GitHub Actions cannot modify .github/workflows/ files (403 "Resource not
accessible by integration"), causing "Error adding to tree" failure.
pyproject.toml is already handled natively by python release type.
Only Dockerfile needs the generic updater for x-release-please-version markers.
- tickets.py: guard against non-text messages in waiting_for_title FSM state
- payments.py: fix Wata webhook using wrong field name (order_id vs orderId),
add full payload to error log
- tariff.py: stop overwriting admin tariff settings on every bot restart,
sync_default_tariff_from_config now only creates if no tariff exists
- start.py: catch TelegramBadRequest specifically for "message is not modified"
instead of bare except with useless retry
- admin/tickets.py: downgrade ticket notification log from error to warning
for expected case of OAuth/email users without telegram_id
- pricing.py, countries.py, purchase.py: guard against expired FSM state
causing KeyError on 'period_days'
- blacklist_service.py: add 5-min in-memory cache to is_user_blacklisted()
to reduce DB load from per-request checks
- remnawave_service.py: fix "Session is closed" race condition — create
new RemnaWaveAPI instance per get_api_client() call instead of reusing
shared instance whose aiohttp session gets overwritten by parallel coroutines
- Fix async context manager usage in sync_users: __aenter__() result
was not assigned, so hwid_api_client held the context manager object
instead of the actual API client, causing AttributeError on
reset_user_devices()
- Add user existence check in _restore_missing_yookassa_payment before
INSERT to prevent ForeignKeyViolationError when user_id from payment
metadata no longer exists in users table
- Switch release-please to manifest mode (config-file + manifest-file)
- Add Dockerfile and docker workflow files as generic extra-files
- Add x-release-please-version annotations for automatic version replacement
- Bump hardcoded v3.6.0 to v3.7.0 to match current release
- Add total_threshold_gb and node_threshold_gb to ExportCsvRequest
- Compute GB/day, risk level, risk ratio for each user when thresholds set
- CSV includes Total GB/day, Risk Level, Risk Ratio, Risk GB/day columns
- Add node filter: filter traffic by selected nodes, recalculate totals
- Add status filter: filter by subscription status (active/trial/expired/disabled)
- Add custom date range: support start_date/end_date params alongside period
- Refactor _aggregate_traffic to use date strings with stable 5-min cache keys
- Add cache eviction for expired entries to prevent memory leaks
- CSV export now respects all active filters and custom date range
- Extract _get_status helper, add _compute_date_range helper
- Add node filter (comma-separated UUIDs) and status filter query params
- Add custom date range (start_date/end_date) as alternative to period
- Fetch connected device count per user via HWID API (semaphore=10)
- Cache key changed to (start_str, end_str) tuple for both modes
- CSV export now respects all active filters and date range
- Backend returns available_statuses and filtered nodes list
- Validate future dates, max 31-day range
Cabinet was calling YooKassaService.create_payment() directly, bypassing
PaymentService which saves the payment record to the local database.
When YooKassa webhook arrived, the payment was not found in the DB,
causing payment processing failures.
Now uses PaymentService.create_yookassa_payment() and
create_yookassa_sbp_payment() consistently with all other payment methods.
Also standardizes metadata key from 'type' to 'purpose' to match bot flow.
- Switch from per-user to per-node API strategy in _aggregate_traffic
(O(nodes) calls instead of O(users), ~10 vs ~200 requests)
- Add retry with exponential backoff for 429 in _make_request
- Reduce concurrency limit from 20 to 5 to prevent request bursts
- Switch from get_bandwidth_stats_node_users (broken UUID matching) to
get_bandwidth_stats_user per user (same API as working detail page)
- Add tariff filter with available_tariffs in response
- Add concurrency-limited parallel per-user bandwidth stats fetching
Sort by tariff_name/full_name crashed with TypeError when some values
were None (fallback to 0) mixed with strings. Use empty string fallback
for string fields with case-insensitive comparison.
Add paginated GET /admin/traffic endpoint aggregating per-user traffic
across all nodes with server-side sorting, search, and 5-min in-memory
cache. Add POST /admin/traffic/export-csv to generate CSV and send
to admin via Telegram DM.
Telegram API rejects messages with mismatched HTML tags. When
truncate_for_blockquote cuts the description mid-way, it can leave
tags like <i>, <b> unclosed inside the blockquote. Telegram then
fails with "Unmatched end tag" error.
Add _close_open_tags helper that scans for unclosed tags and appends
closing tags in reverse order. Also ensure the total length with
closing tags still fits within the message budget.
Always fetch 30 days with daily_bytes per node and categories.
Frontend computes period totals locally without extra API calls.
Removes days query param.
Per-node queries (8+ calls) hit Remnawave rate limit. Switch back to
single get_bandwidth_stats_user call with %Y-%m-%d date format (same
as traffic_monitoring_service). Add response logging to debug format.
Also optimize panel-info to use accessible-nodes instead of all-nodes.
The /api/bandwidth-stats/users/{uuid} endpoint rejects date params.
Switch to querying each accessible node via the working legacy
endpoint /api/bandwidth-stats/nodes/{uuid}/users/legacy and finding
the user in the per-node results.
