You should first install watchme. This will place the executable watchme in your bin folder, which is the client for setting up and running a watcher. You can jump into setting up your first watcher below. If you are interested in using the watchme decorator to monitor resource usage of a function, see the psutils page.




Data and Sharing

or read about the following:

What is a Watcher?

A watcher is a configuration to check a resource like a website at some frequency. We do this with a combination of cron jobs (scheduling) and git repositories (version control). Your watchers will live by default in subfolders of $HOME/.watchme, and it’s easiest to keep this default. If you want to change the default, you can do this with environment variables.

For example, here is what a set of three watchers might look like in your watchme home.


What goes in each folder? It’s actually a git repository, and this helps to track changes for pages. We do this instead of saving duplicates of the pages, and if you are doing research, you have a GitHub repository that updates with each.

Setup WatchMe

The first thing you need to do is an initial setup of WatchMe, which will generate your WatchMe home folder and a global configuration.

$ watchme init --help

usage: watchme init [-h] [--empty] [--watcher WATCHER] [--base BASE]

optional arguments:
  -h, --help         show this help message and exit
  --empty            don't create the default watcher folder
  --watcher WATCHER  the watcher to create (defaults to watcher)
  --base BASE        the watcher base (defaults to $HOME/.watchme)

Notice that you can change the default base, turn off generation of the default watcher (named watcher) with --empty or provide a custom name (instead of watcher). Let’s use all the defaults and just run the init:

$ watchme init
Creating /home/vanessa/.watchme...
Adding watcher /home/vanessa/.watchme/watcher...
Generating watcher config /home/vanessa/.watchme/watcher/watchme.cfg

Notice how it created a default watcher called “watcher?” To take that off, do:

$ watchme init --empty
Creating /home/vanessa/.watchme...

and just the base directory will be created. Note that you can change the root of this with –base:

$ watchme --base /home/vanessa/Desktop/.watchme init --empty
Creating /home/vanessa/Desktop/.watchme...

Notice how the --base argument comes before init? That’s because it’s a global argument, meaning you can supply it to any command group (init, create, etc.). If you don’t want to set it and want to change the default, remember that you can export WATCHME_BASE_DIR in your bash profile for a permanent setting. For the reminder of this tutorial, we will assume that you ran init without --empty, and interact with a watcher named “watcher.”

How do I create a watcher?

If you didn’t create a watcher above (meaning you ran the init command with --empty) this is ok! You can create as many watchers as you like with the “create” command.

$ watchme create --help
usage: watchme create [-h] [watchers [watchers ...]]

positional arguments:
  watchers    watchers to create (default: single watcher)

optional arguments:
  -h, --help  show this help message and exit

As an example (after running watchme init) let’s create a watcher called “weather”

$ watchme create weather
Adding watcher /home/vanessa/.watchme/weather...
Generating watcher config /home/vanessa/.watchme/weather/watchme.cfg

You can then see an empty watcher directory is created with a configuration file:

$ tree /home/vanessa/.watchme/
└── weather
    └── watchme.cfg

and what you can’t see is that there is a Git repo (a .git folder) in the watcher directory too. Good job! The watcher is set up, and ready to add tasks to it.

How do I rename a watcher?

The watcher name coincides with the folder name. For example, if you created the “weather” watcher and want to change the name at a later point, just rename the folder. For the remainder of this getting started guide, we will be using a watcher named “watcher,” so if you didn’t create this watcher, you can either create it or rename an existing folder.

Monitor Tasks on The Fly

If you dont’ want to use WatchMe for tasks, you can use it to monitor on the fly! This could be a simple wrapper to any command:

$ watchme monitor singularity pull docker://busybox
[{"status": "running", "username": "sochat1", "cmdline": ["singularity", "pull", "docker://busybox"], "cpu_num": 29, "num_fds": 6, "create_time": 1659736689.01, "open_files": 0, "ionice": {"value": 4, "ioclass": "IOPRIO_CLASS_NONE"}, "name": "singularity", "cpu_affinity": [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47], "exe": "/usr/bin/singularity", "cpu_percent": 0.0, "num_threads": 11, "io_counters": {"read_count": 21, "write_count": 0, "read_bytes": 0, "write_bytes": 0, "read_chars": 8671, "write_chars": 0}, "memory_full_info": {"rss": 24010752, "vms": 1356939264, "shared": 11788288, "text": 26202112, "lib": 0, "data": 1300967424, "dirty": 0, "uss": 23040000, "pss": 23040000, "swap": 0}, "num_ctx_switches": {"voluntary": 84, "involuntary": 1}, "terminal": "/dev/pts/53", "uids": {"real": 13754, "effetive": 13754, "saved": 13754}, "ppid": 848657, "pid": 848682, "memory_percent": 0.011313591754217435, "cpu_times": [0.01, 0.01, 0.0, 0.0, 0.0], "gids": {"real": 5782, "effetive": 5782, "saved": 5782}, "nice": 0, "connections": [], "cwd": "/home/sochat1", "SECONDS": "3"}]

Or an entire analysis (e.g., running a container) and taking samples every second:

$ watchme monitor --seconds 1 singularity run watchme-mnist_latest.sif plots.png > mnist-external.json

For the interested reader, there is an entire write-up here.

How do I add tasks?

After creation, the watcher is empty. You need to add tasks for it to run. The configuration commands will vary based on the kind of task you want to add, and here is a quick example of adding a task to watch a url (the default task):

$ watchme add-task watcher task-singularity-release url@
url  =
active  = true
type  = urls
save_as = text

In the example above, we added a task called “task-singularity-release” to the default watcher “watcher.” The only required variable is the url, and we provided it in the format <key>@<value>. This is how you will add any parameter to a task, and the parameters allowed will vary based on the task type. We also specified that the page we are going to retrieve is save_as text, meaning that it’s not json (default). In the above, we didn’t define a --type variable. By default, the setting is --type url. After you add a task, you can quickly verify it was added by looking at the configuration file directly:

$ cat /home/vanessa/.watchme/watcher/watchme.cfg

active = false

url =
active = true
type = urls
save_as = text

While you don’t need to manually edit this file, there isn’t any reason that you can’t if you wanted to. To read more about the settings here, see the watcher configuration documentation.

The task is active by default (after you set up its schedule) and you can disable this with –active false:

$ watchme add-task watcher task-singularity-release url@ --active false

The reason we save these parameters in the repo is that if you put it under version control on GitHub (or similar), others will be able to reproduce your protocol.

How do I edit tasks?

After you’ve added a task, you can easily update parameters. The format is:

$ watchme edit <watcher> <action> <task> <key> <value> 

Where action can be one of “update” or “add” or “remove.” For example, add a value:

$ watchme edit watcher add task-harvard-hpc file_name file.txt
Adding file_name:file.txt to task-harvard-hpc

If you try adding again, it won’t let you because the value exists.

$ watchme edit watcher add task-harvard-hpc file_name file.txt
ERROR file_name already exists. Use "update" action to change.

Instead, use update:

$ watchme edit watcher update task-harvard-hpc file_name churro.txt
Updating file_name to churro.txt in task-harvard-hpc

and when you have had enough, remove a parameter entirely:

$ watchme edit watcher remove task-harvard-hpc file_name
Removing file_name

What are the parameters for each task?

The parameters will vary based on the task type. When you are ready, take a look at the watchers page to choose a task type that you want to configure.

How do I inspect my watcher?

If you don’t want to directly inspect the watcher configuration file, there are some easy functions you can run to show them on the screen. First, to inspect an entire watcher:

$ watchme inspect watcher
active  = false

url =
active = true
type = urls

You can also inspect a particular task alone:

$ watchme inspect watcher task-singularity-release
url =
active = true
type = urls

If the task doesn’t exist, it will tell you:

ERROR task-doesnt-exist is not a valid section.

How do I list my watchers?

You can quickly see the watchers installed to your watcher home with:

$ watchme list

You can also list task folders and other files for a particular watcher:

$ watchme list github
watcher: /tmp/tmp.QG2Tuusrrb/github

or finally, if you add a specific task folder, you can list the contents there.

$ watchme list github task-spython
task: /tmp/tmp.QG2Tuusrrb/github/task-spython

If you are interested in the configurations in watchme.cfg, then you can inspect further with inspect.

How do I protect or freeze my watcher?

If you want to prevent deletion of your folder, you can protect it.

$ watchme protect watcher on
active  = false
protected  = on

protection means that you cannot delete the watcher, but you can edit the tasks. To remove protection:

$ watchme protect watcher off
active  = false
protected  = off

Freezing is one step above protected - it means that you cannot delete or edit the tasks for your watcher. You can also turn it on or off:

$ watchme protect watcher freeze
active  = false
frozen  = on

$ watchme protect watcher unfreeze
active  = false

Frozen takes preference to protected, if the settings don’t agree.

How do I activate my watcher?

By default, when you add a watcher, it isn’t active. You can see this by looking at the watcher configuration file in the [watcher] section:

$ watchme inspect watcher
active  = false

To activate or deactivate the entire watcher, run these commands:

$ watchme activate watcher
[watcher|watcher] active: true
$ watchme deactivate watcher
[watcher|watcher] active: false

If you want to activate or deactivate single tasks, just add the task name:

$ watchme deactivate watcher task-singularity-release

By default, when a task is added, it is active.

How do I run a watcher?

Typically, you would schedule your watcher with a cron job, and then forget about it. However, you are also able to run the watcher manually, if you so desire (to test or otherwise.) To run manually:

# watchme run <watcher>
$ watchme run watcher

or to run as a test (not saving data):

$ watchme run watcher --test

where “watcher” corresponds with the watcher name. When this is done, if the watcher has status “active” as true, all the associated tasks that are also active will be run, and the repository updated with results. If you want to show variables and names of the tasks instead of the default progress bar, do:

$ watchme run watcher --no-progress

You can also choose to run in serial, also with or without a progress bar:

$ watchme run watcher --no-progress --serial

After you run your watcher tasks, you will see a directory created named based on the task, along with a result file (if successful) and a TIMESTAMP file to indicate when the task was last run.

$ tree
├── task-harvard-hpc
│   ├── result.txt
└── watchme.cfg

You can also confirm that the watcher has run and commit results to the repo. Here we are in the watcher base folder, at $HOME/.watchme/watcher. We just ran the watcher, and there were two tasks:

git log
commit c8929ea7caf71a4172195b618ff602f9b47686e1
Author: Vanessa Sochat <>
Date:   Mon Apr 1 13:07:47 2019 -0400

    watchme watcher ADD results task-harvard-hpc

commit 95517ff8d4d39e326be5e9638969f385fbe05355
Author: Vanessa Sochat <>
Date:   Mon Apr 1 13:07:47 2019 -0400

    watchme watcher ADD results task-singularity-release

We can see results were added for both.

How do I schedule my watcher?

Now that you’ve configured your watcher, you simply need to schedule it to run. WatchMe uses cron to run scheduled jobs, but will handle the configuration for you. For example, here is how to schedule your watcher to run at different standard time intervals:

#                  <name> <frequency>
$ watchme schedule watcher @hourly
$ watchme schedule watcher @daily
$ watchme schedule watcher @weekly
$ watchme schedule watcher @monthly
$ watchme schedule watcher @yearly

Notice how the times are prefixed with @ to distinguish between watchers and times? If you want to set a custom cron string you can do that too:

$ watchme schedule watcher 0 * * * * *

Here is how to schedule your watcher to run:

$ watchme schedule air-quality @hourly

And then check that an entry has been added to crontab:

$ crontab -l
@hourly watchme run air-quality # watchme-air-quality

You can check to see if cron is running (and the job ran):

# ubuntu
$ service cron status

# linux
$ service crond status

How do I remove a task from a watcher?

To remove a task, just specify it:

$ watchme remove watcher task-reddit-hpc
task-reddit-hpc removed successfully.

If the task doesn’t exist, it will tell you.

$ watchme remove watcher task-reddit-hpc
WARNING task-reddit-hpc does not exist.

If the watcher is frozen, it will not allow you to remove it.

$ watchme remove watcher task-reddit-hpc
ERROR watcher is frozen, unfreeze first.

How do I delete a watcher?

To remove an entire watcher, specify --delete:

$ watchme remove watcher --delete

If the watcher is frozen, you need to unfreeze it first:

$ watchme remove  watcher --delete
ERROR watcher watcher is frozen, unfreeze to delete.

$ watchme protect watcher unfreeze
$ watchme remove  watcher --delete
Removing watcher watcher

The entire folder is now removed.

$ ls /home/vanessa/.watchme/

How do I get a watcher?

Not only can you share your watcher (configuration and data output) on a service like GitHub thanks to version control, you can also easily grab someone else’s watcher repository!

$ watchme get special-name
Added watcher watchme-air-quality

Conflrm that it was added:

$ watchme list

You can also clone (and name it something else):

$ watchme get special-name
Added watcher special-name
$ watchme list

If you attempt to overwrite a folder that exists, you’ll get a warning.

vanessa@vanessa-ThinkPad-T460s:~/Documents/Dropbox/Code/Python/watchme$ watchme get special-name
ERROR /home/vanessa/.watchme/special-name exists. Use --force to overwrite

and need to use --force to remove first, and then clone the watcher to the same location.

$ watchme get special-name --force
Added watcher special-name

You will still need to schedule the water to run (and update the .git folder cloned there) and activate it, in the case that it wasn’t active in the repository.

How do I export data?

The core reason you would want to use WatchMe in the first place is to collect data over time for some task. Let’s say we have a result file, “oakland.txt” in the task folder “task-air-oakland” for the “air-quality” watcher. We can export that from the command line via:

# watchme export <watcher>   <task>           <filename>
$ watchme export air-quality task-air-oakland oakland.txt
git log --all --oneline --pretty=tformat:"%H" --grep "ADD results" cca3c4eb84d9c38527ec93a9a620bfab07d798f2..5d2b047dabe74e76e1585341bb956fd633bd0832 -- task-air-oakland/oakland.txt
    "commits": [
    "dates": [
        "2019-04-02 12:31:19 -0400"
    "content": [

By default, it will print the result to the screen, unless you provide a --out to print it to file. Also notice that the git commands to get the list of commits are shown. We only have one commit for the task file, but we can see that it has an associated date, and the content of the file. For more than one commit, each of the entries in the output dictionary would be a list.

Exporting Json

If your results files have json, you will get a funny looking unparsed json inside json. If you know this about your task, add the --json flag so the entire thing is parsed as a single (loadable) json file:

$ watchme export system task-cpu vanessa-thinkpad-t460s_vanessa.json  --json


This code is licensed under the Mozilla, version 2.0 or later LICENSE.