- Moved the `programme_index` from the `EPGSource` model to a new `EPGSourceIndex` table, ensuring that the large JSON blob is only loaded when explicitly accessed, thus improving query performance and memory efficiency.
- Updated related queries and API views to utilize the new structure, including adjustments to EPG generation and import logic to prevent unnecessary data loading.
- Enhanced memory management in the EPG grid endpoint to reduce worker RSS during response handling.
- Streamlined `generate_epg()` to incrementally stream EPG data without loading the entire guide, improving memory efficiency.
- Reduced database load by deferring the fetching of large `programme_index` blobs, enhancing response times for EPG generation.
- Introduced a composite index on `(epg_id, id)` in `ProgramData` to optimize query performance during EPG exports.
- Updated tests to ensure proper functionality and performance of new features.
- Fix NameError in sd_lineups_search: add dispatcharr_version import
- Fix DELETE handler: persist changesRemaining from SD response to custom_properties
- Fix lockout reset time: calculate next midnight UTC (00:00Z) instead of
rolling now+24h, matching SD's documented reset behavior
- Remove redundant explicit index from SDProgramMD5: unique_together already
creates this index in Postgres
- Rename api_key to password (RemoveField + AddField) for clarity and reusability
- Move sd_changes_remaining/sd_changes_reset_at to custom_properties JSONField
instead of nullable columns on all EPGSource rows
- Replace sd_program_md5 column on ProgramData with SDProgramMD5 relation table
keyed by epg_source + program_id, fixing the key mismatch bug in program
MD5 delta detection
- Fix total_changed counter: .items() -> .values() so len() returns date count
not tuple length
- Add missing SD lineup management endpoints to api_views.py
(sd_lineups GET/POST/DELETE, sd_lineups_search POST)
- Regenerate migration 0023_schedules_direct using Django to ensure correct
index names and schema
- Add EPGSource username field and SDScheduleMD5 model for MD5 delta sync
- Implement full Schedules Direct fetch with SHA1 auth, token caching,
20-day schedule window, and MD5-based delta sync to minimize API usage
- Add lineup manager UI allowing users to search and manage SD lineups
- Add stations-only fetch on source creation for immediate EPG matching
- Enforce 2-hour minimum refresh interval to respect SD rate limits,
with first-fetch bypass when no prior full refresh exists
- Add real-time progress updates via synchronous Redis WebSocket delivery,
bypassing gevent.spawn for reliable delivery from Celery prefork workers
- Refactor EPGsTable status cell into standalone EPGStatusCell component
with direct Zustand subscription for live progress bar updates
- Guard XMLTV parse trigger for channels linked to SD EPG sources
- Add 15 backend tests covering auth, credentials, signals, and serializers
Migrations: 0023 (username field), 0024 (api_key help_text), 0025 (SD fields + SDScheduleMD5)
This enhancement introduces a powerful custom dummy EPG system that allows users to generate EPG programs on-demand by parsing channel or stream names using configurable regex patterns.
Key Features:
- Custom Pattern Matching: Define regex patterns to extract information from channel/stream names (teams, leagues, times, dates, etc.)
- Flexible Name Source: Choose to parse either the channel name or a specific stream name (by index)
- Timezone-Aware Scheduling: Automatic DST handling using pytz timezone names (e.g., 'US/Eastern', 'Europe/London')
- Time Format Support: Parse both 12-hour (AM/PM) and 24-hour time formats
- Date Parsing: Extract dates from names with flexible month/day/year patterns
- Custom Templates: Format EPG titles and descriptions using captured groups with {placeholder} syntax
- Upcoming/Ended Customization: Define custom titles and descriptions for programs before and after scheduled events
- Live Preview: Test patterns and templates in real-time with sample input
- Smart Program Generation: Automatically creates "Upcoming" and "Ended" programs around scheduled events
Use Cases:
- Sports channels with event details in stream names (e.g., "NHL 01: Bruins VS Leafs @ 8:00PM ET")
- Movie channels with genre/title/year information
- Racing events with driver/track/series details
- Any scenario where EPG data is embedded in channel/stream naming conventions
Technical Implementation:
- Backend: Pattern matching engine with timezone conversion and program scheduling logic
- Frontend: Interactive form with validation, pattern testing, and visual group preview
- Name Source Options: Parse from channel name or selectable stream index (1-based)
- Fallback Behavior: Uses standard dummy EPG if patterns don't match
- Custom Properties: Stores all configuration in EPGSource.custom_properties JSON field
Configuration Options:
- Title Pattern: Extract primary information (required)
- Time Pattern: Extract hour/minute/AM-PM (optional)
- Date Pattern: Extract month/day/year (optional)
- Timezone: Event timezone with automatic DST support
- Program Duration: Length of generated programs in minutes
- Title Template: Format EPG title using captured groups
- Description Template: Format EPG description using captured groups
- Upcoming Title Template: Custom title for programs before event starts (optional)
- Upcoming Description Template: Custom description for programs before event starts (optional)
- Ended Title Template: Custom title for programs after event ends (optional)
- Ended Description Template: Custom description for programs after event ends (optional)
- Name Source: Channel name or stream name
- Stream Index: Which stream to use when parsing stream names (1, 2, 3, etc.)
Closes#293
