Only applies to chatting

Launch the bot and net async, so the bot can get online quickly independently of the network.

prints events by how many times events with a specific UID has been intercepted. I.e.:

User     | Events
SomeUser | 912
User2    | 132

applies on Discord too, use site-specific commands to deal with the async bs of getting usernames through Discord.py

This just exists to inform about why the eval command was removed and what's required for it to be re-added.

The problem:

[TRIGGER]eval while(true)print("infinite recursive loop")

Attempted solutions:

• Kotlin corroutines - cancel and cancelAndJoin doesn't actually stop the execution of the script
• Regular threads - does not work. stop, join, interrupt, etc doesn't actually stop it.

Assuming a timeout can be added, and it works, the eval command can be re-added. Until then, it's not an option to do so.

2000 char limit. Splitting is already happening (horribly though, that's another item on the TODO-list), but having the entire help list takes up a lot of space. [trigger]help <command name> exists, so help needs to be added to the commands too.

Add booleans in config.py to disable entire output sites.

I.e.:

useDiscord = True

disabling it disables booting the site and adding it to the site list (for privileges).

I noticed you have several loops to find the site desired.

https://github.com/LunarWatcher/NN-chatbot/blob/master/Commands.py
Lines 28 - 32, 38 - 42, etc.

What I think would help is that since the sites object is an array, we could create an index object. What would be done is making a new line after line 20 that reads: siteIndexs = {}

A new line would be appended after line 31 that reads: siteIndexs[site.name] = sites.indexOf(site)

That enables us to do the following:

instead of: for site in sites: #some verification method
we would use:
sites[siteIndexs[site.name]]

sites: list() = []

Throws a syntax error:

 sites: list() = []
      ^
SyntaxError: invalid syntax

For all the hard-coded commands (learned commands are an exception).

SE OpenID will be deprecated on July 25th. Most likely, this will break chat.SE login. From what I can tell, it looks like MSE and SO is unaffected as the login seems to not use OpenID, though I can't be entirely sure.

It's probable that at least one Stack Exchange chat site breaks, depending on the flow. This might work, but how much of the current login system is affected. There are currently no POST requests to the gifs mentioned, which is why it's possible that SO and MSE login will break too.

It's hard to debug until OpenID is removed completely, since there isn't a problem yet. But it's in less than 20 days, so it still needs fixing soon

Database autosave ends up wiping the data.

Idk why, doesn't work though

The ranks don't work on Discord for some reason. Could be a data storage issue

I think it would be useful if users could make their own vocab the first time before they make saves for the bot. This would require 2 things:

• A way to access the chatroom posts.
• A way to write to a file.

The first one can be solved by SO-chatbot by Zirak, combined with the modification of the console.log function.
The second one can be done by opening a new window and using document.write.

Command to restart the bot. Would make life easier. Most likely, this means adding a gradle task and executing it before calling system.exit, but this would be much easier with a core system where the bot itself is external from the process running it. Using Python for this may be an option, seems fairly easy too

Requests is vulnerable to CVE-2018-18074. IIRC, all the Python code using requests was removed. With the cleanup, there's a bit of code that apparently wasn't updated

Currently, the flow for events on SE is:

• Receives an event from the websocket
• Checks if the event has an r[roomid] (without brackets)`. Creates a new JsonNode
• Checks if the JsonNode created in the previous event has a key called e
  • If yes, moves to a bunch of if-statements with known, mapped, events
  • If no, returns.

While this works, it gets messy. In addition, it's not flexible. The goal should be:

• Receives an event from the websocket
• Parses it like the current flow (necessary to get to the event object)
• Gets the event_id
• Uses the event_id to find related callbacks (if any) and forwards the JsonNode from the parsing to them

This would be more flexible, and also enable the registration of multiple callbacks for different events.

In light of recent bugs (I'm talking to you #50), it would be a good idea to have a backup command. This command would be accessible to rank 8 users and up, and creates a separate file. This is why it's separate from the save command, as this command would create a new file, where as the save command saves into the existing one.

This only applies to SE and Discord. Twitch does not have message reactions, editing, deletion, starring, or something similar

Example of a learned command currently deployed in my instance:

name: charge
output: *charges @%s with %s*

The current usage is:

[TRIGGER]charge who, what

However, other commands wok differently and split by space instead. For consistency, learned command arguments should be:

[TRIGGER]charge who what

But this presents one problem: How are multi-word arguments handled? They should be handled like this:

[TRIGGER]charge who "with what"

Given GitHub doesn't support pull requests for the wiki repository, contributors need a workaround to submit and collaborate on documentation.

I found a few possible suggestions on the internet:

1. Including a github wiki in a repository as a submodule
2. Enabling pull requests on github wikis setup
3. Contributors clone the wiki; make changes to their version of the wiki repo; submit a link to their repo as an Issue; and request for it to be added to the main wiki.

There are quite a lot of them. There's not a list of them at the moment.

For now, it's the obvious stuff that needs to be fixed, like unnecessary calls to get fields, unnecessary iterating, and generally other inefficient calls or design.

Such command. Much fun. WOW!

Most of the API is a mess. It needs to be refactored, and specifically the Java module.

The neural net could use some improvements too, specifically on memory use. It might need to be re-written from scratch.

Save time - add CLI arguments to allow direct boot without prompting the user.

Necessary stuff:

• Parsing
• Set variables if the CLI argument for the relevant var isn't found
• Without getting in the way of the force training system

Relevant code:

Inside if __name__ == "__main__" in bot.py

don ' t you just hate reading messages like this ? Especially , since all the formatting is messed up by the bot and it ' s training . . .

Anyways, the "problem" (by design obviously) created here is (supposed to be) cleaned up here, but it does so in a horrible, non-functioning manner. It works perfectly fine on the first few replaces, but after that the poorly constructed replace statements don't clean up properly.
These need to be fixed

Please mark this as "enhancement".
Also please delay this until the project is finished.

Use any tricks and shortcuts you know to make bot.py smaller for quicker download size.
I know Python cannot be compressed exactly like JS (made into 1 line).
Stil it could be made a bit shorter.

prints whether or not the Flask server is online or not

Could be a version problem for that matter. When using Java 9, the bot crashes at the lombok compiler. As a result, the bot doesn't run.

Learned commands are very static at the moment. They currently only support arguments with %s or {id}, but that's where it stops. By adding functions (i.e. url encoding) the learned commands can be more complex than they currently can be

#54 escalated quickly.

• Clean up the multi-command Kotlin files
• Make the classes more clear
• Merge the site config classes
• Remove redundant classes
• Do API calls instead of storing usernames

Having one instance of every command for every site is a massive memory waste. It should be refactored to have a single instance, but each class still uses the existing functions (i.e. loadSE) to register that site as using a specific set of commands (aside the shared ones). This would also require refactoring with passing the site, but it shouldn't be a problem

I'm going to hate myself for this, but convert the main bot to Java/Kotlin and use Python for the backend

The documentation is a mess and a lot of it is non-existent. Documentation needs to be added as soon as possible

Current progress

• Setup documentation - thanks to Whispres
• Command/listener documentation
  • Extending existing ones
• Code documentation (Javadoc/Kdoc) - not very critical.
• Interaction guide
  • rank documentation

Notes

• Interaction guide

A basic "how to use the bot". Creating command and listener documentation would take care of a large part of this, but stuff like understanding arguments (currently a comma-separated list, #61 changes that a bit but still) for commands, also the CLI-style arguments (--somename "argument content").

Should also cover ranks

You should make it so that the user does not manually have to disable several lines. The lines could simply be wrapped in a if: statement. You could do this by adding these lines in the config.py file:

"useDiscord": "true",
"useStackExchange":"true"

SEChat is a mess, and a bunch of features are missing. In addition, there are a lot of bugged features.

I.e. //getRank Zoe says I'm banned, but //getRank 6296561 gives the appropriate rank

Prints the bot's invite link with instructions

I got most of them down, but when I try to install any of the tensor ones, it fails. And yes, I got python3.6

A script that installs the dependencies, if not present. TheOneWhoMade's suggestion

Reinforcement learning would allow the bot to learn while chatting. A large part of the network has to be re-written though, but it would improve it significantly