jaideepkekre / medibot Goto Github PK

Since the message queuing architecture is now multi-process, it does not seem necessary to stilll have the dispatcher around. One process manages getting and queuing messages and the other processes them and sends them across.

Using an in-memory data store like Redis to store the messages of a conversation would be simpler and cleaner. The core class can simply read the the chat_id and pull the relevant conversation.

• http://blog.pythonisito.com/2012/07/introduction-to-gevent.html
• http://www.gevent.org/

Storing information on the patient will require another schema. Discuss here.

@v0dro You're inputs ?

@jaideepkekre list your thoughts about the json format to be used for DB storage here.

I'll put up a mock from my side in a short time.

Logs generated by dispatcher and core looks exceeding ugly and not very readable. @jaideepkekre clean them up.

Maybe use colors?

The NLP engine tokenizes and labels words based on their occurrence in the string and how previous conversations have proceeded.

It is important to formalize a format that will be passed to the expert system by the NLP engine which will easily convey the relevance of every word in a sentence to the expert system, and make KE easy.

As discussed, sending and getting messages should be based on queues. Current architecture limits max number of messages that can be processed in a finite amount of time.

1. Let's do this , each user gets a thread .

• Text/dict from user is kept in queue , the thread consumes text / dict and only consumes next text/dict after previous request has returned .

2. Let's do this , each NLP op gets a thread .

• each NLP op in core.py gets a thread as NLP will be the heaviest job in our code .

What's your opinion ?

@v0dro

• updater = Updater(token=os.environ.get('TELEGRAM_API_KEY'))
• dispatcher = updater.dispatcher

IMHO the custom variables "updater" , "dispatcher" are VERY similar to Updater() function , updater.distatcher . Won't it be a better convention to make it obvious what is custom and what is imported . You code isn't very readable like this , imagine this on a scale*100 .

./run.sh Traceback (most recent call last): File "/Users/rex/Downloads/MediBot-master/bin/server.py", line 11, in <module> from telegram import Updater ImportError: cannot import name 'Updater'

I've updated telegram but still unable to resolve this issue.