All notable changes to the graph-app-kit are documented in this file.
The changelog format is based on Keep a Changelog. This project adheres to Semantic Versioning and all PyGraphistry-specific breaking changes are explictly noted here.
Related changelogs:
Core:
Extensions:
See projects page and open pull requests
- Update dependency PyGraphistry 0.33.8
- Enable native execution with stronger default env var handling
- Update dependency PyGraphistry 0.33.0
- upgrade from streamlit v1.21.0 to latest --> v1.25.0
- improved the logging behavior and print statements not being written (we recommend using pyton logger instead of print statements for greater functionality and control) issue #104
- fixed unhandled exception in demo_avr when splunk credentials not set issue #105
- fixed incorrect ST_PUBLIC_PORT printed in the logs from entrypoint.sh issue #106
in a previous release two new demo views were added: demo_avr and demo_login which are both cyber security related
- check if upgrade from streamlit 1.21 --> 1.25 might have affected your code
- Infra: CUDA base now 11.5
- Infra: Optimized runner
- Infra: Remove now-unnecessary free space step
- Infra: env var USE_DOCKER=true
- Infra: env var PYTHONPATH=/apps/views
- Infra: env var FAVICON_URL
- Infra: externally overridable src/streamlit/{config.toml,credentials.toml}
- Infra: Update GPU base to 2.39.27
- Infra: Update actions/checkout GHA from v2 (deprecated) to v3
- Infra: Removed autoheal service and labels (likely OK for most users)
- toml: Fixed startup for users of new startup system (master)
- Update dependency PyGraphistry 0.27.1
- Update dependency Streamlit to 1.12.0
- Update RAPIDS-ecosystem packages (2022.04)
- opt-in to minimal CPU-only base image
- alias in
src/docker
:./dc
,./dc.cpu
- test python 3.7, 3.8, 3.9
- test cpu and gpu docker building
- build now uses buildkit
- minimal
- cpu mode
- aliases of
./dc
and./dc.cpu
- Temporarily switch dockerhub build to CPU minimal while too-large base image kinks get worked out
- GraphistrySt: Switched from unsafe mode markdown to component
- PyGraphistry: Updated to 0.21.2
- UI: Switch default to wide mode
- Infra: Upgrade to Graphistry 2.38.10 base
- Infra: Unpinned plotly and updated protobuf, streamlit
- Infra: New
docker-compose.override.yml
symlink mode (WIP)
- Infra: Replaced failing GHA collaborators-only with repo setting
- CI: Graphistry AMI list generator takes VERSION parameter
- Infra: Upgrade to Graphistry 2.36.6
- CI: Graphistry AMI list generator usage
- README: Removed dangling link
- README: Quicklaunch links and admin commands
- Private Streamlit dashboards instance now bound to Jupyter notebook private dashboards folder. Was incorrectly using the public folder: https://github.com/graphistry/graph-app-kit/pull/51/commits/3872de053b7d2888ce271acf395f112491742606
- CI: Publish cloud formation templates to s3 on merge (#48)
- TigerGraph support (#36)
- Changelog
- AMI enumeration script
- Tests: flake8, docker build
- CI: GHA
- CI: Badges
- Versions: Streamlit 0.70 -> 0.77, PyGraphistry 0.14 -> 0.17.2, Graphistry -> 2.35.9 (including AMIs)
- Dev docs: Tagging, building
- Graphistry 2.35+ Support: Swaps in old < 2.34-style Caddy 1.x container as Graphistry 2.35's Caddy 2.x is currently tricky for auth reuse
- Plotter auto-memoizes with
.plot(as_files=True, ...)
- Flake8 warnings
- Default base container now CUDA 11.0