wrolpi/modules/videos/schema.py
Roland 52cf7a9520
Search Channels by name in the Other search tab
The Other tab only listed Channels with the selected Tags, so typing a
Channel's name found nothing.  Channels now match the search text by name,
and the Tags when there are both.  The list and its estimate share one query.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-07 15:37:07 -06:00

304 lines
6.3 KiB
Python

from dataclasses import dataclass, field
from typing import List, Optional
from modules.videos.video.lib import DEFAULT_VIDEO_ORDER, VIDEO_QUERY_LIMIT
from wrolpi.errors import ValidationError
@dataclass
class ChannelPostRequest:
name: str
directory: str
calculate_duration: Optional[bool] = None
download_missing_data: Optional[bool] = True
generate_posters: Optional[bool] = None
source_id: Optional[str] = None
tag_name: Optional[str] = None
url: Optional[str] = None
def __post_init__(self):
self.name = self.name.strip() or None
self.directory = self.directory.strip() or None
self.url = self.url.strip() if self.url else None
self.source_id = self.source_id.strip() if self.source_id else None
@dataclass
class ChannelPutRequest:
calculate_duration: Optional[bool] = None
directory: Optional[str] = None
download_missing_data: Optional[bool] = True
generate_posters: Optional[bool] = None
mkdir: Optional[bool] = None
name: Optional[str] = None
url: Optional[str] = None
def __post_init__(self):
self.name = self.name.strip() or None
self.directory = self.directory.strip() or None
self.url = self.url.strip() if self.url else None
@dataclass
class ChannelTagRequest:
tag_name: Optional[str] = None
directory: Optional[str] = None
def __post_init__(self):
if self.directory and self.directory.startswith('/'):
raise ValidationError('Directory must be relative to media directory')
@dataclass
class ChannelTagInfoRequest:
channel_id: Optional[int] = None
tag_name: Optional[str] = None
@dataclass
class ChannelDownloadRequest:
url: str
frequency: int
settings: dict = field(default_factory=dict)
def __post_init__(self):
self.url = self.url.strip()
if not self.url:
raise ValidationError('url cannot be empty')
# Validate settings contents. Remove empty values.
from wrolpi.schema import DownloadSettings
self.settings = {k: v for k, v in DownloadSettings(**self.settings).__dict__.items() if v not in ([], None)}
@dataclass
class RefreshRequest:
channel_name: str
@dataclass
class ChannelModel:
id: int
url: str
name: str
directory: str
@dataclass
class ChannelResponse:
channel = ChannelModel
@dataclass
class ChannelPostResponse:
success: str
@dataclass
class ChannelsChannelModel:
id: int
name: str
@dataclass
class ChannelsResponse:
channels: List[ChannelsChannelModel]
@dataclass
class StreamResponse:
success: str
stream_url: str
@dataclass
class Video:
id: int
description_path: str
ext: str
poster_path: str
source_id: str
title: str
video_path: str
name: str
caption_path: str
info_json_path: str
channel_id: int
@dataclass
class VideoWithChannel:
id: int
description_path: str
ext: str
poster_path: str
source_id: str
title: str
video_path: str
name: str
caption_path: str
info_json_path: str
channel_id: int
channel = ChannelModel
@dataclass
class ExtendedVideo(Video):
info_json: dict
@dataclass
class VideoResponse:
file_group = ExtendedVideo
prev = ExtendedVideo
next = ExtendedVideo
@dataclass
class VideoCommentsResponse:
comments: list[dict]
@dataclass
class VideoCaptionsResponse:
captions: list
@dataclass
class VideoDescriptionResponse:
description: Optional[str]
@dataclass
class VideoSearchRequest:
search_str: Optional[str] = None
tags: List[str] = field(default_factory=list)
offset: Optional[int] = None
limit: Optional[int] = VIDEO_QUERY_LIMIT
order_by: Optional[str] = DEFAULT_VIDEO_ORDER
channel_id: Optional[int] = None
tag_names: List[str] = field(default_factory=list)
headline: bool = False
censored: bool = False
deep: bool = False # When True, search all text including captions (d_text)
@dataclass
class VideoUpdateRequest:
# Only the fields given are changed.
title: Optional[str] = None
description: Optional[str] = None
@dataclass
class VideoTranscodeRequest:
video_codec: Optional[str] = None # A key of TRANSCODE_VIDEO_TARGETS, REMOVE_VIDEO, or None to copy the stream.
audio_codec: Optional[str] = None # A key of TRANSCODE_AUDIO_TARGETS, REMOVE_AUDIO, or None to copy the stream.
container: str = 'mp4'
fragmented: bool = False # Fragmented mp4: quick start/seek for very long videos (mp4/m4a only).
@dataclass
class VideoTranscodeResponse:
job_id: str
@dataclass
class VideoSearchResponse:
videos: List[VideoWithChannel]
tsquery: str
@dataclass
class PaginationQuery:
offset: int
@dataclass
class TagRequest:
video_id: int
tag_name: str
@dataclass
class WROLModeRequest:
enabled: bool
@dataclass
class VideoStatistics:
videos: int
week: int
month: int
year: int
sum_duration: int
sum_size: int
max_size: int
have_comments: int
missing_comments: int
failed_comments: int
@dataclass
class ChannelStatistics:
channels: int
tagged_channels: int
@dataclass
class VideosStatisticsResponse:
videos = VideoStatistics
channels = ChannelStatistics
@dataclass
class CensoredVideoRequest:
limit: int
offset: int
@dataclass
class CensoredVideoResponse:
videos: List[VideoWithChannel]
@dataclass
class ChannelSearchRequest:
tag_names: List[str] = field(default_factory=list)
search_str: Optional[str] = None
@dataclass
class VideoFileFormatRequest:
video_file_format: str
@dataclass
class CookiesUploadRequest:
"""Request to upload and encrypt cookies."""
cookies_content: str
password: str
def __post_init__(self):
if not self.cookies_content:
raise ValidationError('cookies_content cannot be empty')
if not self.password or len(self.password) < 8:
raise ValidationError('password must be at least 8 characters')
@dataclass
class CookiesUnlockRequest:
"""Request to unlock cookies with password."""
password: str
def __post_init__(self):
if not self.password:
raise ValidationError('password cannot be empty')
@dataclass
class CookiesStatusResponse:
"""Response with cookies status."""
cookies_exist: bool
cookies_unlocked: bool