Zum Inhalt

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.

Unreleased

1.7.1-beta.1 - 2026-06-01

Added

  • MIT LICENSE file with full license text; linked from README
  • "Über Mich" page with humanist values and personal stance (#96)

1.7.0 - 2026-03-07

Added

  • Fullscreen charts support x-axis zoom and pan via pinch gesture; double-tap resets the view; axis labels recalculate for the visible range (#91)
  • Category (Epistemisch/Aleatorisch) and prediction type (Wahrscheinlichkeit, Ja/Nein, Faktual, Intervall) filter chips in the prediction list; chips appear only when relevant (#93)
  • Overdue filter chip cycles through three states: no filter, overdue only, not overdue only
  • Manual import shows full checkbox list with per-question selection; duplicates are pre-deselected, struck through, and non-selectable; imported count shown on button and success card (#94)

Fixed

  • Back navigation from a Winkler Score prediction detail now returns to the fullscreen chart in its previous zoom state (#91)
  • Factual prediction type missing from type filter chips (#93)
  • Overdue filter had no effect on the resolved tab (#93)

1.7.0-beta.7 - 2026-03-07

Fixed

  • Factual prediction type missing from type filter chips (#93)
  • Overdue filter chip had no effect on the resolved tab (#93)

1.7.0-beta.6 - 2026-03-07

Added

  • Manual import shows full checkbox list with per-question selection; duplicates are pre-deselected, struck through, and non-selectable (#94)
  • Import button and post-import success card show actual imported question count for both manual and AI import (#94)

1.7.0-beta.5 - 2026-03-07

Added

  • Overdue filter chip cycles through three states: no filter, overdue only, not overdue only

1.7.0-beta.4 - 2026-03-07

Added

  • Category (Epistemisch/Aleatorisch) and prediction type (Wahrscheinlichkeit, Ja/Nein, Intervall) filter chips in the prediction list; chips appear only when relevant (#93)

1.7.0-beta.3 - 2026-03-07

Fixed

  • Back navigation from a Winkler Score prediction detail now returns to the fullscreen chart in its previous zoom state (#91)

1.7.0-beta.2 - 2026-03-07

Added

  • Fullscreen charts support x-axis zoom and pan via pinch gesture; double-tap resets the view; axis labels recalculate for the visible range (#91)

1.6.0 - 2026-03-07

Fixed

  • Winkler Score data point links now active only in the fullscreen detail view, not on the summary chart; any tap on the summary chart opens fullscreen, consistent with Brier and Log Loss (#90)

1.5.0 - 2026-03-07

Added

  • Winkler Score chart: tapping a data point opens the corresponding prediction detail; back navigation returns to statistics (#90)

1.4.2 - 2026-03-06

Fixed

  • Winkler Score chart y-axis is now logarithmic; scores spanning multiple orders of magnitude are readable without visual compression of small values (#89)

1.4.1 - 2026-03-06

Fixed

  • Replace misleading Winkler Score mean card with a per-estimate scatter chart; dots coloured green (hit) and red (miss) (#88)

1.4.0 - 2026-03-06

Added

  • Winkler Score for interval predictions on the statistics screen; shows mean interval score and hit rate (#88)

Fixed

  • Template share buttons now write to a temporary file before sharing; XFile.fromData() produced no valid URI for Android share intents (#81)

1.3.0 - 2026-03-06

Added

  • AES-256-GCM encrypted full backup with PBKDF2-SHA256 key derivation; create and restore .kbak files from Settings (#85)
  • Delete button in prediction detail screen AppBar with confirmation dialog (#87)
  • Per-tab sort direction persisted via SharedPreferences; each tab remembers its own sort order across sessions (#86)
  • Skip duplicate questions by default on import; configurable toggle in the import screen (#84)
  • Full multi-select list in AI generator preview, all pre-selected; replaces the static "first 5" preview (#83)
  • YAML template buttons for each prediction type in import screen
  • Share button in prediction list view; exports the active tab as JSON
  • Tag chip selector on new-prediction screen
  • Preselect existing tags in the bulk-edit dialog
  • "Ohne Tag" filter chip in statistics screen
  • Exact 5 % steps in calibration chart x-axis

Fixed

  • Invalid confidence values (< 0.5) in documentation examples (#82)

1.2.0 - 2026-03-05

Breaking Changes

  • Database schema updated; migration runs automatically on first launch. Downgrading to v1.0.0 is not supported without data loss.

Added

  • Overdue indicator on home screen stat cards; cards turn red with a warning icon when unresolved predictions have passed their deadline (#54)
  • "Überfällig" filter chip in prediction list to show only overdue predictions across open and pending tabs (#55)
  • Delete button in resolve screen AppBar with confirmation dialog (#56)
  • Sort order toggle in prediction list; reverses per-tab default (open/pending: oldest first; resolved: newest first) (#57)
  • Tag selection via FilterChips in the AI generator (#68)
  • Calendar toggle in pending tab AppBar to sort by deadline; entries without a deadline stay at the end regardless of direction
  • Global tag manager in Settings to delete tags from all predictions at once, with per-tag confirmation dialog (#70)
  • "Ohne Tag" filter chip in prediction list; visible only when untagged entries exist, visually distinct via secondary-color border (#72)
  • Rename action in global tag manager; inline dialog with pre-filled name, updates all predictions at once (#73)
  • Confidence slider snaps to 5 % steps from 50 to 100 % (#71)

Fixed

  • Interval confidence slider minimum raised to 50 % (#67)
  • Calibration dot size scaled proportionally to the most-populated bin instead of a hard clamp (#69)
  • Type cast error in v3 schema migration (#73)

1.2.0-beta.1 - 2026-03-05

Added

  • "Ohne Tag" filter chip in prediction list; visible only when untagged entries exist, visually distinct via secondary-color border (#72)
  • Rename action in the global tag manager; inline dialog with pre-filled name, updates all predictions at once (#73)
  • Confidence slider now snaps to 5 % steps from 50 to 100 %; one-time startup migration rounds existing values (#71)

Fixed

  • Type cast error in v3 schema migration (addColumn expected GeneratedColumn<Object>) (#73)

1.1.1-beta.1 - 2026-03-05

Added

  • Calendar toggle in the pending tab AppBar to sort by deadline; entries without a deadline stay at the end regardless of direction
  • Global tag manager in Settings to delete tags from all predictions at once, with per-tag confirmation dialog (#70)

1.1.0-beta.1 - 2026-03-05

Added

  • Overdue indicator on home screen stat cards; cards turn red with a warning icon when unresolved predictions have passed their deadline (#54)
  • "Überfällig" filter chip in prediction list to show only overdue predictions across open and pending tabs (#55)
  • Delete button in resolve screen AppBar with confirmation dialog (#56)
  • Sort order toggle in prediction list; reverses per-tab default (open/pending: oldest first; resolved: newest first) (#57)
  • Tag selection via FilterChips in the AI generator, sourced from existing database tags (#68)

Fixed

  • Interval confidence slider minimum raised to 50 %; calibration bins restructured to cover 50–100 % range (#67)
  • Calibration dot size scaled proportionally to the most-populated bin instead of a hard clamp (#69)

1.0.0 - 2026-03-04

First stable release of Kailibrate. All features from the v0.20–v0.28 beta cycle are considered production-ready. The database schema (v5), import/export format (v2), and public API surface are now stable; breaking changes will follow semver and require a major version bump.

0.28.0 - 2026-03-04

Added

  • AI-powered question catalog generator via OpenRouter; configurable prompt templates, question count (5/10/15/20), and model selection
  • API key for OpenRouter stored securely; write-only after saving (#40)
  • Preview screen before AI import with share-as-obfuscated-JSON option
  • Tags input in AI generator restricts LLM-assigned labels (#36)
  • Generation cost and token count shown after each AI run (#39, #50)
  • Configurable model list (one per line); selection persists per session
  • Aleatory AI templates for binary and interval predictions (#46)
  • {date} placeholder in aleatory templates for future-only deadlines
  • Warning and exclusion checkbox for past-deadline questions in preview
  • New factual prediction type for epistemic Wahr/Falsch questions (#43)
  • Wahr/Falsch UI in estimate, resolve, feedback, card, and detail views (#43)
  • Deadline shown on prediction cards with color-coded urgency (#48)
  • Deadline editing from detail view for open/pending predictions (#47)
  • "Überfällig" badge for overdue predictions (#47)
  • Binary confidence slider restricted to 50–99 % (#33)
  • Default prompt templates can be deleted and recovered (#34)

Changed

  • probability type replaced by binary (aleatory) and factual (epistemic); import parser remaps legacy entries; schema migrated to v5 (#45)
  • Tapping any prediction card opens detail view; primary action via FAB (#47)
  • Mixed AI template removed in favour of focused aleatory templates (#46)
  • Docs restructured with why/what/how narrative and new concept pages

Fixed

  • Calibration scores use directional confidence for binary/factual predictions; correct "99 % FALSCH" appears in 99 % bin, not 1 % (#52)
  • Resolution outcome hidden in detail view until user has estimated (#51)
  • Generation cost shown prominently on AI import success screen (#50)
  • Settings navigation in AI generator corrected (#31)
  • AI import preview no longer reveals resolution icons (#32)
  • Yes/no prompts set predictionType "binary" correctly (#35)
  • Binary feedback shows actual Ja/Nein answer below verdict (#42)
  • Binary correctness based on binaryChoice == outcome (#41)

0.28.0-beta.1 - 2026-03-04

Fixed

  • Calibration curve and scores now use directional confidence for binary/factual predictions; a "99 % FALSCH" estimate that was correct now appears in the 99 % bin, not the 1 % bin (#52)

0.27.0-beta.1 - 2026-03-04

Fixed

  • Generation cost and token count now shown prominently above the success icon after AI import (#50)
  • Resolution outcome hidden in detail view until user has estimated; shows locked indicator instead (#51)

0.26.0-beta.1 - 2026-03-04

Changed

  • Restructured docs with why/what/how narrative flow
  • New Konzepte page covering calibration, categories, types, and states
  • New KI-Generator page with step-by-step usage guide
  • Removed deprecated probability type from examples
  • Fixed stale navigation description in prediction workflow

0.25.0-beta.1 - 2026-03-04

Added

  • Warning in AI generator preview when questions have past deadlines
  • Checkbox to exclude past-deadline questions before import
  • {date} placeholder in aleatory prompt templates so the model generates future-only deadlines

0.24.0-beta.1 - 2026-03-04

Added

  • Deadline shown on prediction cards with color-coded urgency: red for overdue, orange for due within 7 days (#48)

0.23.0-beta.1 - 2026-03-04

Added

  • Deadline can be set, changed, or cleared from the detail view for open and pending predictions (#47)
  • Overdue open/pending predictions flagged with "Überfällig" badge in the overview (#47)

Changed

  • Tapping any prediction card now opens the detail view; primary action (Schätzen/Auflösen) surfaced via FAB for non-resolved predictions (#47)

0.22.0-beta.1 - 2026-03-04

Added

  • Aleatory AI templates for binary (Ja/Nein) and interval predictions on future events without known answers (#46)

Changed

  • probability prediction type removed; replaced by binary (aleatory) and factual (epistemic) throughout the app (#45)
  • Import parser remaps legacy probability entries to binary/factual based on category; schema migrated to v5 (#45)
  • Stats and feedback screens updated to reflect new type set (#44, #45)
  • Mixed AI template removed in favour of the two focused aleatory templates (#46)
  • LLM prompt docs and type reference updated (#45, #46)

0.21.0-beta.5 - 2026-03-04

Added

  • New 'factual' predictionType for epistemic Wahr/Falsch questions, distinct from 'binary' (Ja/Nein) used for aleatory events (#43)
  • Wahr/Falsch buttons in estimate, resolve, feedback, card, and detail views for factual questions (#43)

Changed

  • DB migration v4 converts existing epistemic binary entries to factual (#43)
  • New-prediction screen shows Wahr/Falsch segment only for epistemic category; Ja/Nein stays for aleatory (#43)
  • Default AI prompt template updated to use predictionType "factual" (#43)

Fixed

  • Binary feedback banner now shows actual Ja/Nein answer below Richtig/Falsch verdict (#42)

0.21.0-beta.4 - 2026-03-04

Fixed

  • Yes/no questions now show green for a correct "Nein" prediction; color and icon are based on binaryChoice == outcome instead of outcome alone (feedback banner, prediction card, detail view) (#41)

0.21.0-beta.3 - 2026-03-04

Added

  • Tags input in AI generator: restricts which labels the LLM assigns to generated questions, for consistent in-app filtering (#36)
  • Generation cost and token count shown in the preview after each run (#39)
  • Info icon in model settings links to openrouter.ai/models (#37)
  • Clipboard copy button for the model list (#37)

Changed

  • Model list editor replaced with a multiline text field (one model per line); easier to manage and paste from (#37)
  • Selected model is persisted and restored on next session (#38)
  • API key is now write-only: after saving, only bullet placeholders are shown with an "Ändern" button (#40)

Fixed

  • Yes/no question prompts (default and mixed templates) now correctly set predictionType: "binary" so the binary estimate UI is shown after import (#35)

0.21.0-beta.2 - 2026-03-03

Added

  • Default prompt templates can now be deleted; they are hidden via a per-device suppression list and remain recoverable from source code (#34)

0.21.0-beta.1 - 2026-03-03

Added

  • Configurable model list in Settings: add/remove OpenRouter models, first entry used as default (#34)
  • Model dropdown in AI generator form, pre-filled with the first configured model; selection persists per session (#34)
  • Binary confidence slider now spans 50–99 % (50 % = maximum uncertainty / guessing); below 50 % the answer direction should be flipped instead (#33)

Fixed

  • "Einstellungen" button in the missing-API-key card now navigates correctly with go_router (context.push instead of Navigator.pushNamed) (#31)
  • Import preview in AI generator no longer reveals correct/incorrect icons for questions with embedded resolutions (#32)

0.20.0-beta.1 - 2026-03-03

Added

  • Generate epistemic quiz questions via OpenRouter AI; supports custom prompt templates (editable and deletable) and a configurable question count (5 / 10 / 15 / 20)
  • API key for OpenRouter stored securely via flutter_secure_storage; configurable in Settings
  • Preview screen before import: category, source, question count, first 5 questions with resolution indicators
  • Share generated catalog as obfuscated v2 JSON directly from the preview screen, without importing first
  • ImportParser.obfuscateResolution() as public static method for encoding resolutions outside the database layer

0.18.2 - 2026-03-03

Fixed

  • "Select all" button now toggles: first tap selects all visible predictions in the active tab, second tap deselects all (#29)
  • Button icon and tooltip reflect current state (select_all ↔ deselect)

0.18.1 - 2026-03-03

Changed

  • Project renamed from Calibrate to Kailibrate: package name, applicationId (dev.kailibrate.app), database filename (kailibrate.db), notification channel ID, and app title

Fixed

  • Release workflow: APK filename in files: parameter corrected (calibrate- → kailibrate-); fail_on_unmatched_files: true added

0.18.0 - 2026-03-03

Changed

  • Project renamed from Calibrate to Kailibrate: package name, applicationId (dev.kailibrate.app), database filename (kailibrate.db), notification channel ID, app title, and all documentation updated accordingly

0.17.4 - 2026-03-03

Fixed

  • Interval unit (e.g. "Liter", "km") is now stored in the Questions table (schema v3) instead of only in Estimates; previously the unit was lost for imported questions without bounds, causing the estimate screen to ask for it again
  • Estimate screen shows a read-only label instead of an editable text field when the unit is already known from the question or an existing estimate

0.17.3 - 2026-03-03

Fixed

  • Unit field in v2 interval imports now read from question level as fallback when absent from the estimate sub-object; data from callibrate-gen and similar tools was silently dropped (#28)
  • exportForSharing now includes unit for interval predictions so the unit survives re-import (#28)

0.17.2 - 2026-03-03

Fixed

  • Unit (e.g. km, °C) now persisted and shown when estimating interval predictions via the estimate screen; previously lost on every save (#28)
  • Prediction card shows numeric outcome with unit for resolved interval predictions instead of plain "Ja"/"Nein" (#28)

0.17.1 - 2026-03-03

Fixed

  • Tapping anywhere on a chart now opens the fullscreen view; previously only the bottom-left blank area responded because fl_chart's internal GestureDetector consumed touch events across most of the surface (#27)

0.17.0 - 2026-03-03

Added

  • Tap any chart on the statistics screen to open it fullscreen in landscape orientation for easier reading (#27)

0.16.1 - 2026-03-03

Fixed

  • Redundant floating action button for new prediction removed from home screen; the navigation tile serves as the single entry point (#26)

0.16.0 - 2026-03-02

Added

  • Brier Score and Log Loss history charts on the statistics screen; each chart shows the cumulative average after every resolved estimate, sorted chronologically, with a dashed coin-flip reference line (0.25 / ln 2) and a segmented button to limit the view to the last 25, 50, or 100 estimates

0.15.6 - 2026-03-02

Fixed

  • Interval bounds and numeric outcomes now display as integers when the stored value has no fractional part (e.g. "45 km" instead of "45.0 km"); unit is also shown in the feedback sheet's outcome row

0.15.5 - 2026-03-02

Fixed

  • Feedback sheet after estimating a pre-resolved question now displays resolution notes and the numeric outcome stored with the import (#24)

0.15.4 - 2026-03-02

Fixed

  • Resolve screen now shows the known answer for questions imported with an embedded answer (e.g. trivia catalogues) (#24)

0.15.3 - 2026-03-02

Fixed

  • Switch from exact to inexact alarm scheduling to prevent a PlatformException when SCHEDULE_EXACT_ALARM is not granted on Android 12+ (#22)

0.15.2 - 2026-03-02

Fixed

  • Inline estimate state (slider value, Ja/Nein selection) no longer resets when switching prediction type or toggling the deadline; root cause was ListView position-shift destroying the Consumer element and triggering an autoDispose provider reset (#21)

0.15.1 - 2026-03-02

Fixed

  • Tag filter no longer shows an empty list after all entries with a selected tag are deleted; stale filters are cleared on data refresh (#20)

0.15.0 - 2026-03-02

Added

  • Trash icon in selection mode AppBar to permanently delete all selected predictions including their estimates and resolutions; a confirmation dialog warns before the irreversible action (#19)

0.14.1 - 2026-03-02

Fixed

  • Interval outcome now recomputed from the current estimate bounds when estimating a pre-resolved question; previously a stale outcome from the original resolution could mark a wrong estimate as correct (#18)

0.14.0 - 2026-03-02

Added

  • Feedback sheet shown immediately after estimating a question that already has a resolution (e.g. trivia imports with embedded answers); same CalibrationFeedbackSheet as the resolve flow (#16)

0.13.0 - 2026-03-02

Added

  • Bottom sheet after resolving shows Brier contribution of the current estimate, overall Brier Score and Log Loss, and a type-specific section when predictions of multiple types exist (#16)

0.12.2 - 2026-03-02

Fixed

  • Statistics tag filter replaced with FilterChips; the previous autocomplete showed no options on empty input, making the filter undiscoverable (#9)

0.12.1 - 2026-03-02

Fixed

  • Tag dialog now flushes the text field before saving; typed input was silently discarded when "Setzen" was pressed without Enter, deleting all existing tags instead of replacing them (#17)

0.12.0 - 2026-03-02

Added

  • Multi-select mode in the predictions list; long-press activates selection, "Select All" covers the active tab and tag filter (#15)
  • Bulk tag editing for selected predictions; the dialog replaces tags on all selected items at once (#15)

0.11.0 - 2026-03-02

Added

  • Tag filter in the sharing export dialog; select one or more tags to export only matching resolved predictions (#8)

0.10.1 - 2026-03-02

Fixed

  • Import parser now reads resolution fields in version 1 files; previously they were silently ignored, preventing auto-resolution (#14)

0.10.0 - 2026-03-02

Added

  • Clipboard import now recognizes ```json and ```yaml fences; LLM-generated output can be pasted directly without editing (#13)

0.9.0 - 2026-03-02

Added

  • Export resolved questions for others without own estimates; optional category filter (epistemic/aleatory) (#8)
  • Version tile in Settings shows build number and share icon for JSON debug info (OS, device model, API level) (#11)

0.8.0 - 2026-03-02

Added

  • Statistics screen now supports three combinable filters: category (single-select), prediction type (multi-select), and tags (autocomplete, OR-linked) (#9)

0.7.1 - 2026-03-02

Changed

  • Import format reference restructured with per-field descriptions and an import behavior table explaining the "Lösung vorhanden" flow
  • LLM prompt guide added to docs: three copy-ready prompts for generating calibration exercises with hidden answers

0.7.0 - 2026-03-01

Added

  • MkDocs documentation site deployed to GitHub Pages with versioned URLs via mike; triggered automatically on every release tag
  • Settings screen links to the documentation version matching the installed APK (#6)

0.6.0 - 2026-03-01

Added

  • Tapping a resolved prediction card opens a detail view showing question text, category, tags, estimate, and resolution with outcome, notes, and numeric value (#5)

0.5.0 - 2026-03-01

Added

  • Import now supports resolutions: questions with estimate and resolution are marked resolved immediately; questions with a resolution but no estimate show a "Lösung vorhanden" hint and auto-resolve after estimating (#4)
  • Export obfuscates resolution data with ROT13 + Base64 to prevent accidental spoilers when sharing question sets (#4)

0.4.3 - 2026-03-01

Fixed

  • Import parser now accepts version 2 export format: reads per-question categories, hasKnownAnswer/knownAnswer, and nested estimate objects (#2)

0.4.2 - 2026-03-01

Fixed

  • Add missing Java 17 (Temurin) setup step to release workflow
  • Strip newlines before base64 decode to prevent keystore corruption

0.4.1 - 2026-03-01

Fixed

  • FilterTab enum was renamed from _FilterTab in predictions_screen.dart but the old private name remained in app.dart, breaking the CI build

0.4.0 - 2026-03-01

Added

  • Dashboard stat cards (Offen, Ausstehend, Aufgelöst) are now tappable and navigate directly to the predictions list on the matching tab

Fixed

  • Release APK signing: replaced machine-specific debug keystore with a dedicated release keystore loaded from key.properties or CI secrets; prevents "Update not installed" errors when sideloading updates

0.3.0 - 2026-03-01

Added

  • Optional inline estimation when creating a new prediction; saves the separate estimate step for users who already know their probability
  • Estimate fields in JSON/YAML import files: predictionType, probability, binaryChoice, confidenceLevel, lowerBound, upperBound, and unit — imported estimates are saved automatically
  • Sample data updated with embedded estimate examples for all three prediction types

Changed

  • Estimate form logic extracted to shared/widgets/estimate_inputs.dart (EstimateFormState, EstimateFormNotifier, BinaryEstimateInput, IntervalEstimateInput, ConfidenceSlider) and reused across EstimateScreen and NewPredictionScreen

0.2.1 - 2026-03-01

Fixed

  • Enable core library desugaring in android/app/build.gradle so that flutter_local_notifications builds on all supported Android runtimes; resolves checkReleaseAarMetadata failure in CI

0.2.0 - 2026-03-01

Added

  • Binary prediction type: choose Yes/No and set confidence level; probability is derived as confidence (Yes) or 1 − confidence (No)
  • Interval prediction type: define a numeric range with confidence level; resolves as true when the measured value falls within the interval
  • Optional unit field on interval predictions (e.g. m, °C, kg)
  • Local deadline notifications: reminded the day before and on the deadline day at 09:00; rescheduled automatically on app start to survive device reboots (flutter_local_notifications, timezone)
  • Type-selector (SegmentedButton) on the new prediction screen
  • Type-aware estimate screen: slider for probability, Yes/No buttons plus confidence slider for binary, numeric fields plus confidence slider for interval
  • Type-aware resolve screen: numeric input for interval resolutions; outcome computed automatically from the stored bounds

0.1.2 - 2026-03-01

Fixed

  • Add mipmap launcher icons (mdpi–xxxhdpi) and adaptive icon for API 26+; resolves AAPT build error "resource mipmap/ic_launcher not found"

0.1.1 - 2026-03-01

Fixed

  • Drift column name conflict (Questions.text shadowed inherited text() method); renamed to questionText with .named('text')
  • Replace SharePlus.instance/ShareParams with stable Share.shareXFiles API (share_plus 10.x)
  • Bump compileSdk/targetSdk to 36 (required by path_provider and flutter_plugin_android_lifecycle)
  • Upgrade Gradle wrapper to 8.10.2, AGP to 8.7.0, Kotlin to 2.1.0

0.1.0 - 2026-03-01

Added

  • Core app: probability estimation, resolution, calibration stats, and JSON/YAML import
  • Settings screen, tag filter, and clipboard import for question sets
  • GitHub Actions release workflow for tag-triggered APK builds