Changelog

​

Version 0.2.8

​

Improvements

• Reduced PDF ingestion memory usage with optimized docling configuration
• Improved subprocess handling on Windows with better event loop compatibility
• Enhanced ChromaDB provisioner reliability and dataset deletion safety

​

Bug fixes

• Fixed subprocess creation issues on Windows SelectorEventLoop
• Fixed registry race conditions during dataset operations
• Resolved asyncio subprocess compatibility issues

​

Version 0.2.7

​

Improvements

• Added file logging with rotation and configurable log levels
• Enhanced exception logging for 500 errors with full tracebacks
• Improved error diagnostics for troubleshooting

​

Bug fixes

• Fixed ChromaDB provisioner to use cross-platform os.kill instead of Unix-only commands
• Fixed ChromaDB client host configuration to point to localhost
• Fixed model ID type handling in ChatContext schema
• Fixed backend console window visibility on Windows desktop app
• Added Windows/Linux Tauri origin to CORS allowed origins
• Disabled strip and UPX on all platforms to prevent .so file corruption

​

Version 0.2.6

​

Improvements

• Enhanced ChromaDB provisioner error logging for debugging import failures

​

Bug fixes

• Fixed python312.dll load failure by disabling UPX and strip on Windows
• Resolved desktop app bundling issues on Windows

​

Version 0.2.5

​

Features

• Added configurable search filters for remote Weaviate datasets
• Implemented system tray functionality for desktop app
• Added “About” window with version information

​

Improvements

• Updated branding with new OpenMined logos and icons
• Simplified dataset health checks by separating type and provisioner validation
• Enhanced remote Weaviate filter support with choice of different data types
• Reduced Linux bundle size from 7.5GB to ~200MB by using CPU-only PyTorch
• Improved graceful shutdown handling

​

Bug fixes

• Fixed ChromaDB health check to use localhost instead of DOCKER_NETWORK_HOST
• Fixed Docker health check URL configuration
• Fixed process-wick path in build script
• Resolved AppImage bundling issues on GitHub Actions Ubuntu runners
• Fixed dataset deletion cleanup and removed dead code
• Added ORM cascade delete for dataset and model endpoints
• Fixed dataset ID type handling as UUID in context classes

​

Version 0.2.0

​

Features

• Enhanced dataset handling and context management
• Integrated dataset type partition deletion on dataset removal
• Improved data source browsing and management

​

Improvements

• Refactored dataset deletion to properly clean up all resources
• Better integration of dataset types with core system

​

Version 0.1.0

​

Initial release

• FastAPI backend with RAG endpoint support
• Vue 3 frontend with modern UI
• Dataset management with ChromaDB and Weaviate support
• Model integration with OpenAI-compatible providers
• Policy system for rate limiting and access control
• File ingestion with automatic indexing
• SyftHub marketplace integration
• Docker container auto-provisioning
• Desktop app for macOS and Linux

​

Upgrade guide

​

From 0.2.x to 0.2.8

docker pull ghcr.io/openmined/syft-space:latest
docker stop syft-space
docker rm syft-space
# Run with your previous configuration
docker run -d --name syft-space ...

git pull origin main
./run.sh

​

From 0.1.x to 0.2.x

# Copy your database file
cp ~/.syft-space/app.db ~/.syft-space/app.db.backup

​

Deprecations

​

Known issues

​

Version 0.2.8

• Large PDF files (>100MB) may cause high memory usage during ingestion
• Windows desktop app may show console window briefly on startup (cosmetic only)

​

Version 0.2.7

• No known issues

​

Contributing

​

Release notes