xklb 1.21.2

xk library

xk media library

A wise philosopher once told me, "The future is autotainment".

Manage large media libraries. Similar to Plex but more minimalist. Primary usage is local filesystem but also supports some virtual constructs like tracking video playlists (eg. YouTube subscriptions) or daily browser tabs.

Required: ffmpeg

Recommended: mpv, fish, firefox

Install

Linux recommended but Windows setup instructions available.

pip install xklb

$ library
xk media library subcommands (v1.21.002)

local media:
  lb fsadd                 Create a local media database; Add folders
  lb fsupdate              Refresh database: add new files, mark deleted

  lb listen                Listen to local and online media
  lb watch                 Watch local and online media
  lb read                  Read books
  lb view                  View images

  lb bigdirs               Discover folders which take much room
  lb dedupe                Deduplicate local db files

  lb christen              Cleanse files by giving them a new name

online media:
  lb tubeadd               Create a tube database; Add playlists
  lb tubeupdate            Fetch new videos from saved playlists

  lb redditadd             Create a reddit database; Add subreddits
  lb redditupdate          Fetch new posts from saved subreddits

downloads:
  lb download              Download media
  lb block                 Prevent downloading specific URLs
  lb merge-dbs             Merge multiple SQLITE files
  lb merge-online-local    Merge local and online metadata

playback:
  lb now                   Print what is currently playing
  lb next                  Play next file
  lb stop                  Stop all playback
  lb pause                 Pause all playback

statistics:
  lb playlists             List added playlists
  lb dlstatus              Show download status

browser tabs:
  lb tabsadd               Create a tabs database; Add URLs
  lb tabs                  Open your tabs for the day
  lb surf                  Load browser tabs in a streaming way (stdin)

mining:
  lb reddit-selftext       db selftext external links -> db media table
  lb pushshift             Convert Pushshift jsonl.zstd -> reddit.db format (stdin)
  lb nfb-films             NFB Director links -> film links (stdin)
  lb nouns                 Unstructured text -> compound nouns (stdin)
  lb hnadd                 Create a hackernews database (this takes a few days)

Quick Start -- watch online media on your PC

wget https://github.com/chapmanjacobd/library/raw/main/examples/mealtime.tw.db
library watch mealtime.tw.db

Or hook into HackerNews videos

wget https://github.com/chapmanjacobd/hn_mining/raw/main/hackernews_only_direct.tw.db
library watch hackernews_only_direct.tw.db --random --ignore-errors

Quick Start -- listen to online media on a chromecast group

wget https://github.com/chapmanjacobd/library/raw/main/examples/music.tl.db
library listen music.tl.db -ct "House speakers"

Start -- local media

1. Extract Metadata

For thirty terabytes of video the initial scan takes about four hours to complete. After that, subsequent scans of the path (or any subpaths) are much quicker--only new files will be read by ffprobe.

library fsadd tv.db ./video/folder/

2. Watch / Listen from local files

library watch tv.db                           # the default post-action is to do nothing
library watch tv.db --post-action delete      # delete file after playing
library listen finalists.db --post-action=ask # ask whether to delete after playing

To stop playing press Ctrl+C in either the terminal or mpv

Start -- online media

1. Download Metadata

Download playlist and channel metadata. Break free of the YouTube algo~

library tubeadd educational.db https://www.youtube.com/c/BranchEducation/videos

And you can always add more later--even from different websites.

library tubeadd maker.db https://vimeo.com/terburg

To prevent mistakes the default configuration is to download metadata for only the most recent 20,000 videos per playlist/channel.

library tubeadd maker.db --dl-config playlistend=1000

Be aware that there are some YouTube Channels which have many items--for example the TEDx channel has about 180,000 videos. Some channels even have upwards of two million videos. More than you could likely watch in one sitting--maybe even one lifetime. On a high-speed connection (>500 Mbps), it can take up to five hours to download the metadata for 180,000 videos.

1a. Get new videos for saved playlists

Tubeupdate will go through the list of added playlists and fetch metadata for any videos not previously seen.

library tubeupdate

2. Watch / Listen from websites

library watch maker.db

To stop playing press Ctrl+C in either the terminal or mpv

Start -- tabs (visit websites on a schedule)

tabs is a way to organize your visits to URLs that you want to visit every once in a while.

If you want to track changes to websites over time there are better tools out there, like huginn, urlwatch, or changedetection.io.

The use-case of tabs are websites that you know are going to change: subreddits, games, or tools that you want to use for a few minutes daily, weekly, monthly, quarterly, or yearly.

1. Add your websites

library tabsadd --frequency monthly --category fun \
    https://old.reddit.com/r/Showerthoughts/top/?sort=top&t=month \
    https://old.reddit.com/r/RedditDayOf/top/?sort=top&t=month

2. Add library tabs to cron

library tabs is meant to run once per day. Here is how you would configure it with crontab:

45 9 * * * DISPLAY=:0 library tabs /home/my/tabs.db

Or with systemd:

tee ~/.config/systemd/user/tabs.service
[Unit]
Description=xklb daily browser tabs

[Service]
Type=simple
RemainAfterExit=no
Environment="DISPLAY=:0"
ExecStart="/usr/bin/fish" "-c" "lb tabs /home/xk/lb/tabs.db"

tee ~/.config/systemd/user/tabs.timer
[Unit]
Description=xklb daily browser tabs timer

[Timer]
Persistent=yes
OnCalendar=*-*-* 9:58

[Install]
WantedBy=timers.target

systemctl --user daemon-reload
systemctl --user enable --now tabs.service

You can also invoke tabs manually:

library tabs -L 1  # open one tab

Things to know.db

When the database file path is not specified, video.db will be created / used.

library fsadd ./tv/

The same for audio: audio.db will be created / used.

library fsadd --audio ./music/

Likewise, fs.db from:

library fsadd --filesystem /any/path/

If you want to specify more than one directory you need to mention the db file explicitly.

library fsadd --filesystem one/
library fsadd --filesystem fs.db one/ two/

Organize via separate databases.

library fsadd --audio both.db ./audiobooks/ ./podcasts/
library fsadd --audio audiobooks.db ./audiobooks/
library fsadd --audio podcasts.db ./podcasts/ ./another/more/secret/podcasts_folder/

Usage

$ library watch -h
usage: library watch [database] [optional args]

Control playback:
    To stop playback press Ctrl-C in either the terminal or mpv

    Create global shortcuts in your desktop environment by sending commands to mpv_socket:
    echo 'playlist-next force' | socat - /tmp/mpv_socket

If not specified, watch will try to read video.db in the working directory:
    library watch
    library watch ./my/other/database/is-a/db.db

Override the default player (mpv):
    library does a lot of things to try to automatically use your preferred media player
    but if it doesn't guess right you can make it explicit:
    library watch --player "vlc --vlc-opts"

Cast to chromecast groups:
    library watch --cast --cast-to "Office pair"
    library watch -ct "Office pair"  # equivalent
    If you don't know the exact name of your chromecast group run `catt scan`

Play media in order (similarly named episodes):
    library watch --play-in-order
    There are multiple strictness levels of --play-in-order:
    library watch -O   # slow, more complex algorithm
    library watch -OO  # above, plus ignores most filters
    library watch -OOO # above, plus ignores include/exclude filter during ordinal search

Play recent partially-watched videos (requires mpv history):
    library watch --partial       # play newest first
    library watch --partial old   # play oldest first
    library watch -P o            # equivalent
    library watch -P fo           # use file creation time instead of modified time

    library watch -P p            # sort by progress / duration
    library watch -P s            # skip partially watched (only show unseen)

Print instead of play:
    library watch --print --limit 10  # print the next 10 files
    library watch -p -L 10  # print the next 10 files
    library watch -p  # this will print _all_ the media. be cautious about `-p` on an unfiltered set

    Printing modes
    library watch -p    # print in a table
    library watch -p p  # equivalent
    library watch -p a  # print an aggregate report
    library watch -p f  # print fields -- useful for piping paths to utilities like xargs or GNU Parallel

    Check if you have downloaded something before
    library watch -u duration -p -s 'title'

    Print an aggregate report of deleted media
    library watch -w time_deleted!=0 -p=a
    ╒═══════════╤══════════════╤═════════╤═════════╕
    │ path      │ duration     │ size    │   count │
    ╞═══════════╪══════════════╪═════════╪═════════╡
    │ Aggregate │ 14 days, 23  │ 50.6 GB │   29058 │
    │           │ hours and 42 │         │         │
    │           │ minutes      │         │         │
    ╘═══════════╧══════════════╧═════════╧═════════╛
    Total duration: 14 days, 23 hours and 42 minutes

    Print an aggregate report of media that has no duration information (likely corrupt media)
    library watch -w 'duration is null' -p=a

    Print a list of filenames which have below 1280px resolution
    library watch -w 'width<1280' -p=f

    Print media you have partially viewed with mpv
    library watch -p=v

    View how much time you have watched
    library watch -w play_count'>'0 -p=a

    See how much video you have
    library watch video.db -p=a
    ╒═══════════╤═════════╤═════════╤═════════╕
    │ path      │   hours │ size    │   count │
    ╞═══════════╪═════════╪═════════╪═════════╡
    │ Aggregate │  145769 │ 37.6 TB │  439939 │
    ╘═══════════╧═════════╧═════════╧═════════╛
    Total duration: 16 years, 7 months, 19 days, 17 hours and 25 minutes

    View all the columns
    library watch -p -L 1 --cols '*'

    Open ipython with all of your media
    library watch -vv -p --cols '*'
    ipdb> len(media)
    462219

Set the play queue size:
    By default the play queue is 120--long enough that you likely have not noticed
    but short enough that the program is snappy.

    If you want everything in your play queue you can use the aid of infinity.
    Pick your poison (these all do effectively the same thing):
    library watch -L inf
    library watch -l inf
    library watch --queue inf
    library watch -L 99999999999999999999999

    You may also want to restrict the play queue.
    For example, when you only want 1000 random files:
    library watch -u random -L 1000

Offset the play queue:
    You can also offset the queue. For example if you want to skip one or ten media:
    library watch -S 10  # offset ten from the top of an ordered query

Repeat
    library watch                  # listen to 120 random songs (DEFAULT_PLAY_QUEUE)
    library watch --limit 5        # listen to FIVE songs
    library watch -l inf -u random # listen to random songs indefinitely
    library watch -s infinite      # listen to songs from the band infinite

Constrain media by search:
    Audio files have many tags to readily search through so metadata like artist,
    album, and even mood are included in search.
    Video files have less consistent metadata and so only paths are included in search.
    library watch --include happy  # only matches will be included
    library watch -s happy         # equivalent
    library watch --exclude sad    # matches will be excluded
    library watch -E sad           # equivalent

    Search only the path column
    library watch -O -s 'path : mad max'
    library watch -O -s 'path : "mad max"' # add "quotes" to be more strict

    Double spaces are parsed as one space
    -s '  ost'        # will match OST and not ghost
    -s toy story      # will match '/folder/toy/something/story.mp3'
    -s 'toy  story'   # will match more strictly '/folder/toy story.mp3'

Constrain media by arbitrary SQL expressions:
    library watch --where audio_count = 2  # media which have two audio tracks
    library watch -w "language = 'eng'"    # media which have an English language tag
                                                (this could be audio _or_ subtitle)
    library watch -w subtitle_count=0      # media that doesn't have subtitles

Constrain media to duration (in minutes):
    library watch --duration 20
    library watch -d 6  # 6 mins ±10 percent (ie. between 5 and 7 mins)
    library watch -d-6  # less than 6 mins
    library watch -d+6  # more than 6 mins

    Duration can be specified multiple times:
    library watch -d+5 -d-7  # should be similar to -d 6

    If you want exact time use `where`
    library watch --where 'duration=6*60'

Constrain media to file size (in megabytes):
    library watch --size 20
    library watch -z 6  # 6 MB ±10 percent (ie. between 5 and 7 MB)
    library watch -z-6  # less than 6 MB
    library watch -z+6  # more than 6 MB

Constrain media by time_created / time_played / time_deleted / time_modified:
    library watch --created-within '3 days'
    library watch --created-before '3 years'

Constrain media by throughput:
    Bitrate information is not explicitly saved.
    You can use file size and duration as a proxy for throughput:
    library watch -w 'size/duration<50000'

Constrain media to portrait orientation video:
    library watch --portrait
    library watch -w 'width<height' # equivalent

Constrain media to duration of videos which match any size constraints:
    library watch --duration-from-size +700 -u 'duration desc, size desc'

Constrain media to online-media or local-media:
    Not to be confused with only local-media which is not "offline" (ie. one HDD disconnected)
    library watch --online-media-only
    library watch --online-media-only -i  # and ignore playback errors (ie. YouTube video deleted)
    library watch --local-media-only

Specify media play order:
    library watch --sort duration   # play shortest media first
    library watch -u duration desc  # play longest media first
    You can use multiple SQL ORDER BY expressions
    library watch -u 'subtitle_count > 0 desc' # play media that has at least one subtitle first

Post-actions -- choose what to do after playing:
    library watch --post-action delete  # delete file after playing
    library watch -k ask  # ask after each whether to keep or delete

    library watch -k askkeep  # ask after each whether to move to a keep folder or delete
    The default location of the keep folder is ./keep/ (relative to the played media file)
    You can change this by explicitly setting an *absolute* `keep-dir` path:
    library watch -k askkeep --keep-dir /home/my/music/keep/

Experimental options:
    Duration to play (in seconds) while changing the channel
    library watch --interdimensional-cable 40
    library watch -4dtv 40

    Playback multiple files at once
    library watch --multiple-playback    # one per display; or two if only one display detected
    library watch --multiple-playback 4  # play four media at once, divide by available screens
    library watch -m 4 --screen-name eDP # play four media at once on specific screen
    library watch -m 4 --loop --crop     # play four cropped videos on a loop
    library watch -m 4 --hstack          # use hstack style

You can pipe stuff

lowcharts

$ lb watch -p f -col time_created | lowcharts timehist -w 80
Matches: 445183.
Each ∎ represents a count of 1896
[2022-04-13 03:16:05] [151689] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
[2022-04-19 07:59:37] [ 16093] ∎∎∎∎∎∎∎∎
[2022-04-25 12:43:09] [ 12019] ∎∎∎∎∎∎
[2022-05-01 17:26:41] [ 48817] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
[2022-05-07 22:10:14] [ 36259] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
[2022-05-14 02:53:46] [  3942] ∎∎
[2022-05-20 07:37:18] [  2371] ∎
[2022-05-26 12:20:50] [   517]
[2022-06-01 17:04:23] [  4845] ∎∎
[2022-06-07 21:47:55] [  2340] ∎
[2022-06-14 02:31:27] [   563]
[2022-06-20 07:14:59] [ 13836] ∎∎∎∎∎∎∎
[2022-06-26 11:58:32] [  1905] ∎
[2022-07-02 16:42:04] [  1269]
[2022-07-08 21:25:36] [  3062] ∎
[2022-07-15 02:09:08] [  9192] ∎∎∎∎
[2022-07-21 06:52:41] [ 11955] ∎∎∎∎∎∎
[2022-07-27 11:36:13] [ 50938] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
[2022-08-02 16:19:45] [ 70973] ∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎∎
[2022-08-08 21:03:17] [  2598] ∎

$ lb watch -p f -col time_deleted -w time_deleted'>'0 | lowcharts timehist -w 80  # for some cols like time_deleted you'll need to specify a where clause so they aren't filtered out

mnamer

To rename poorly named files I recommend mnamer

pip install mnamer
mnamer --movie-directory ~/d/70_Now_Watching/ --episode-directory ~/d/70_Now_Watching/ \
    --no-overwrite -b (library watch -p fd -s 'path : McCloud')
library fsadd ~/d/70_Now_Watching/

rsync

I use rsync to move files instead of copy-on-write duplication because I want deletions to stick.

function mrmusic
    rsync -a --remove-source-files --files-from=(
        library lt ~/lb/audio.db -s /mnt/d/80_Now_Listening/ -p f \
        --moved /mnt/d/80_Now_Listening/ /mnt/d/ | psub
    ) /mnt/d/80_Now_Listening/ /mnt/d/

    rsync -a --remove-source-files --files-from=(
        library lt ~/lb/audio.db -w play_count=0 -u random -L 1200 -p f \
        --moved /mnt/d/ /mnt/d/80_Now_Listening/ | psub
    ) /mnt/d/ /mnt/d/80_Now_Listening/
end

Datasette

Explore library databases in your browser

pip install datasette
datasette tv.db

TODOs (PRs welcome)

• create/update views: especially dl queue, etc so that it is easier for people to access data externally db.create_view("items", ITEM_VIEW_DEF, replace=True)
• SELECT * FROM playlists p WHERE PATH NOT IN (SELECT DISTINCT playlist_path FROM media WHERE playlist_path NOT NULL )