Move the desktop app onto a Vite-powered React shell with username/password setup, diagnostics, themed core views, offline sync queueing, and tighter Electron runtime boundaries.
Allow the desktop renderer to authenticate through the app login endpoint and call API routes from its packaged origin without weakening non-API responses.
Client portal: add min-w-0, break-words, and flex gap/shrink utilities on
the projects grid cards so long project names no longer force horizontal
overflow and clip against the viewport edge.
Desktop: add app and tray icons, adjust Electron main process (window,
tray, lifecycle), renderer connection and API client updates, build script
and package metadata, and regenerate the bundled renderer script.
Add render_sandboxed_string() using SandboxedEnvironment so stored invoice and
quote HTML, ReportLab text templates, admin PDF previews, and invoice email
HTML are not evaluated with Flask's full template globals (mitigating SSTI).
Add regression tests for sandbox behavior and demo user permissions.
Create the auto-provisioned demo user with legacy role "user" and the RBAC
user role so public demos cannot reach settings or the PDF layout editor.
On startup, downgrade an existing demo user that still has admin or
super_admin roles left over from older releases.
Document behavior in docs/deploy/RENDER.md and README.md.
Introduce a connection manager (state machine, listeners, offline handling,
token/server binding) with timer start/stop reconciliation helpers. Refactor
app.js to use it and extend the first-run wizard (welcome + server + token).
Add node:test coverage for the manager, timer operations, and an /api/v1/info
integration harness. Rebuild renderer bundle for packaging.
Bump desktop toolchain (electron, electron-builder, esbuild) and Python
package version to 5.3.2 in setup.py.
Client-portal-enabled users (main app login, typically viewer) were not
included in get_allowed_client_ids(), so ProjectService and other callers
saw scope_client_ids=None and listed every project.
- Return [client_id] for is_client_portal_user in User.get_allowed_client_ids
- Derive get_allowed_project_ids from allowed client IDs for all non-admins
- Apply client/project scope and access checks from allowed IDs, not only
subcontractor is_scope_restricted (fixes user_can_access_* for portal)
FixesDRYTRIX/TimeTracker#592.
Tests: extend test_scope_filter with client_portal_scoped_user and API
isolation for GET /api/v1/projects.
Desktop (Electron):
- Add two-step first-run wizard: test TimeTracker via GET /api/v1/info, then log in with API token
- Replace bogus token check with validateSession (users/me, fallback to timer/status for narrow scopes)
- Normalize base URLs; classify TLS/DNS/timeout errors; periodic 401 forces re-login
- Settings save/test use public + authenticated checks; prebuild/prestart and npm test
Server:
- Exempt /api/v1/info, /api/v1/health, and POST /api/v1/auth/login from HTML setup redirect
- Include setup_required on GET /api/v1/info for unfinished installs
Mobile (Flutter):
- Validate saved token against new server URL before persisting settings change
- Remove unused lib/core/config.dart; point BUILD_CONFIGURATION at app_config.dart
Docs: DESKTOP_SETTINGS, desktop README, mobile-desktop-apps README, REST_API /info
Delete app/utils/posthog_features.py. It was unused by routes, and
is_posthog_enabled() always returned False, so flags never activated and
the API was misleading.
Document the real model: deployment behavior uses environment variables
and app/config.py; per-user UI preferences stay on the user record.
Refresh PostHog monitoring guides and README accordingly, update the
incomplete-implementations note, and soften posthog_segmentation wording
so it does not imply an in-app PostHog flag layer.
Register optional blueprints and the optional audit_logs module with full tracebacks (logger.exception and stable extra fields). Re-raise registration errors when FLASK_ENV is development or DEBUG is enabled so local misconfiguration fails fast; production and testing keep skipping optional modules after logging.
Update REST API, API versioning, architecture, project structure, contributor guide, and CONTRIBUTING for global search responses (partial and per-domain errors), shared run_global_search in app/services/global_search_service.py, and blueprint registry observability.
Document the dual HTTP surface everywhere integrators look: OpenAPI intro and servers, ARCHITECTURE, REST_API, API_VERSIONING (deprecated vs internal routes, shared modules), and CONTRIBUTING (v1-first rule).
Session JSON routes in app/routes/api.py that overlap REST v1 now return X-API-Deprecated and a Link header with rel successor-version, implemented via app/utils/api_deprecation.py.
Extract shared global search into app/services/global_search_service.py for both GET /api/search and GET /api/v1/search while preserving legacy short-query 200 empty responses and v1 400 validation.
Refactor legacy POST /api/timer/start, /api/timer/stop, and the start path of /api/timer/resume to use TimeTrackingService; keep existing socketio emits for the web UI.
Add tests/test_api_deprecation_headers.py and adjust search partial-failure tests to patch Project.query on the service module.
Describe how GET /api/openapi.json sets info.version via get_version_from_setup()
(setup.py at runtime, with TIMETRACKER_VERSION or APP_VERSION overrides and Flask
APP_VERSION fallback), and point contributors to the same rules in version
management and project structure docs.
Update the documentation audit entry for the former hardcoded OpenAPI version
gap, document tests/test_api_route_contract.py and the refined test_api_v1
isolation guidance in the testing strategy, add a quick-reference pytest
invocation for the contract suite, and note info.version in the API consistency
audit next to the OpenAPI references.
Document session-auth GET /api/stats/value-dashboard in REST_API.md
(response shape, last_7_days, estimated value fields, 10-minute Redis
cache). Link dashboard session JSON routes from docs/API.md and note
that /api/v1 scopes do not apply to those legacy paths.
Update the comprehensive API test so GET /api/projects/<id>/tasks is
expected to return every task status, including done and cancelled,
which matches the time-entry UI.
Changelog: record the documentation update and the test correction.
Global search referenced Client.company, which is not a column on Client, so client matches failed at runtime. Legacy and v1 search, plus search_clients(), now filter on name, email, description, and contact_person; result descriptions use the same fields. Legacy /api/search returns count: 0 for queries shorter than two characters so responses stay consistent.
OpenAPI info.version is taken from get_version_from_setup(), with a config fallback when the resolved version is unknown. get_version_from_setup() also honors TIMETRACKER_VERSION and APP_VERSION for CI and container builds.
Client.__init__ accepts custom_fields. ClientService no longer passes status= into Client(), which the initializer does not support.
Tests add HTTP route contract checks and OpenAPI version alignment, fix subcontractor search fixtures (Client/Task construction and v1 client fixture naming), and update related API integration tests.
Adds out_of_scope_entities fixture and a session-authenticated search test so projects, tasks, and clients from unassigned clients are excluded alongside the API v1 search scope refactor.
Smart notifications (opt-in under user settings): NotificationService builds candidates from the user's local day and active timers; GET /api/notifications and POST /api/notifications/dismiss; migration 150 adds user columns and user_smart_notification_dismissals. /api/summary/today uses the same local-day totals. Client polls from smart-notifications.js; toastManager.show gains onDismiss for server dismiss sync. Config and env.example document SMART_NOTIFY_* variables.
Value dashboard: StatsService with Redis-backed caching, GET /api/stats/value-dashboard, dashboard template and dashboard-enhancements polling alongside existing widgets.
API v1 token search now uses apply_project_scope and apply_client_scope on queries; scope_filter adds apply_project_scope; tests extended for the new helper.
Add a support modal with usage stats, tier and license links, share control, and offline-safe outbound CTAs. Surface support from the header, sidebar, user menu, dashboard card, and settings "Support & Community" section without hiding entry points when a supporter license is active.
Introduce UsageStatsService and a persisted users.support_stats_reports_generated counter incremented on key report exports and custom report views. Add SupportPromptService for session-scoped soft toasts (after export, dashboard milestones, long session via POST /donate/request-soft-prompt).
Wire consent-aware track_event names support.* and mirror funnel rows in DonationInteraction; fix has_recent_donation_click to treat link_clicked as a recent click. Document events and SUPPORT_* / migration notes in docs.
Tests: tests/test_support_services.py for prompt and usage stats behavior.
Add VersionService to fetch and cache the latest GitHub release, compare it to the installed semver (APP_VERSION when valid, else setup.py), and expose admin-only GET /api/version/check and POST /api/version/dismiss on the legacy /api blueprint (session or Bearer token).
Persist per-user dismissal in users.dismissed_release_version (Alembic 148) and show a non-blocking update card in base.html for administrators. Add packaging for semver parsing and tests for comparison, service, and routes.
Document configuration in docs/admin/deployment/VERSION_MANAGEMENT.md and endpoints in docs/api/REST_API.md and docs/API.md.
Document https://crowdin.com/project/drytrix-timetracker in CONTRIBUTING.md, docs/TRANSLATION_SYSTEM.md, and contributor guides. Update CONTRIBUTING_TRANSLATIONS.md with the public project URL, clearer translator vs maintainer onboarding, and an optional "Further Crowdin integration" section (GitHub app, automation, QA, glossary, notifications).
Refresh crowdin.yml header comments (project URL, secrets pointers), normalize preserve_hierarchy to a boolean, and keep the nb→no languages_mapping quoted for YAML 1.1 compatibility.
Add crowdin.yml mapping the English gettext catalog to per-locale messages.po paths, with an explicit nb→no mapping so Norwegian matches app/config.py.
Add a manual GitHub Action (Crowdin sync) to upload sources and download translations when CROWDIN_PROJECT_ID and CROWDIN_PERSONAL_TOKEN are configured.
Extend CONTRIBUTING_TRANSLATIONS with maintainer steps, cross-link TRANSLATION_SYSTEM and contributor guides, refresh TRANSLATION_SYSTEM metadata, and note the integration under [Unreleased] in CHANGELOG.md.
Add Move up / Move down controls and an Order column for quote line items, Costs, and Extra goods on create and edit pages (Issue #584). DOM row order matches POST field order so existing QuoteItem.position handling stays correct.
Fix the quote detail Valid until row by using quote.is_expired instead of an undefined Jinja now() (Issue #583). Submit a past valid_until in the web regression test so the view path is exercised.
Document translation contributions without Git: add docs/CONTRIBUTING_TRANSLATIONS.md, a Translation improvement issue template, links from CONTRIBUTING.md and TRANSLATION_SYSTEM.md, and a Translations subsection in docs/development/CONTRIBUTING.md. Refresh CHANGELOG [Unreleased] for these items.
The will-navigate guard compared file URL origins to the literal string
file://, but the URL API reports an opaque origin for file pages, so
legitimate file navigations were blocked and the window could fail to
load reliably on Windows.
Import utils/helpers from the renderer entry so esbuild includes
window.Helpers in the bundle, restoring formatters and isValidUrl after
build:renderer.
Documentation: desktop README explains the renderer bundle workflow;
Windows desktop troubleshooting covers stuck loading; frontend quality
gates table notes the app.js entry and rebuild step.
Add quote_items.line_kind (item | expense | good) plus display_name,
category, line_date, and sku so one table still drives totals, PDFs,
and acceptance stock logic.
- Migration 147: new columns with backfill line_kind=item
- Quote create/edit: three sections; stock and warehouse only when a
line item is sourced from inventory; shared JS partial
- POST parsing, positions, and duplicate-quote copying for all fields
- API v1 quote items accept the new fields with defaults for old clients
- View, client portal, and PDF/fallback rendering use display_name and
line metadata where relevant
- Integration tests: stock POST shape; expense and good line creation
Docs: extend INVENTORY_MANAGEMENT_PLAN QuoteItem and migration notes.
Add Move up / Move down controls and a 13-column desktop grid on quote
item rows so users can change line order without deleting entries. Order
matches persisted quote_items.position (migration 146) for PDFs and views.
Update CHANGELOG [Unreleased] with Issue #584. Fix inventory integration
test to read quote.items[0] (relationship returns a list, not a query).
Quotes (#583):
- Add requires_approval, approval_level, and can_be_sent; wire create form
- Migrations 145 (approval columns) and 146 (quote_items.position)
- Order Quote.items by position; set positions on create/edit/duplicate/API
- Fix view template approval branch (not_required); add web regression test
Invoices / PEPPOL:
- Use the same Factur-X embed and PDF/A-3 normalization for export and
email attachments; Associated File Data + text/xml metadata
- CII/UBL validators, pdfa3, zugferd, and invoice_pdf_postprocess helper
- Bundle compact sRGB ICC (app/resources/icc/); INVOICE_SRGB_ICC_PATH override
- Package data in setup.py; extend PEPPOL_EINVOICING.md and tests
- Generate and send Idempotency-Key for queued POST /api/v1/time-entries creates.
- Wire sync queue and API client so retries do not duplicate entries after reconnect.
- Adjust settings and sync use case for the new behavior.
- Add Linear import (GraphQL, personal API key, optional team key filter).
- Centralize integration HTTP via integration_session and session_request.
- Add integration_sync_context for project/task refs and custom_fields metadata.
- Refactor Asana, GitHub, GitLab, Jira, Trello, ActivityWatch, and QuickBooks to use helpers.
- Extend integration UI, settings, and scheduled sync behavior as needed.
- Add POST /api/v1/time-entries/import-csv and POST /api/v1/time-entries/bulk.
- Support Idempotency-Key on POST /api/v1/time-entries with 24h replay per token.
- Enforce per-token minute/hour limits (Redis when available, local fallback otherwise).
- Extract bulk and CSV import logic into dedicated services; trim legacy api route surface.
- Add tasks.custom_fields (JSONB) for integration metadata (migration 143).
- Add api_idempotency_keys table keyed by token, scope, and key hash (migration 144).
- Register ApiIdempotencyKey model for safe replay of POST /api/v1/time-entries.
Detect SQLAlchemy 2.0 flush execution via _warn_on_events in
Settings._session_in_flush so get_settings() does not add/commit
during flush_context.execute(), avoiding ResourceClosedError in
invoice currency smoke tests.
Payment smoke tests now set a password on factory users and submit
it on login, matching local AUTH_METHOD requirements (fixes 302 to
login on /payments).
Allow HTTP 403 in test_non_admin_cannot_access_roles when the app
denies non-admins directly instead of redirecting with 200.
exchange_code_for_tokens called logger.debug without a defined logger, which triggered flake8 F821 in CI. Import logging and use getLogger(__name__) at module scope.
Flutter rejects release mode with --simulator; use the default debug
simulator build so CI can compile and zip Runner.app. Document the
constraint in workflow comments.
Screens referenced dateFormat and timeFormat, which do not exist on
UserPrefs. Align with the model and date_format_utils key-based API so
flutter analyze and tests compile.
Implement the missing Flutter data layer so release builds compile: Dio ApiClient for /api/v1 (timer, time entries, projects, tasks, finance, time-off, users/me), JSON models, Hive LocalStorage, and offline SyncService queue.
Add OpenTelemetry (opentelemetry package) with initMobileOpenTelemetry() reading OTEL_EXPORTER_OTLP_ENDPOINT and OTEL_EXPORTER_OTLP_TOKEN via --dart-define, matching server OTLP base URL and Basic auth behavior. Instrument login token validation, timer start/stop, and sync pending.
Fix SyncUseCase to import storage SyncService, use trusted insecure hosts, and call syncAll().
GitHub Actions (build-mobile.yml, cd-release.yml): run flutter test; pass OTLP secrets into flutter build apk/appbundle/ios; switch iOS CI to release simulator builds and package build/ios/iphonesimulator/Runner.app to avoid requiring an Apple Development Team for generic device builds.
.gitignore: allow tracking mobile/lib/data/ despite the repo-wide data/ ignore rule.
Introduce a centralized OTel layer (app/telemetry/otel_setup.py) that reuses
the existing OTLP endpoint and token, exports traces and metrics over OTLP/HTTP,
and instruments Flask plus SQLAlchemy. Manual OTLP log export remains for base
and product analytics; log records now include trace_id, span_id, and
event_category where tracing is active.
Business spans and product metrics cover invoices, timers, reports, auth,
webhook delivery, and scheduled jobs. RED-style HTTP metrics are recorded in
after_request alongside existing Prometheus counters. ENABLE_TRACING and
ENABLE_METRICS default on when credentials exist; graceful no-op when they do not.
Privacy is preserved: user_id appears on traces only when detailed analytics is
opted in; metrics never carry user_id; _remove_pii behavior for analytics is
unchanged. Responses inject traceparent when tracing is enabled for future
browser correlation.
Tests: test_otel_integration.py and per-test reset_for_testing() in conftest
so each app factory can reinitialize OTel.
Remove default form-input shadow and top margin inside tinted invoice and
quote line rows. The row already has a border and hover shadow; inputs
were doubling that edge so the top of each row looked overly heavy.
Cover GET edit form and PATCH API when the user has edit_own_time_entries
on their own entries, complementing the timer route and API changes in the
mobile shell / timer edit work.
- Allow schedule edits (project, task, start/end, break) for users with
edit_own_time_entries on their own entries in API update_entry and
timer edit; scope project lists for subcontractors; admin-only source
dropdown on edit timer form.
- App shell: min-width/overflow fixes, header layout, compact bottom nav
on very narrow viewports (#573), dashboard timer block responsive layout.
- Invoice and quote edit: min-w-0 on grids/cells; scoped stronger neutral
borders for .form-input on #editInvoiceForm and #quote-form (#574).
Add Settings.mail_test_recipient (migration 142) and wire Admin → Email Configuration to save, load, and prefill the SMTP test recipient field.
Extract invoice email PDF/HTML building into _build_invoice_email_payload for reuse. New send_invoice_template_test_email sends a real message with subject "Invoice template test: …", reusing production rendering and PDF attachment, without updating invoice status or creating InvoiceEmail records.
Add POST /admin/email-templates/<id>/send-test (optional recipient override, optional invoice_id) and Send test email UI on template view and edit pages. Extend smoke and unit tests.
Purchase orders with line items crashed the detail page because the template
referenced item.stock_item without a SQLAlchemy relationship, and used
datetime.now() in Jinja where datetime was not in context.
Add PurchaseOrderItem.stock_item, pass default_received_date from the view
route, and add a regression test for viewing a PO that includes line items.
Read telemetry/sentry status in the admin dashboard from build-injected analytics defaults with env fallback, so release images correctly show OTLP configuration. Include both install_id and telemetry_fingerprint in base and detailed telemetry payloads to reliably differentiate installations, and add tests for the new fingerprint fields.
Switch product and installation telemetry to OTLP/Grafana across runtime config, CI injection, docs, and tests to unify telemetry transport and simplify privacy-focused opt-in behavior.