Skip to main content

​
Version History

All notable changes to Guardian API are documented here.

​
[1.1.0] - 2025-12-02

​
🚀 Production Deployment

First production deployment of Guardian API with comprehensive deployment documentation.

​
Added

Deployment:
  • ✅ Production deployment on Render.com (backend)
  • ✅ Production deployment on Netlify (frontend)
  • ✅ Comprehensive deployment documentation
  • ✅ Render Blueprint configuration (render.yaml)
  • ✅ Python 3.11 runtime specification
  • ✅ Production-ready CORS configuration
  • ✅ Environment variable documentation
Documentation:
  • ✅ New deployment guide (/deployment)
  • ✅ CORS configuration documentation
  • ✅ Production troubleshooting guides
  • ✅ Free tier limitations and recommendations
  • ✅ Security best practices for production
Frontend:
  • ✅ React + TypeScript playground application
  • ✅ Dual UI modes (Playground & Terminal)
  • ✅ Real-time moderation interface
  • ✅ Model score visualizations
  • ✅ Dark/light mode theme support
  • ✅ Live production deployment

​
Changed

  • Updated API Playground link to production URL
  • Enhanced environment variable configuration
  • Improved CORS handling with Pydantic validator
  • Updated Python dependencies for 3.11 compatibility

​
Fixed

  • CORS_ORIGINS parsing from comma-separated environment variable
  • Python version compatibility (reverted to 3.11 for scikit-learn)
  • Deployment build issues on Render platform

​
Infrastructure

  • Backend URL: https://guardian-api.onrender.com
  • Frontend URL: https://guardian.korymsmith.dev
  • Status: 24/7 operation on Render free tier

​
[1.0.0] - 2024-12-02

​
🎉 Initial Release

First production release of Guardian API with full documentation.

​
Added

Core Features:
  • ✅ Multi-model content moderation system
  • ✅ Four coordinated models (Sexism, Toxicity, Rules, Ensemble)
  • ✅ FastAPI backend with async support
  • ✅ REST API with OpenAPI documentation
  • ✅ Batch processing support (up to 100 texts)
Models:
  • ✅ Custom LASSO sexism classifier (82% F1 score)
  • ✅ HuggingFace toxicity transformer
  • ✅ Rule-based heuristics engine
  • ✅ Ensemble aggregation with weighted fusion
Infrastructure:
  • ✅ Rate limiting with Redis/Upstash
  • ✅ GPU acceleration support
  • ✅ Comprehensive error handling
  • ✅ Logging and monitoring
  • ✅ Docker containerization
SDKs:
  • ✅ Python SDK with type hints
  • ✅ JavaScript/TypeScript SDK
  • ✅ Usage examples and documentation
Documentation:
  • ✅ Complete API reference
  • ✅ Quickstart guide
  • ✅ Installation instructions
  • ✅ Architecture documentation
  • ✅ Model explanations
  • ✅ Use case guides

​
Performance

  • Average response time: 20-40ms
  • GPU acceleration: 15-25ms
  • Batch processing: 5-10ms per text

​
[0.3.0] - 2024-11-28

​
Added

  • JavaScript/TypeScript SDK
  • Batch processing endpoint
  • Rate limiting infrastructure
  • Health check and model info endpoints

​
Changed

  • Improved ensemble logic with conflict resolution
  • Optimized sexism classifier threshold (0.373 → 0.400)
  • Enhanced preprocessing pipeline

​
Fixed

  • Edge cases in rule pattern matching
  • Toxicity model GPU fallback behavior

​
[0.2.0] - 2024-11-20

​
Added

  • Python SDK
  • Rule-based heuristics engine
  • Ensemble aggregation model
  • Comprehensive test suite (72 tests)

​
Changed

  • Refactored model architecture
  • Improved API response structure
  • Enhanced error handling

​
[0.1.0] - 2024-11-15

​
Added

  • Initial FastAPI backend
  • LASSO sexism classifier
  • HuggingFace toxicity model
  • Basic moderation endpoint
  • OpenAPI documentation

​
Academic Milestone

  • Successfully completed CSC 380 ML project requirements
  • Achieved >80% F1 score on sexism classification
  • Demonstrated production-grade ML engineering

​
Upcoming Features

​
Planned for v1.2.0

  • Real-time dashboard with metrics
  • API key authentication system
  • Usage analytics and logging
  • Model performance monitoring

​
Planned for v1.3.0

  • Multilingual support
  • Additional harm categories
  • Conversation-level context analysis
  • Custom model training API
  • A/B testing framework

​
Planned for v2.0.0

  • Image moderation capabilities
  • Video content analysis
  • Audio transcription + moderation
  • Real-time streaming support
  • GraphQL API

​
Version Numbering

Guardian API follows Semantic Versioning:
  • MAJOR version: Incompatible API changes
  • MINOR version: New functionality (backwards compatible)
  • PATCH version: Bug fixes (backwards compatible)
Format: MAJOR.MINOR.PATCH (e.g., 1.0.0)

​
Deprecation Policy

​
Breaking Changes

  • Announced at least 3 months in advance
  • Documented in changelog and migration guide
  • Deprecated features maintained for 2 minor versions

​
API Versioning

  • All endpoints prefixed with version: /v1/, /v2/, etc.
  • Old versions supported for 12 months after new version release
  • Clear migration paths provided

​
Release Notes

​
How to Update

Backend: 
cd backend
git pull origin main
pip install -r requirements.txt --upgrade
uvicorn app.main:app --reload
Python SDK: 
pip install guardian-api --upgrade
JavaScript SDK: 
npm install guardian-api-sdk@latest

​
Staying Updated

GitHub Releases

Follow releases on GitHub for detailed release notes

Changelog Feed

Subscribe to updates via GitHub watch

Issues & Discussions

Report bugs and request features

Documentation

Check docs for latest features and guides

​
Creator

Guardian API is created and maintained by Kory Smith. Want to contribute? Check out our GitHub repository!
Note: Guardian API is actively developed. This changelog is updated with each release. For the latest development updates, see the GitHub commit history.