The idea is to use torrent statistics to define which media (movies, series or other) is in trend.

The first step is to initialize a torrent database. Then, retrieve the statistics (completed, leechers, seeders) and store them. Finally, define which torrents are trending thanks to its statistics.

Each torrent has a status (new, follow, unfollow). These statuses allow you to continue collecting stats or not.

Statistics are collected for torrents with a new and follow status.

By default, a torrent has the status new for three days. It changes to the follow status if it is trending at the end of the new period.

A simple clustering on torrent names allows to attach torrents to a movie (or a series - in dev)

• Python >= 3.6
• Linux OS (debian, ubuntu)
• set MEDIASTRENDS_MODE global environment variable
• create configuration file to put in <project>/mediastrends.MODE.ini

> mediastrends
usage: mediastrends [-h] [-v] [-f CONFIG_DIR] [-m MODE]
                    {database,torrents,movies} ...

CLI to interact with mediastrends'tasks 

The logic steps to update database are:
    1. torrents add
    2. torrents stats
    3. torrents trends compute
    4. torrents status
    5. torrents trends get

optional arguments:
  -h, --help            show this help message and exit
  -v, --verbose         Set mediastrend logger level, 0 to 5.
  -f CONFIG_DIR, --config-dir CONFIG_DIR
                        Configuration directory. Load mediastrends.MODE.ini
  -m MODE, --mode MODE  Mode. Override MEDIASTRENDS_MODE environment

Top level commands:
  {database,torrents,movies}
    database            Commands on database
    torrents            Commands on torrent
    movies              Commands on movies

pip install git+https://github.com/prise6/medias-trends@master

User have to override some values of default config:

Content of mediastrends.ini (default config)

[directory]
base=
data=
logs=${base}/logs
sqlite=

[db]
database=sqlite

[sqlite]
path=${directory:sqlite}/database.db
backup_dir=${directory:data}

[indexers]
indexer_file=

[trackers]
tracker_file=

[requests]
user_agent=
batch_size=50

[retry]
tries=10
delay=5

[torrents_manager]
; new_delay=3
; delta_hours=1
; min_date=YYYY-mm-dd HH:MM
; max_date=YYYY-mm-dd HH:MM
; candidates_status=new,follow

[trends_manager]
; min_date=YYYY-mm-dd HH:MM
; max_date=YYYY-mm-dd HH:MM
; delta_days=31
; torrents_status=new,follow
; tau=0.2
; max_trendings=50
; weight_seeders=0.4
; weight_completed=0.4
; weight_leechers=0.2
; lambda=0.8

[jackettapi]
scheme=http
domain=jackett
port=9117
netloc=${domain}:${port}
path=api/v2.0/indexers/{indexer}/results/torznab/api
apikey=

user config file have to be named with this pattern: mediastrends.MODE.ini, MODE is the value of MEDIASTRENDS_MODE global var.

mediastrends look for user config in these directory:

1. $MEDIASTRENDS_DIRCONF if set
2. . current directory
3. ~ home directory
4. /etc/mediastrends directory

This file lists indexer website to collect new torrents.

fill the filepath of this file in config file:

[indexers]
indexer_file=

Example of content of the file:

indexers:
  indexer_1: # /!\ must be a tracker known by jackett
    cat_1:
      active: True
      action: search
      params:
        cat: 2000
    cat_2:
      active: True
      action: search
      params:
        cat: 5000
  indexer_2:
    movies:
      active: True
      action: search

This file lists bittorrent tracker to collect statistics from torrents.

fill the filepath of this file in config file:

# ...
[trackers]
tracker_file=
# ...

Example of content of the file:

trackers:
  tracker_1:
    active: True
    scheme: http
    netloc: domain:port
    path: 
  
  tracker_2:
    active: True
    scheme: udp
    netloc: domain:port
    path: 

Full pipeline only working with movies for now

# add torrents to DB
mediastrends torrents add -c movies -i indexer_1
mediastrends torrents add -c movies -i indexer_2
# collect statistics for new and followed torrents
mediastrends torrents stats -c movies -t tracker_1
mediastrends torrents stats -c movies -t tracker_2
# compute if new or follow torrents is still trending
mediastrends torrents trends compute -c movies
# compute new statuses based on new data
mediastrends torrents status -c movies
# get trending torrents
mediastrends torrents trends get -c movies
# compute movies trends based on torrents data
mediastrends movies compute
# get trending movies
mediastrends movies get

These methods need to have the entire repo (not only the package).

For movie pipeline

# reset old runs
make clean
# get trending movies
make get_movie_trends
# see Makefile file for more details

Docker image of this project: prise6/mediastrends

Use latest version, available for amd64, armv7.

This image will use an entrypoint to make mediastrends use easier. See dockerhub page for more details.

To-do

• fix score (size colum was wrong)
• add trends tests
• I recommand to truncate ptrends table and set status=1 for torrents

• resolve issue #12 (wrong score order)
• new score logic with parameter to tune in config

• resolve issue #9
• add fields to movies
• execute scripts/migrate_db_v0.1.1_to_v0.1.2.py to update database fields

• first version

One main goal of this project is learning:

• more advanced topics in Python:
  • linting (flake8)
  • unittest and mocking
  • typing
  • packaging
  • cli with argparse
  • peewee ORM
• github action (see .github/package.yaml)
• airflow
• bittorent protocol
• vscode + remote container dev with python

• implements a publisher package (create static website)
• Develop serie pipeline
• Use airflow instead of Make
• ...

Old run

          Fantasy Island
Year:2020
id:0983946
Nombre de torrents:2

          Dangerous Lies
Year:2020
id:10183816
Nombre de torrents:3

          Rich in Love
Year:2020
id:10329566
Nombre de torrents:1

          Rendez-vous chez les Malawas
Year:2019
id:10537978
Nombre de torrents:1

          Dreamkatcher
Year:2020
id:10553210
Nombre de torrents:1

GNU GPLv3

This project use mainly peewee under MIT license, jackett under GNU GPLv2.