v0.4.3 - load testing and cloud-inference rate limiting.

What's Changed

• Add rate limit to background escalations by @f-wright in #115
• Add load testing scripts by @CoreyEWood in #108
• Include detail about model fetching errors by @CoreyEWood in #116
• Fix inference model updates by @CoreyEWood in #117
• Dealing with kernel issue on raspberry pi with k3s by @robotrapta in #119
• Rewording things so we use "cluster" in the canonical sense of the word by @robotrapta in #118

Full Changelog: v0.4.2...v0.4.3

v0.4.2 - fps improvements

What's Changed

• Add TTL cache for edge inference ready checks by @CoreyEWood in #113
• Reduce cache TTL to avoid getting stuck in failure for too long by @CoreyEWood in #114

Full Changelog: v0.4.1...v0.4.2

v0.4.1 - bug fix and efficiency improvements

What's Changed

• Fix for submit_image_for_inference by @CoreyEWood in #111
• Avoid deserializing image on the edge-endpoint by @tyler-romero in #112

Full Changelog: v0.4.0...v0.4.1

v0.4.0 - remove motion detection, refactor config options

What's Changed

• Fix deployment bug for unconfigured detectors; reset database upon startup by @tyler-romero in #107
• Increase nginx worker connections by @CoreyEWood in #109
• Rip out Motion Detection by @tyler-romero in #110

Full Changelog: v0.3.0...v0.4.0

v0.3.0 - Support for additional prediction modes (count, multiclass, bboxes, text)

What's Changed

• Small fix for filename by @CoreyEWood in #94
• Better logging inference by @robotrapta in #83
• Support all detector modes by @CoreyEWood in #95
• General improvements (pydanticv2, health checks, better testing, cleanup) by @tyler-romero in #98
• Sync public model bucket to edge device upon inference server startup by @tyler-romero in #96
• Remove det ids from edge-config.yaml by @CoreyEWood in #104
• Fix count result parsing by @CoreyEWood in #105
• Edge only inference with cloud training by @f-wright in #101
• Add timestamps to inference-model-updater logs by @CoreyEWood in #106
• Trying to prewarm inference image by @robotrapta in #102

Full Changelog: v0.2.0...v0.3.0

v0.2.0 - FastApi-based Edge-Inference Server

What's Changed

• FastApi based inference-server by @tyler-romero in #93

Full Changelog: v0.1.1...v0.2.0

v0.1.1

What's Changed

• Enable per-detector edge-only mode by @CoreyEWood in #91
• Edge-Endpoint pyproject.toml cleanup and vscode/settings.json by @tyler-romero in #92

Full Changelog: v0.1.0...v0.1.1

v0.1.0

Initial public release of Groundlight's edge-endpoint 🥳 🎊 🎉

The edge-endpoint is under active development, so expect improvements coming soon!

What's Changed

• PK-929 Stub edge API by @michael-groundlight in #1
• CI/CD via Proxy by @blaise-muhirwa in #6
• Set Up Default Linting by @blaise-muhirwa in #9
• Add Detectors Endpoint and YAML Configuration for Upstream Edge Server Endpoint by @blaise-muhirwa in #7
• Integrate Motion Detection by @blaise-muhirwa in #8
• Fix post_image_query Endpoint by @blaise-muhirwa in #11
• Rename Endpoints to Routes and Add Log-Level to Docker Command by @blaise-muhirwa in #12
• Configure Maximum Time Between Cloud-Submitted Images and Add Specific GH Action for Motion Detection by @blaise-muhirwa in #16
• Cache for Image Query ID's Originating from the Edge Endpoint by @blaise-muhirwa in #17
• Updating README by @robotrapta in #13
• Update SDK version by @blaise-muhirwa in #18
• Updating README - build, security, install by @robotrapta in #20
• Fix Race Condition in Motion Detection by @blaise-muhirwa in #21
• Setup K3s on Edge by @blaise-muhirwa in #23
• Update README.md by @blaise-muhirwa in #24
• k3s deployment for inference container by @tyler-romero in #25
• [small] validate request body in post image query by @blaise-muhirwa in #26
• Submit inference request to edge inference container by @tyler-romero in #27
• Support Motion Detection for Multiple Detectors by @blaise-muhirwa in #22
• Use a ConfigMap for edge.yaml inside Kubernetes by @blaise-muhirwa in #30
• Remove edge endpoint-level API Token by @blaise-muhirwa in #33
• Support building image on both arm64 and amd64 by @tyler-romero in #34
• Re-escalate Unconfident Cached Image Query Responses by @blaise-muhirwa in #29
• Optimize Docker Builds by @blaise-muhirwa in #36
• Remove --target production-image requirement while building docker image by @blaise-muhirwa in #37
• Model updating from the edge! by @tyler-romero in #35
• Github action that sets up k3s by @robotrapta in #39
• Fix bug that was causing us to poll for the cloud response on the edge-endpoint by @tyler-romero in #40
• Fix polling behavior by @blaise-muhirwa in #42
• Smarter model fetching - no download if no new model available by @tyler-romero in #41
• Create a separate inference deployment per detector by @blaise-muhirwa in #38
• adding container to update models on a schedule by @blaise-muhirwa in #46
• delete poetry cache after installing dependencies in Dockerfile by @blaise-muhirwa in #47
• add option to build images for dev by @blaise-muhirwa in #49
• Improve uptime of edge inference container + small fixes by @tyler-romero in #50
• README for Jetson install by @robotrapta in #54
• Add want_async query parameter to post_image_query by @blaise-muhirwa in #53
• Add commands for setting up nvidia drivers and nvidia container runtime in install-k3s-g4.sh by @blaise-muhirwa in #55
• update backend triton server image by @blaise-muhirwa in #58
• Fix Triton configuration that was causing CPU memory errors by @blaise-muhirwa in #60
• Add timestamp field to Example on Edge, plus supporting some SDK features, QOL improvements, etc by @tyler-romero in #61
• Setup sqlite database by @blaise-muhirwa in #56
• Enable using EFS Volumes for sqlite database and model repository by @blaise-muhirwa in #62
• edge-inference on k3s on balena! by @tyler-romero in #64
• Set k3s to use mysql as datastore, not etcd by @tyler-romero in #66
• Increased file size limit to 10MB and new build image by @honeytung in #67
• Autolaunch Cluster when starting Edge-Endpoint by @honeytung in #69
• Fixed CI/CD Tests for Edge-Endpoint by @honeytung in #70
• Reviewer lottery for dependabot PRs by @tyler-romero in #71
• Switching edgedevice docker images to public facing AWS account by @f-wright in #72
• Updated SDK and Fixed Pydantic Model Validation Error by @honeytung in #74
• Updated Inference Image by @honeytung in #75
• Add optional flag to run edge endpoint for balena by @f-wright in #73
• Added Detector Group Support and Deployment Instructions for Balena devices by @honeytung in #76
• Fix auto-install of nvidia drivers by @tyler-romero in #79
• Add Support for Edge to use Latest Build Images for Endpoint and Inference by @honeytung in #80
• Fix docker run command in README to work on Ubuntu by @CoreyEWood in #81
• README and setup fixes/improvements by @CoreyEWood in #82
• Logging FPS and shell usability by @robotrapta in #68
• Update README by @CoreyEWood in #84
• Update issue templates by @tyler-romero in #85
• Fixed communication error due to incorrect endpoint by @honeytung in #86
• change parameters to correspond to secondary predictions change by @brandon-groundlight in #88
• Text support bugfix by @brandon-groundlight in #89
• Enabling EDGE_ONLY option by @CoreyEWood in #87
• Update k3s and nvidia install by @tomfaulhaber in #90

New Contributors

• @michael-groundlight made their first contribution in #1
• @robotrapta made their first contribution in #13
• @tyler-romero made their first contribution in #25
• @f-wright made their first contribution in #72
• @CoreyEWood made their first contribution in #81
• @brandon-groundlight made their first contribution in #88
• @tomfaulhaber made their first contribution in #90

Full Changelog: https://github.com/groundlight/edge-endpoint/commits/v0.1.0