Telegram webhook handler crashed with sqlalchemy.exc.MissingGreenlet
when processing any incoming message after committing the chat row:
TelegramChat.bot_id == bot.id
^^^^^^
MissingGreenlet: greenlet_spawn has not been called
AsyncSession expires all instances on commit. Accessing bot.id/bot.token
after that triggers implicit lazy-load I/O from a sync attribute getter,
which can't enter the greenlet dispatcher → crash.
Fix: snapshot bot.id + bot.token to locals before commit, refresh the
ORM instance after a successful commit so handle_command() can still
use it, and route the remaining call sites through the snapshot
variables.
Startup was crashing on fresh databases because:
- init_db() calls SQLModel.metadata.create_all(), which builds tables
from the model classes. CommandConfig didn't declare `locale`, so
the created command_config table lacked the column.
- The seeder then issued INSERTs that included locale='en', causing
`OperationalError: table command_config has no column named locale`.
The legacy migration #6 in migrate_schema creates command_config WITH
locale via raw SQL, so upgraded databases worked. Only fresh installs
broke.
Fix:
- Add `locale: str = Field(default='en')` to CommandConfig model so
create_all() produces a consistent schema.
- Add a defensive ALTER TABLE ... ADD COLUMN locale in migrate_schema's
else-branch, so any existing command_config table missing the column
(from a broken v0.1.0 install) is backfilled on next startup.
Previous attempt used `python3 -c "..." KEY=VALUE` which passes
KEY=VALUE as positional args, not environment variables — the python
block then crashed with KeyError: 'BODY' because nothing actually set
it in the environment.
Consolidate into a single heredoc-fed python3 block that reads
RELEASE_NOTES from the already-exported env var and reads TAG/VERSION/
IS_PRE after an explicit `export`. Uses <<'PY' so shell metachars in
the Python source (backticks, $, quotes) are not interpreted.
Also drops the redundant intermediate BODY variable — body is built
directly inside the single python invocation.
Previous implementation silently assumed any missing 'id' in POST
response meant "release already exists", then called an unguarded
python3 on the fallback response — which crashes (exit 1) if the
fallback also fails (e.g. release really doesn't exist).
New logic:
- Build JSON payload in Python (avoids shell escaping + CLI length limits)
- Capture HTTP status explicitly
- 201 → success
- 409 or "already exists" message → reuse existing (with HTTP check on fetch)
- Anything else → fail loudly with the response body printed
This also unblocks diagnosis of the current v0.1.0 failure by surfacing
the actual error the Gitea API is returning.
- Set fetch-depth: 0 so previous tag lookups work across full history.
- Use `-n 20` instead of HEAD~20..HEAD, which fails when the repo has
fewer than 20 commits (e.g. on the first release).
The two-step pattern (sparse-checkout RELEASE_NOTES.md, then full
checkout) left sparse-checkout config active on the workspace, so the
second checkout still only restored RELEASE_NOTES.md. Docker build
then failed with "open Dockerfile: no such file or directory".
Since both RELEASE_NOTES.md and the full source are needed in the same
job, one full checkout is simpler and correct.
- Skip token clear/redirect on 401 for unauthenticated requests
- Fix typo in test secret key in restart-backend script
- Remove completed plan documents (entity-relationship-refactor, ux-notification-improvements)
- Use one DEPLOY_TOKEN for both registry login and Gitea release API,
matching the claude-code-facts convention.
- Rename "Trigger Portainer redeploy" to "Trigger redeploy webhook" —
the step calls a generic DOCKER_REDEPLOY_WEBHOOK_URL, not a
Portainer-specific endpoint.
- Add .facts-sync.json to pin this project to the facts repo commit.
- Remove top paginator from dashboard events, keep only bottom
- Fix test message locale: pass UI locale to email/matrix bot tests
- Convert webhook auth mode from text input to icon grid selector
- Generate secure UUID tokens for webhook URLs instead of sequential IDs
- Move Recent Payloads into per-provider expandable container (lazy-loaded)
- Make template config languages dynamic via app settings instead of hardcoded
- Change default dev port to 5175
Add person exclude criteria to Immich auto-organize — assets containing
excluded persons are filtered out after candidate gathering. Also adds
full backup/restore system with export, import, scheduled backups, and
retention management.
- Fix github.* → gitea.* context consistency
- Add pre-release detection (skip :latest for alpha/beta/rc)
- Add release fallback (reuse existing if creation fails)
- Add prerelease field to release API call
- Use sparse-checkout for RELEASE_NOTES.md
- Skip Portainer redeploy for pre-releases
- Add version tag without v prefix
- Add manual build.yml for Docker image verification
Add a new "webhook" provider type that accepts arbitrary HTTP POST payloads,
extracts template variables via user-defined JSONPath mappings, and dispatches
notifications through the existing pipeline. Supports three auth modes
(HMAC-SHA256, Bearer token, none), bounded JSONPath cache, and 1MB payload limit.
Full stack: core provider + event parser, API endpoint, DB migration,
capabilities, seeds, default templates (EN/RU), frontend descriptor, i18n.
- Tracker now has default_tracking_config_id and default_template_config_id
that apply to all linked targets unless overridden per-target
- Dispatch falls back to tracker defaults when per-link configs are null
- Email bot creation validates SMTP connection before saving
- Email notifications sent as HTML (links render properly)
- Linked target items are expandable: collapsed shows config CrossLinks,
expanded shows config selectors; action buttons always visible
- Fix email bot test button icon (mdiEmailSend → mdiSend)
- Fix target type icons in LinkedTargetsSection for all types
- Provider filter moved above search in sidebar
- Route scheduled/memory test sends through the same NotificationDispatcher
the watcher uses — identical template rendering, media handling, caching
- Add preview_url field to MediaAsset (transcoded mid-size), separate from
thumbnail_url (small) and full_url (original). Dispatcher prefers preview_url
- Fix sendMediaGroup cache: extract file_ids from Telegram response and store
via async_set_many so repeat sends use cached file_ids
- Parallelize asset downloads in _send_media_group with asyncio.gather
- Filter unprocessed assets (archived/trashed/offline/no-thumbhash) at album
parse time in ImmichAlbumData.from_api_response
- Extract shared asset_to_media + collect_scheduled_assets into asset_utils.py
(single source for test dispatch and future real scheduler)
- Respect tracking config filters: limit, asset_type, favorite_only, min_rating
- Random asset sampling for scheduled sends
- Memory mode: "On This Day" date filter (same month+day, previous year)
- Skip dispatch when no matching assets found
- Remove ~250 lines of duplicated send logic from notifier.py
- Fix restart-backend.sh: proper env var export, Python path resolution, error log
Move "Favorites only" from a separate checkboxes array into the regular
fields grid as a toggle switch, aligning Scheduled Assets and Memory Mode
sections visually. Memory source moved to last position.
Template slot editors (notification + command) now use collapsible
accordion rows instead of showing all editors at once. Each slot
displays a compact header with status pill (empty/valid/warning/error).
Adds slot name filtering and a preview toggle button that swaps
between editor and rendered preview views.
- Command templates now match notification template style: type icons,
linked filenames via album shared links, location, favorite status
- Media mode sends text message first, then media as reply (was media-only)
- Search/find/person/place resolve asset public URLs from tracked albums'
shared links (share/{key}/photos/{id})
- Albums/summary commands include album public_url in context
- Enriched command template preview sample context with public_url, city,
country, is_favorite
- Extract sanitizePreview to shared lib/sanitize.ts
- Command template preview now renders HTML links (was raw text)
- Global provider filter moved above search in sidebar
- CLAUDE.md: template consistency + context variable sync rules
When both internal URL and external domain are configured, rewrite
asset download URLs from external to internal before fetching.
This avoids routing through public internet when the bot and
provider are on the same LAN.
Add broadcast target type that fans out notifications to multiple
child targets. Dispatch expands broadcast into children in
load_link_data() — dispatcher stays unaware. Children can be
toggled on/off via disabled_child_ids in config.
Also: dashboard provider card smaller font for names, scroll-to-form
on target edit, broadcast nav tab with counter, flag_modified fix
for JSON column updates, CLAUDE.md nav tree docs.
Replace all if/else chains keyed on provider type strings with a
descriptor-driven architecture. Each provider type (immich, gitea,
planka, scheduler, nut, google_photos) has a descriptor in
frontend/src/lib/providers/ that declares config fields, event
tracking fields, collection metadata, validation, and hooks.
Components now use getDescriptor(type) and render dynamically.
Dashboard provider card shows provider name + type when global
filter is active. Grid-items derived from registry.
Provider selectors in notification tracker, command tracker, and actions
forms now only show providers matching the global provider type filter.
Command config selector in command trackers also filters by provider type.
- Dashboard providers card now shows count of providers matching the
global provider type filter instead of special name/type display
- Fix provider update sending empty config when only name/icon changed,
causing 400 validation error (api_key required)
- Fix bot card header overflow by replacing "Sync with Telegram" text
button with icon button, add flex-wrap
- Rename sync button label to "Sync Commands"
- Remove decorative dashes from selector placeholders (— X — → X)
- Show selected provider name/icon in dashboard stat card when global
provider filter is active
- Add selector placeholder convention to frontend-architecture.md
- Introduce Receiver base class + typed subclasses (TelegramReceiver,
WebhookReceiver, EmailReceiver, etc.) in core/notifications/receiver.py
- Dispatcher uses typed Receiver objects instead of raw dicts, with
per-receiver locale-aware template rendering
- load_link_data resolves locale from TelegramChat.language_override at
load time: TargetReceiver.locale || chat.language_override || chat.language_code
- Add language_override field to TelegramChat (separate from auto-detected
language_code), with per-chat commands toggle and command dispatch using
override language
- Add locale field to TargetReceiver for explicit per-receiver overrides
- Add commands_enabled field to TelegramChat (default off) with
migration, gating command dispatch in both poller and webhook
- Show toggle switch per chat in bot tab for enabling/disabling commands
- Fix listener response to include bot name instead of just type
- Replace listener "Enabled" label + "Edit" link with toggle switch
and crosslink to command-trackers page
- Add locale support to notification templates (matching command template
pattern): TemplateSlot now has locale field with (config_id, slot_name,
locale) uniqueness, nested API format {slot: {locale: template}}
- Migration merges separate EN/RU system configs into unified per-provider
configs; seeds create one config per provider with multi-locale slots
- Locale-aware dispatch with EN fallback in NotificationDispatcher
- Frontend locale tabs (EN/RU) on template config editor
- Fix tracking config cards not showing default provider icons
- Global provider filter, search palette, and various UX polish
Video size warnings:
- Add file_size field to ImmichAssetInfo from exifInfo.fileSizeInByte
- Expose per-target max_video_size (50 MB for Telegram, none for others)
- Compute has_oversized_videos and per-asset oversized flag in template context
- Default templates show warning only when videos actually exceed the limit
- Templates no longer hardcode Telegram-specific logic
Template autocomplete:
- New jinja-autocomplete.ts engine with contextual completions
- Top-level variables ({{ }}), asset/album fields (dot access in loops),
Jinja2 filters (|), block tags ({% %}), and loop.* special vars
- JinjaEditor accepts optional variables prop via CodeMirror Compartment
- Wired into template-configs and command-template-configs pages
Also: fix template emoji (📷 → 📎) and sync sample_context with new vars.
- Added desc text to all 40+ grid items (EN + RU)
- compact prop on all IconGridSelect in compact form sections
- Fixed compact width to fill grid cells (removed width:auto)
- Replaced <select> filter dropdowns with IconGridSelect on config pages
- Replaced <select> provider filters with EntitySelect on tracker pages
- Dashboard filters constrained to fixed widths (not full row)
- Added filtering to command-template-configs and providers pages
- providerTypeFilterItems() with "All" option for filter contexts
Chat language:
- Added language_code field to TelegramChat model + migration
- Saved from message.from.language_code on webhook/polling
- Displayed as badge on bot chat cards and target receiver items
- Resolved from DB in target API response (works for existing receivers)
- Shown in chat picker dropdown (desc includes language)
EntitySelect improvements:
- Tracker-target link selector shows all targets, already-linked ones
appear disabled with "Already linked" hint
- Receiver chat picker shows already-added chats as disabled
Dev scripts:
- scripts/restart-backend.sh and restart-frontend.sh
- Updated .claude/docs/dev-servers.md to reference scripts
chat_action was stored on NotificationTarget model but never injected
into the config dict passed to the dispatcher. Now injected in both
watcher and webhook handler, and read by the dispatcher to pass to
TelegramClient.send_notification().
Any transform (even transform:none) in animation keyframes with
fill-mode creates a containing block that traps position:fixed
overlays. Removed transform entirely — fade-in only with opacity.
The fadeSlideIn animation used transform:translateY which creates a
new containing block, trapping position:fixed children (EntitySelect).
Switched to the CSS translate property which doesn't create a
containing block.
The translateY(-2px) transform on Card hover created a new containing
block, trapping position:fixed EntitySelect overlays inside the Card
instead of rendering relative to the viewport.
alexei.dolgolyov