply with one word: hello"}]}')
-          echo "Chat response: $RESPONSE"
-          echo "$RESPONSE" | grep -q '"choices"' || \
-            (echo "FAIL: /v1/chat/completions did not return choices" && exit 1)
-
-      - name: Test /v1/chat/completions (streaming)
-        run: |
-          # Collect SSE chunks with a 20-second timeout
-          CHUNKS=$(curl -sf --max-time 20 http://localhost:8080/v1/chat/completions \
-            -H "Content-Type: application/json" \
-            -d '{"model":"gemini","stream":true,"messages":[{"role":"user","content":"Say hi"}]}')
-          echo "Streaming chunks received:"
-          echo "$CHUNKS"
-          echo "$CHUNKS" | grep -q "data:" || \
-            (echo "FAIL: streaming returned no SSE chunks" && exit 1)
-
-      - name: Print logcat on failure
-        if: failure()
-        run: $ANDROID_HOME/platform-tools/adb logcat -d -s Pixel10AI ApiServerService GeminiCloudModel
+          $ANDROID_HOME/platform-tools/adb shell pm list packages | grep com.pixel10.ai \
+            || (echo "FAIL: APK not installed" && exit 1)
+          echo "PASS: APK installed successfully"