sate_ai 0.11.0

A fault injection framework for testing on-device AI models in Flutter. Simulate memory pressure, malformed inputs, and degradation to catch failures before deployment.

Changelog

All notable changes to SATE AI are documented here.

The format is based on Keep a Changelog and this project adheres to Semantic Versioning.

Unreleased

0.11.0 - 2026-10-03

Added

• Multi-language support for reports (Issue #95)
  • ReportLanguage enum supporting English (en), Spanish (es), French (fr), German (de), and Portuguese (pt)
  • I18n class and ReportLocalizer for zero-dependency localized report rendering
  • toMarkdownLocalized() and toHtmlLocalized() methods on StressReport
  • CLI flag --language to generate reports in target languages
  • 10+ unit tests covering all supported languages
• Webhook notifications for stress test results (Issue #94)
  • WebhookNotifier supporting Slack, Discord, and Microsoft Teams
  • WebhookPayload builders for each provider
  • CLI flags --webhook-url, --webhook-type, --webhook-on-pass
  • 10+ unit tests
• Detailed model performance profiling (Issue #93)
  • ProfileResult with per-stage timing and memory
  • InferenceProfiler for staged profiling
  • CLI flags --profile and --profile-runs
  • 10+ unit tests

Added

• CLI code generation for custom injectors (sate_ai create injector <name>)
• Stress test scheduling via cron expressions
• Golden baseline comparisons with tolerance thresholds
• Batch mode for running stress tests on multiple models
• Auto-detect model type from file extension in batch mode
• Report comparison and diff view with HTML output
• Performance benchmarking mode (p50, p90, p99 percentiles)
• Stress test retry and flaky test detection
• HTML report export with Chart.js charts
• Real-time SSE monitoring dashboard
• MediaPipe adapter for on-device vision tasks
• Core ML adapter for iOS (with simulation mode)
• Google ML Kit adapter
• CI/CD SVG status badges generator
• Custom report templates in JSON and YAML
• SQLite storage for historical report tracking
• Model health check API (SateAI.healthCheck)
• VS Code extension for editor integration
• Comprehensive documentation rewrite (README, CONTRIBUTING)
• Website redesign with feature guides
• 8 in-depth feature guides in docs/guides/

Changed

• README.md rewritten to reflect all new features
• CONTRIBUTING.md rewritten with current tooling and workflows
• pubspec.yaml description and topics updated
• example/README.md refreshed
• docs/paper.html updated with current status
• .github/ISSUE_TEMPLATE/* verified and updated
• Adapters table expanded to 7 adapters
• Fault injectors table expanded to 11 injectors

Fixed

• Various CI fixes for Android SDK setup and NDK installation
• Manifest merger conflicts between example app and fllama plugin

0.10.0 - 2026-09-01

Added

• Real-time monitoring dashboard (Issue #33)
• HTML report export with Chart.js (Issue #32)
• MediaPipe adapter (Issue #29)
• Core ML adapter (Issue #30)
• Google ML Kit adapter (Issue #31)

0.9.0 - 2026-08-21

Added

• TensorFlow Lite adapter (Issue #4)
• Confidence threshold injector (Issue #6)
• Data corruption injector (Issue #26)
• Model version mismatch injector (Issue #27)
• Network latency/drop injector (Issue #25)

0.8.0 - 2026-08-12

Added

• Fllama (llama.cpp) adapter (Issue #28)

0.7.0 - 2026-08-05

Added

• Latency injector
• Model swap injector
• Interactive demo on website

0.6.0 - 2026-07-30

Added

• Thermal throttle injector (Issue #5)

0.5.0 - 2026-07-29

Added

• CLI executable sate_ai
• GitHub composite action

0.4.0 - 2026-07-28

Added

• Quantization drift injector (Issue #2)

0.3.0 - 2026-07-28

Added

• Web dashboard with dark mode and exports

0.2.0 - 2026-07-28

Added

• ONNX Runtime adapter (Issue #1)

0.1.0 - 2026-07-28

Added

• Initial release: core framework, MockAdapter, MemoryPressureInjector, MalformedInputInjector
• 59 unit tests