fix: Hindi/Hinglish fraud-phrase matching was word-order dependent

feat: add copilot (RAG chat) endpoint alongside apk-scan and smishing

fix: bake multilingual smishing model into Docker image to fix deploy timeout

feat: migrate multilingual smishing classifier service and model to apk-scanner

feat: add SHAP explainability and APK DNA fingerprinting

sync: update apk intelligence, identity, risk fusion logic, and YARA rules from backend

Align APK intelligence models and engines with backend

Add Reverse Engineering Intelligence engine (capability findings, cert/network IOCs)

Retrain APK classifier on real malware corpus and remove MobSF integration

feat: enable RandomForest ML classifier for APK scoring

Revert Androguard log level suppression to allow detailed logs

fix(logging): PYTHONUNBUFFERED=1 + uvicorn --log-level info --access-log

deploy: apk-scanner microservice (androguard _safe_call fix)

Fix imports: optional ApkRepository, add risk/fusion

Initial APK scanner microservice deployment

initial commit