A telegram pipeline helper, to help organise, prepare, and send videos for gif channels. It also handles tagging, and can post videos to twitter.
Installing should be as easy as:
poetry install
The application will then install and automatically update yt-dlp while running.
To run, use the command:
poetry run python ./main.py
Configuration is done via the config.json
file. This file is updated manually only. The application state is then stored in pipeline.sqlite
At the base level of the config file are these keys:
api_id
:int
, the telegram client API key, as obtained from https://my.telegram.orgapi_hash
:str
, the api hash to login to the user's telegram account who is running the application. This is required to allow the bot access to extra information about channel state not allowed by bots.pipeline_bot_token
:str
(optional), A bot token, as obtained from botfather, for the bot that will be sending messages in gif workshop groups. If not provided, the user account will be used.public_bot_token
:str
(optional), A bot token, as obtained from botfather, for the bot that will handle public queries, such as tagging and sourcing and searching, for subscribers to channels to use.channels
:list[Channel]
, A list of channel configurations, further detailed belowworkshop_groups
:list[Workshop]
, A list of workshop configurations, further detailed belowapi_keys
:dict
, A dictionary of API keys to various services, as detailed belowwebsite
:WebsiteConfig
, A dictionary of website configuration information, for the backend and frontend deployments
Each channel is a dictionary in the base channels
list. They have these keys:
handle
:str|int
, the telegram handle of the channel. A string if public channel, or the channel ID integer if it's a private channel. Does not need any "-100" prefix._note
:str
(optional), Not parsed, but can be used to clarify the purpose of channels, especially useful for private ones.queue
:Queue
(optional), queue configuration, see below for details. If provided the channel will use a group chat as a queue, to buffer a set of gifs before sending them.read_only
:boolean
(optional, default: False). If set to true, the channel will be considered read only. It will be indexed for duplicate detection, but will not be given as a destination for sendingsend_folder
:str
(optional), If provided, this channel will be in the folder with the given name when user requests to send a video (and has permission to access this channel). Folders can be nested by using a slash (/
). For example, providinganimals/mammals
will mean that this channel will appear in the "mammals" folder inside the "animals" folder, along with any other channels configured with the samesend_folder
.note_time
:boolean
(optional, default: False). If provided, and true, the send confirmation menu will note to the user when the last message was sent to the destination channel.tags
:Tags
(optional). If provided, this will provide configuration for suggested tags for videos sent to the given channel, see below for details.twitter
:Twitter
(optional). If provided, videos sent to this channel will also be sent to twitter. See below for detailscaption
:str
(optional), If provided, used as a jinja2 template string, which is passed a dictionary of tag names to lists of tag values, to format a caption for the destination.website
:Website
(optional), website configuration, see below for details. If provided, the channel will be available on the website.
Configuration for a channel queue. Must be a group chat, rather than another channel.
handle
:str|int
. The telegram handle of the queue. A string if public (which would be odd), or the chat ID if private. Should be a positive integer, not prefixed by 100duplicate_detection
:boolean
(optional, default: True). Whether to enable duplicate detection in this queue, checking whether videos here have been seen before, or whether videos elsewhere are here.schedule
:Schedule
(optional), if provided, gif pipeline will automatically post from the queue to the channel according to the schedule. It will prompt which video will be sent next, and when, and give you the option to enable automatically posting it at the specified time.
Provided to a queue, this will cause it to prompt to automatically post from the queue to the channel according to this schedule.
min_time
:str
, an iso8601 duration string, the minimum time between automatically prompted posts in the channel.max_time
:str
(optional), an iso8601 duration string which, if provided, will serve as the maximum time between posts in the channel, and the automatic prompt will pick a random time between the minimum time and maximum time.order
:str
(optional, default: "random"). Can be one of:random
,oldest_first
,newest_first
, and will decide whether the automatic posting helper will suggest posting the oldest video in the queue first, the newest video, or will select a video at random.target_queue_length
:str
(optional), an iso8601 duration string which, if provided, changes the behaviour of max and min time, and instead of being a random time between those bounds, will cause the delay to be such the number of posts in the queue should aim to last as long as the target length.schedule_variability_percent
:int
(optional, default: 15), If provided along with a target queue length, this will be how much randomness to add to the calculated delay required to make the queue reach the target length.
The tag configuration is the suggested tags for videos being posted to a given channel. The keys of the dictionary should be the names of the tags, and then the values should be a dictionary with the following keys, or an empty dictionary to default to defaults:
type
(str, default: "normal"), Can be one of:normal
,single
,text
, orgnostic
.type=normal
: A list of all the tag values which have been seen before will be presented to the user, who can then select whichever are appropriate, which will be stored for the video.type=single
: A list of all the tag values which have been seen before will be presented to the user, who can then select one of the listed tags which will be stored for the video.type=text
: The user will be prompted to enter a custom text value for the tagtype=gnostic
: A list of all the tag values which have been seen before will be presented to the user, who can then select which tags apply, and which do not. On saving, the list of which apply and the list of which are not selected, will both be saved, which means future tag values can be added, and will not automatically not-apply to previous videos. By default, asource
tag will be added to videos by download helpers, which will contain the link the video was originally downloaded from.
If this is provided, videos sent to this channel will also be posted to twitter. To use this functionality, you must also provide "consumer_key" and "consumer_secret" in the API keys configuration section for twitter.
account
:TwitterAccount
. The account tokens for the twitter account to post this video.text
:str
(optional, default is empty). The string to use as a caption for the tweet. These are provided as jinja template strings, withtags
variable provided as a dictionary of tag names to lists of tag values. e.g. A "text" value of "Source: {{ tags['source'] | first }}" will result in a caption "Source: https://..." for your given source URL.reply
:dict
(optional). If this is provided, it is another twitter configuration object, with information for automatically posting another tweet as a reply to the original one. The reply will not have a video, so thetext
field is no longer optional. Thereply
field is available, and optional, in case you want to make a longer chain of tweets. Theaccount
field is also available, and is optional, so that the reply can be posted from a separate twitter account. If theaccount
field is not provided, the reply will be posted by the same account as the tweet it is replying to.
A dictionary containing the account access token and access secret for the given twitter account. These are generated via oauth, and can be generated by running the get_twitter_tokens.py
script in the scripts
directory.
access_token
:str
The access token for the account posting videos on twitteraccess_secret
:str
The access token secret for the account posting videos on twitter
Configuration for how the channel will be (or won't be) available on the website
enabled
:boolean
(optional, default: "true") Defaults false if no website configuration is set, but if any website configuration is provided, this will default to false. Whether the channel has a page available on the websitepublicly_listed
:boolean
(optional, default: "false") Whether to list the channel on the index home page.
Workshops are group chats where a user can interact with the gif pipeline to edit and process videos and links. Permissions are handled on the telegram side, meaning anyone in one of these groups can use the functionality of the bot to download videos or edit them. (When sending to channels however, the users' permissions are checked for each available channel)
handle
: (str|int
) The telegram handle for the workshop group. Should be the public username (if it's a public groupchat) or an integer chat ID, which should be positive and without "100" prefix._note
:str
(optional), Not parsed, but can be used to clarify the purpose of workshops, especially useful for private groupchats, with otherwise only have a numeric ID in the config file.duplicate_detection
: (boolean
) Whether to enable duplicate detection notifications (and video hashing) for the workshopdefault_destination
: (str|int
) (optional), If set, is the telegram handle for the default channel videos sent from this workshop will go to.ß
This section stores various API keys or other details for third party services. Generally used by helpers. Most helpers should check for the presence of the API keys they require before attempting to support those services.
reddit
: API access details for reddit, a social media site. Details are generally obtained at reddit's API preferences pagereddit.client_id
: (str
) This is the client_id for the reddit APIreddit.client_secret
: (str
) This is the client_secret for the reddit APIreddit.owner_username
: (str
) This is your reddit username, which is provided in the user agent of requests to reddit's API
twitter
: Details for twitter subscriptions, as well as for posting to twitter accountstwitter.nitter_url
: (str
) This is used by the twitter subscription type. It should be the URL of a nitter instance with RSS enabled.twitter.consumer_key
: (str
) Used by twitter posting, consumer_key as required byscripts/get_twitter_token.py
. Generated on the twitter developer portal site. This guide can also be helpful.twitter.consumer_secret
: (str
) Used by twitter posting, consumer_secret as required byscripts/get_twitter_token.py
in the same was as the consumer_key above.
instagram
instagram.bibliogram_url
: (str
) This us used by the instagram subscription type. It should be the URL of a bibliogram instance with RSS enabled.
frigate
: Details for a Frigate cctv webserverfrigate.url
: The base web URL for the frigate webserver, including protocol and port
This section stores configuration values for the backend API and the frontend website.
backend_port
:int
(optional, default: 3000) Which port the backend API should be listening on for connectionsbackend_url
:str
(optional, default: "http://localhost:3000") The URL that the frontend should use to connect to the backendfrontend_port
:int
(optional, default: 3100) Which port the frontend should be listening on for connections
The pipeline has many "helpers", which are classes which handle different types of user requests in workshop groups. These are used to edit videos, and manage tags, and such. As a general rule, commands should be posted as a reply to the video they are referring to, and will then reply to the command with their results.
Converts a video to audio. Use the command audio
to convert the video to mp3 (for music), or use the command voice
to send the audio as a voice note.
Adds a text caption to a video message (for easier forwarding, mostly). Use the command caption
with your caption, replying to the video you wish to add a message caption to.
Just sets the source tag for any videos forwarded from a public channel.
Takes commands of the form: chart {destination} {tag_name}
where {destination}
is the handle of a telegram channel the bot is managing, and {tag_name}
is the name of a tag field. It will then generate a pie chart of from the frequency distribution of all the values of that tag in that channel.
Cuts a video into regularly sized chunks with commands of the form chunk {duration}
. The duration can either be given as a number (iterpreted as a number of seconds), or an iso8601 duration.
Takes commands of the form: delete family
or delete branch
, as a reply to another message. This helper checks that the user has telegram permissions to delete things in this chat.
If delete branch
is specified, it will delete the message the comamnd is replying to, as well as any messages which are replies to that, or replies to that, all the way down. This includes the command message.
If delete family
is specified, it will delete the message the command is replying to, any messages those messages are replying to, all the way up, and anything replying to any of those messages, all the way down. This includes the command message.
Scans messages for links, and will attempt to download them using the Youtube Downloader tool. Skips certain links that specialised handlers can handle, such as fa and msg handlers. If the link is to a playlist, it will only download the first video of the list.
Attempts to check videos for whether duplicates have been seen elsewhere in the system, whether in channels or workshops. When a new video is posted, it will automatically be split into frames (at 5 fps), and those images will be hashed with dHash from the python imagehash library. If there are any matches, a notice will be posted as a reply to the video.
You can also reply to a video with check
and it will check that video and post a reply with the results.
A specific handler for downloading and processing gif files from the furaffinity website.
When a command is sent as a reply to a video, this helper uses ffprobe to get information about the video.
ffprobe
orstats
will return the full ffprobe information dump for the video.duration
will return the video duration, in seconds.resolution
orsize
will return the video resolution in pixels.
This handler will attempt to find a video in a playlist, which matches the video which the user is replying to.
Use find {playlist link}
and it will search the playlist for a matching video, and respond with the video and link.
This handler is to export and download video clips from a Frigate web server (see https://frigate.video).
Use frigate {camera_name} {start_time} {end_time}
to specify which camera clip to export.
Has no specific commands, but handles all menus in the application. (Such as those generated by send helper, scene split helper, schedule helper, tag helper, etc.) Some menus can also be replied to, such as tagging menus, and this helper will handle those messages.
Merges or joins videos together. If videos are not the same resolution, videos will be scaled and black bars will be padded to make all videos the same resolution as the first video in the sequence. This helper takes 3 different commands
merge {link...}
Can be sent as a reply to a video, or with a list of video message links, and will combine those videos into one video, with the reply message or the first link in the list being first, and the other links in the list in order.append {link...}
Sent as a reply to a video with a list of video message links, and will combine those linked videos onto the end of the reply target video.prepend {link...}
Sent as a reply to a video with a list of video message links, and will combine those linked videos onto the start of the reply target video.
A specific handler for downloading and processing gif and webm submissions on the e621 website.
Takes commands of the form: reverse
, and will reverse the video the command is replying to.
Takes commands of the form: split scenes {threshold}
where {threshold}
is an optional argument, defaulting to 30.
Uses the PySceneDetect library to detect the scenes in a video, and then splits the video up into those scenes. It will give a menu prompt saying how many scenes the video will be split into before it actually splits the video and posts the scenes.
Use a higher threshold to make it split into fewer scenes, and a lower threshold to make it split into more scenes.
Handles channel queue scheduling, and automatic posting. Will ensure every non-empty queue with a scheule will have a menu asking whether to post a specified video into the channel at a specified time.
If a queue becomes empty, a message will be posted that the queue is empty.
Also takes commands of the form schedule
, as a reply to a video in a queue, to schedule that video as the next one to send to the channel.
Takes commands of the form send
to present a menu to the user asking which channel to send the specified video to. Will only suggest channels that the user has telegram permissions to post in.
If there are any configured tags for the channel, it will prompt the user to set those tags.
If there is a queue configured for the channel, it will ask the user whether to send straight to the channel, or whether to send to the queue.
Takes commands of the form stabilise
, (stabilize
, stab
, unshake
, deshake
all also accepted) and will attempt to stabilise the video using ffmpeg's "deshake" video filter
Manages subscriptions. There are various types of subscriptions for different sites. All subscriptions are checked once an hour, and results, or errors, are posted to the chat they were created in. Takes commands of the form:
sub list
: Lists subscriptions active in the current chatsub {link}
: Creates a subscription to the specified link. Subscription handlers will attempt to determine which handler can handle the linksub remove {link}
: Removes the subscription to the specified link. The available types of subscription are listed below
Subscription to a given instagram account. This subscription type is only available if instagram.bibliogram_url
is specified in the API keys configuration.
Subscription to a given subreddit. This subscription type is only available if reddit API keys have been added to configuration.
Subscription to an RSS feed. Entries on the feed will attempt to download using youtube DL
Subscriptions to a telegram caht or channel. Can be a public channel, or any chat that the user who is running gif pipeline is in. The chat will be checked every hour, and any videos, gifs, or links to videos/gifs will be posted to the subscription target.
Subscription to a twitter feed. Uses nitter rss feeds, and hence twitter.nitter_url
must be configured in the API key configuration for this subscription type to be enabled. Feed links can be suffixed with /media
to only check tweets with media, or /with_replies
to check tweets which are posted as replies.
This subscription exists for startup purposes. As youtube-dl support can change with time, any subscriptions which do not seem to be able to be handled when gif pipeline starts up, will be loaded as unitialised subscriptions, and will post an error every hour until they are removed, or they can find a subscription type that supports them.
This subscription type uses youtube-dl to download the feed link as a playlist, and checks that playlist for new items, posting any new items in the subscription target chat.
Handles tagging of videos. All tags are handled as a set of values for a given tag name. Takes replies to videos with commands of the form:
tags
: Posts all the tags and tag values for the videotag {tag_name}
: Posts all the values the video has for a specific tag nametag remove {tag_name}
: Removes all values for a specific tag nametag {tag_name} {tag_value}
: Adds a tag value to the video for the specified tag nametag remove {tag_name} {tag_value}
: Removes the given value from the video for the specified tag name
Converts a video into a telegram "gif", which means removing the audio track and ensuring it's an .mp4 under 8MB and under 1280px wide/tall. It also defaults to ensuring video is 30fps, as 60fps video plays slow on older phones. Additional arguments can be provided, for example:
720x720
would ensure the video is no more than720
pixels wide and720
pixels tall. These numbers are used as a bounding box, and aspect ratio is preserved. This defaults to1280x1280
- bitrate can be specified with an argument ending with
bps
orb/s
. This bitrate will be the target for the video.mbps
andkbps
can also be specified. This will default to the highest it can be for the video to fit in 8MB - framerate can be specified with an argument ending with
fps
. This defaults to 30fps
Takes an argument of the form update youtube downloader
(and synonyms for yt-dl) and will download a new version of yt-dlp directly from the github and install it for future downloads and subscriptions.
Crops a video. All measurements are given in percentages. Arguments are given as keyword value pairs, for example, crop left 30 top 20
. Available keywords are:
left
orl
. Crops a specified percentage of the total width of the video from the left side of the videoright
orr
. Crops a specified percentage of the total width of the video from the right side of the videotop
ort
. Crops a specified percentage of the height of the video from the top side of the videobottom
orb
. Crops a specified percentage of the height of the video from the bottom of the videowidth
orw
. Crops the width of a video down to the specified percentage of the current width. Cannot be specified withleft
orright
height
orh
. Crops the height of a video down to the specified percentage of the current height. Cannot be specified withtop
orbottom
Cuts the length of a video. Takes a command of the form cut {start} {end}
where start and end are timestamps in seconds. User can also use the strings "start" to specify the start of the video, and "end" to specify the end. For example, cut start 5
to get the first 5 seconds of the video, or cut 15 end
to get the video from 15 seconds in, to the end.
Commands can also be given as cut out {start} {end}
which will then cut the video from the start to the specified start timestamp, and join that to the video from the specified end timestamp to the end. Effectively cutting the specified time range out of the video.
Takes the same arguments as the telegram gif helper, but does not remove the audio track. Takes commands of the form video
Rotates or flips video. Takes commands of the form rotate {direction}
or flip {axis}
.
Available rotation directions are: clockwise
, anticlockwise
, and 180
. There are aliases such as right
and left
.
Available flip axes are: horizontal
and vertical
Speeds up video. Takes commands of the form speed {speed}x
.
To slow down a video, a fraction can be provided, such as speed 1/2x
.
If a zip file is sent to a workshop, this helper will check the zip file for any video or .gif files, and turn them into videos and post them. This can be useful for importing .gif files into telegram without the telegram client mangling them into a small resolution version.
There are also public helpers, which are helpers that channel subscribers can use with the public bot instance, to request information on gifs from the channels.
If a message is forwarded from one of the channels monitored by gif pipeline, this helper will tell you the tags for that video.