* Update vc plugin
* Update tests with breaking change: combined_score now 0-100
* Frame processor predictions are in range of 0-100 with new key
* Remove offline convert score to confidence
* Update calculateConfidence.js
* Common ancestor should use combine_score
* Offline suggestions have combined_scores now instead of scores
* Change filtering to range 0-100
* Only use combined_score for Match screen
* Update debug data
* Fic typo in tests
* Remove key from type
* fix: show observation datetime in the obs time zone
I.e. it doesn't offset the observation datetime into the viewer's time zone.
* test: adjust to literal times by default
* chore: update to date-fns 3.0
* wip: show time zone names with all times
* show time zone name whenever a time zone is passed to a formatting function
* store observation IANA time zone in Realm
Note that this required patching around a bug in Hermes in which it should be
returning a GMT offset for the short time zone but is instead just returning
GMT.
* fix: omit time zone for unuploaded obs
* feat: show relative time differences on ActivityItem headers
* fix: hide zone/offset on ObsEdit before upload when signed in
* fix: hide clock icon in activity item header in new default mode
Also
* stop using checkCamelAndSnakeCase when not necessary in DetailsTab.js
* make POJO types only refer to other POJO types
* Update vision-plugin
* useOfflineSuggestion TS
* Use location for prediction from file
* Remove log
* Use plugin function to lookup location and elevation based on h3 cells
* taxonomyRollupCutoff is now done in the plugin on basis of the tops score for a frame
* Fix wrong commit reference. Now references main branch
* Fix has location check to account for 0.0 locations
* Only prop in lat and long
* Add explanation
* Fix: Import vision-plugin function from wrapper file
* Add jest mock for new plugin function
* Fix#2212 - use physical volume buttons to capture photo
* Fixed tests (mocking VolumeManager)
* Fixed tests (mocking VolumeManager)
* #2212 - physical volume buttons for camera capture - also in AI camera
* #2212 - additional fixes for AI camera
* Update pods
* Update pods
---------
Co-authored-by: Amanda Bullington <albullington@gmail.com>
* Support Sign in with Apple on iOS, w/o design
* Typescript conversions
* Add authenticateUserByAssertion that should be usable w/ Google OAuth-based auth
Closes#583
* Also collect altitude from geolocation calls
* Update geomodel plugin
* Prop user location down into ai camera
* Add geo model file to be linked into xcode project
* Update env.example
* Update env.example
* Add geo model to cv model info file
* Update Developer.js
* Use user location for geo model while frame processing
useGeoModel boolean signals the frame processing plugin to use the geo model as well. It is turned on when we have a location.
* Update README.md
* Geo model is optional
* Adding a dummy neural network called geomodel.placeholder
This model doesn't do anything just returns the input. This is a overhead we could justify in my view as not to spend some more time figuring out how to make a model optional in an Xcode project.
* Geo model files are optional
* Update README.md
* Fix wrong model usage
* Update README.md
* This comment is no longer true
* feat: use full-text index on new _searchableName field (#2487)
Supports searches across non-contiguous tokens, e.g. finding "Mount Diablo
fairy lantern" with a query like "diablo lantern"
Related to #2348
* feat: add pull to refresh on notifications (#2489)
Closes#2349
* feat: show notifications from others (#2491)
* split notifications into tabs
* lots of TypeScript conversion
* feat: resize Heading5 and add Heading6 (closes#2480)
* fix: mark remote observations as viewed from ObsDetails
* feat: show indicator in Notifications tabs if unviewed notifications
Closes#2451
* feat: add option to save all observations from batch obs edit (#2493)
* fix: bring KebabMenu in line with designs
Closes#2268
* Update testing instructions
* v0.55.2
* v0.55.2+129
* fix: show observation from external link when Explore filters open (#2453)
* Open observation in app when link is pressed even if modals are open
* Remove comments
* Add comment and reset obsid
* fix: show onboarding content when signed in user has 0 observations (#2510)
Closes#2508
* fix: load new Explore results when filters change (#2507)
URL params got removed from the query key as a part of pull-to-refresh, but
that's how changes to filters were triggering a new request. This restores
that behavior, adds a test to ensure changes to filters call the API again,
and fixes the prior test of pull to refresh which was not testing the
equivalent behavior when pulling to refresh (API on first load, so the mock
needed to be cleared).
* fix: restore observation fetch when focusing on myobs (#2511)
* fix: center number in photo count (#2512)
Closes#2506
* refactor: minor edits for clarity
* fix: specify CV model version in e2e test
Hoping this deflakes the e2e a bit. Not sure why the AICamera test ever worked
without this.
* v0.55.3
* v0.55.3+130
* New Crowdin translations by GitHub Action (#2481)
Co-authored-by: Crowdin Bot <support+bot@crowdin.com>
* fix: get quality grade descs closer to designs; add Joanne to staff (#2517)
* fix: prevent BottomSheet title from overflowing over close button
* Update to Geomodel name
* Another name change
* Rename var and const names
* Rename const file references
* Also collect altitude from geolocation calls
* Update geomodel plugin
* Prop user location down into ai camera
* Add geo model file to be linked into xcode project
* Update env.example
* Update env.example
* Add geo model to cv model info file
* Update Developer.js
* Use user location for geo model while frame processing
useGeoModel boolean signals the frame processing plugin to use the geo model as well. It is turned on when we have a location.
* Update README.md
* Geo model is optional
* Adding a dummy neural network called geomodel.placeholder
This model doesn't do anything just returns the input. This is a overhead we could justify in my view as not to spend some more time figuring out how to make a model optional in an Xcode project.
* Geo model files are optional
* Update README.md
* Fix wrong model usage
* Update README.md
* This comment is no longer true
* refactor: minor edits for clarity
* Update to Geomodel name
* Another name change
* Rename var and const names
* Rename const file references
---------
Co-authored-by: Ken-ichi <kenichi.ueda@gmail.com>
Co-authored-by: Angie <angie@inaturalist.org>
Co-authored-by: Crowdin Bot <support+bot@crowdin.com>
Main fix was updating inaturalistjs to get a recent bug fix. Also submits a
JSON body instead of a pseudo-formdata-style body.
Also:
* Only try to process queued items that have not failed in the last minute
* Delete queued items after failing 3 times
* Fix index.js error handler bug that was squelching errors that weren't
getting caught by the ErrorBoundary
Technically this means we could now support geomodel on iOS. We need additional work though to add the model to the app and connect to the frame processor.
Note that offsetting the onboarding images ended up being a bit hard. We're
mostly doing so with a cropped version of one of the images.
Closes#1906
Also tried to de-flake some tests:
* Use fake timers when possible
* Ensure time travel helper doesn't un-fake our timers
* Wait for time travel helper to finish so changes don't happen outside of
act
* Stop mocking useCurrentUser in integration tests; that's app code and should
not be mocked when integrating all parts of the app; instead use signIn /
signOut helpers
* Remove unnecessary direct uses of act()
* Mocked some inatjs calls to prevent React Query complaints about undefined
query return values
* Close onboarding before all e2e tests
---------
Co-authored-by: Ken-ichi Ueda <kenichi.ueda@gmail.com>
* Use a wrapper around Text for all typography components that sets
app-specific defaults, allowing default RTL to work
* feat: make iOS permission prompt text translatable
* fix: allow RTL layout of char limit on text input bottom sheet
* fix: allow RTL photo carousel on ObsDetails and TaxonDetails
Removes react-native-reanimated-carousel, which needs to be patched to support
RTL and maybe wasn't providing anything beyond FlatList for us anyway.
InfoPlist.xcstrings will not be automatically updated on Crowdin sync pending
a Crowdin bug.
Work toward #2099
* Add a cameraZoomRange default
* Small detox version bump
* Add a mocked camera with a take photo button
* An e2e mock for the computer vision plugin predicting a frame
* Mock taking a photo by loading the first photo from the CameraRoll
Closes#1981
* Refactor SliderControl into own file, migrate to TS
* Update vision plugin to customize a taxonomy rollup cutoff threshold
A number of 0.01 means all nodes with a cv score lower than that are excluded from calculating the top predictions. Thereby saving resourced on computing those.
* Enable setting of threshold via debug menu
The threshold defaults to 0.0, i.e. no nodes are filtered out, because score is always positive.
* Update vision plugin
* More appropriate range of values
