Releases: mdrideout/junjo
0.45.0 - Graphviz Visualizations
- Junjo workflow graphs can now be visualized directly with graphviz
- Every graph object can execute
to_dot_notation()to generate dot notation output of the graph (compatible with any graphviz renderer) - Every graph object can execute
export_graphviz_assets()to directly generate and export graphviz rendered assets (.svg files of the graphs / subflow graphs, .dot notation files, and a .html gallery file to display them)
Full Changelog: 0.44.1...0.45.0
0.44.1 - Documentation and Example Updates
- Improved documentation to discuss using Junjo for Eval Driven Development: https://python-api.junjo.ai/
- Added a pytest based evaluator to demonstrate eval-driven development with a Junjo node: code folder
Full Changelog: 0.44.0...0.44.1
0.44.0 - State Calculation Fixes
Fixes issues with calculating the accumulated state changes for any given telemetry snapshot. Only minor changes required to this end of the library.
Full Changelog: 0.43.1...0.44.0
0.43.1 - Small Library Fixes, Documentation Improvements
0.42 - Examples Improvement
- Improved module organization at the library level allows for cleaner imports at consumption
- Base example improved to align with latest implementation (incomplete, but working)
0.41 - PyPI Release
- Minor changes to pyproject.toml for release on PyPI at https://pypi.org/project/junjo/
0.40 - Alpha Preview
This is the first public release from Junjo that is stable enough for third party evaluation. This release that works well E2E with the AI Chat Example application.
Developers can test this release in their own Python applications for providing feedback.
Last ReactFlow Commit
Switching to MermaidJS for future graph needs due to improved diagram types, easier code gen, and better auto-layout capabilities.
0.20 - Last Custom gRPC Protocol Version
This is the last commit that contains custom gRPC profobuf schema. I am moving everything to utilize OpenTelemetry spans, attributes, and events.
AI Demo Chat working well at this release.
AI Dating Chat - Running
- The AI Dating Chat example is working E2E to create new dating profiles, and send / respond to messages.
- There have been no workflows setup yet, but the limitations of single LLM calls are already a major limitation.
- This app is ready for co-development with Junjo to demonstrate improvements in creating, testing, and debugging dynamic AI graph workflows.
- The readme's for the base and ai_chat examples are working correctly.