Testing Skill

Use this skill when the user wants to run tests, benchmarks, or verify functionality on Android devices or locally.

Capabilities

1. Android Run (Automated)

Use the helper script to Build -> Push -> Run in one go.

Command:

./.agent/skills/testing/scripts/run_android.sh <binary_name> [args...]

Examples:

• •

  Run Inference:

  bash

  ./.agent/skills/testing/scripts/run_android.sh generate \
      --model-path /data/local/tmp/llm_rs2/models/llama3.2-1b \
      --prompt "Hello" \
      -n 128
  

• •

  Verify Backend:

  bash

  ./.agent/skills/testing/scripts/run_android.sh test_backend
  

3. Stress & Stability Testing

Simulate real-world background usage (e.g., user switching apps while model loads) to ensure stability.

Use stress_test_adb.py: This script runs your command on-device while simultaneously switching foreground apps (Settings <-> Home) via ADB.

python3 ./.agent/skills/testing/scripts/stress_test_adb.py \
    --cmd "/data/local/tmp/generate --model-path ... -n 256" \
    --duration 60 \
    --switch-interval 5

Example: YouTube Background Play: Runs the inference while playing a YouTube video (simulates foreground multimedia usage).

python3 ./.agent/skills/testing/scripts/stress_test_adb.py \
    --cmd "/data/local/tmp/generate --model-path ... -n 128" \
    --background-apps "http://youtube.com/watch?v=dQw4w9WgXcQ" \
    --switch-interval 20

2. Local Unit Tests

Run standard Rust unit tests on the host machine.

• •Command: cargo test
• •For specific backend: cargo test --bin test_backend

Common Issues

• •Device Not Found: Check adb devices.
• •Permission Denied: The script automatically runs chmod +x, but ensure the device is accessible.