Robo Reminder
This project - a continuation of HW 10 AWS Unit 13 - and even though it hasn't been that long, feels like forever ago! How does this all work again? Yikes.
This project is to support the process of a new accounting employee who has daily tasks that she doesn't want to miss on any day. So she has requested this robo-reminder, which she can run towards the end of the day, to make sure she hasn't missed anything.
The more I think about it, the ongoing utility becomes clear. It's changing in my imagination to more of an ongoing process.
There is always a need to prioritize any free time amongst competing demands. It would be interesting to have a robo-reminder (RR) do that as possible.
Here's how that would work (maybe a subsequent activity after this project iteration): I would launch the RR when I have a period of time free. It would ask some questions to determine tasks necessary and what the rest of the day looks like, and then it would present a triaged list of tasks for that period. And possibly a b-list for the later period that day as well.
(Images are an area I've mostly avoided up to now, so trying to get more comfortable with them. Perhaps would use this next set of images as choices in a section of the bot dialogue)
So I'm thinking to consider both sorts of functionalities.
Another set of functionalities would be the remind-other-staff set -- maybe would be a whole different robot though.. or a robot that informed the main reminder-robot. Just spitballing..
Task List - Accounting/Finance employee of business
[] Coding of all transactions for day reviewed, any corrections communicated and confirmed
[] Costs per product line items updated (weekly)
[] In-kind donations booked to expense and donation, so costs included in strategic planning
[] Cash Flow forecast prepared for tomorrow /n
[] Dashboard metrics prepared for tomorrow
Task List - Accounting/Finance contractor hired by entrepreneur to help with their personal and business finance
[] Year End Cycle - Pay Pal 1099 form - reconcile to internal records
[] Review employee transit card balances, load in additional funds equal to prior quarter or more if usage warrants.
[] Monthly Cycle - Sales Tax review of any internet purchases from small vendors - record transactions for reporting
Logic/Decision tree
To really complete this and make it work well, the lambda program would be a bit complex. It would take in the answers to the questions about what needed to be done today still; and based on the date, it would draw from it's internal programming regarding other tasks. It would also incorporate the amount of time available during this 'free time' session, and the percentage likelihood of there being a second 'free time' period later in the day.
From all of that info, it would instantly provide a list of 3-5 tasks that were most important in the current period.
I don't know that I'll get that all done any time soon, but it seems like an instance of the journey being more important than the destination.
Questions
I will be noting questions along the way, to check off as I discover things during this process, or for future continued explorations.
[] Is there a built-in time function within Lex or Lambda or something, so that my Robo-Reminder will always already know the date? I will need for it to know if it's week one of the month vs. week two vs. week three, and hopefully the day of the week as well. If there is not that built in, then I would need to have that be part of the dialogue.
Technical Notes
Region:
I am continuing to work in the 'Oregon' region, as we were told that one had the set of functionalities that was the best fit for our activities - and am continuing with those same activities.
Bucket setup (for slot images.. I think..):
[] Versioning - enabled I'm worried that this might be one of those things that trips into more costs, but sounds like could be useful to keep multiple versions and not lose anything accidentally. So I'll try it, and just closely monitor the costs dashboard.
[] Tags - used for tracking costs more closely.. Since I live in accounting land, this sounds like a really good idea to me, but from the documentation scan I did, there's a lot to it. I think I'll have to hold off for now.
[] Encryption - also sounds like a great value-add, also more in-depth than I have the appetite for, so skipping for now (leaving as disabled).
Amazon Lex:
Aspects of Amazon Lex that I'll be working with include:
Bot -- the visible outcome that interacts with the user to achieve a goal
Intent Name -- the request that the bot user is making to the bot (naming the goal)
Slot -- these are the details of the request that the user is making
Slot type -- Description of the type of variable that is being used for each slot - numeric, date, string, etc.. If it is a prebuilt Amazon type, then it will do validation inherent in that prebuilt functionality. If you tell it the slot is Amazon.number, and then during testing you provide an answer for that slot that is not a number, it will just ask again until you provide a number. Same with Date or Time. However if the Slot type is a custom type, then you need to also provide values for that and/or have it be variable.
Sample Utterances -- the wording that the user may use in expressing their intent
Lambda program -- the connective tissue between the sample utterances, and the resulting actions of the bot. We used a python variant of lambda, noting that it was quite a different animal than the python language that we had been used to. I'm saving that part for when I've become more familiarized.
The three types of functionality that the lambda function can provide include:
- Initialization (dialog code hook) [*] Validation (dialog code hook) [~] Fulfillment (fulfillment code hook)
Validation will come in to play in terms of timing of tasks, and confirming what day it is, when the various deadlines are, etc..
Response Cards -- This is a visual cue that prompts the user to make a choice between specified options. That was part of the original assignment, and I had trouble with it at the time. I really wanted to get that part down this time around. So am trying to do that, and it just is eluding me. The one thing says to click on the gear to the right of the intent - no gear. It references the editor in the left bar - no editor. I'm seeing helpful text sections within AWS, reading and saving those, it's just still not lining up for me. But I am learning that the message capabilities of the bot are quite extensive.
I did just set up a new intent from the beginning, in case I missed something during my earlier two attempts. Again, I am not seeing any gear icon next to my intent, anywhere, in order to add in a response card. All I can think at this point is that it's an AWS change, or it is within the role definition step (outside of the instructions provided originally, and not addressed in the AWS example). Drat.
Example Bot
I'm making use of the example bot to get back in the swing of all this - very helpful! Much appreciated. Working through that reminds me that the bot (using the data in the slots only) replies the way it thinks it is supposed to, regardless of the text it is receiving. It doesn't actually recognize or process the actual content provided at all. The correct terminology includes validating the data -- meaning that the lambda function will make sure the input text fits some pre-built text pattern that has been provided to it. Also initialization is involved - I'm less clear on what that refers to in this context. I guess like initializing variables, which are then modified in the process (as applicable)?..
The functionality required to have the bot respond specifically to various text content aspects is the lambda functionality. I'm working my way towards reviewing that again soon.
So, seems amazing, but the example flower-ordering bot does work (ignoring content).. seems like so little structure is provided yet it works. One new feature I've caught up with via this bot is the 'conversation flow' option - clicking on that arrow shows the steps within the intent/answer next intent/next answer through to fulfillment. Excellent!
Slots
The slots in the left bar are the more detailed constructions of the custom slots that are within the intent.
Termination:
My notes from that first AWS homework/class period indicated that it was optimal to terminate services, but that's just too much work right now. Hoping that my usage will be minimal enough that I can skip that step and still not incur any costs. So far A-okay! From final day presentations: some classmates utilized AWS, and reviewed the termination steps - very helpful! For the storage item, will have to empty it and then delete it. So far am not using a notebook. They didn't cover the bot, so will have to grapple with that myself when the time comes (ha ha).
Suspending process for now, and terminations
I am going to step away from this all for now, have started review of all things Python which interests me more even at this point. Also my next step in this reminder-bot process seems to be to delve deeper into the whole permissions functionality of AWS, and right now at work I'm immersed in permissions functionality within the software in use there, and I don't want to have dueling permissions frameworks in my head. So releasing this all for now.
But I'm thinking to leave my existing bots in place for now, to check billing regularly, but I feel like they might not be as cost-involved as other elements. Will also try and review my materials about that. But I've got stuff in place with those etc.. just would be easier to be able to pop back in at will.
Storage bins - will empty (gah) and delete (gah) those though, as I think more costs might be involved with those. And all I had done was put things in, so will just need to put things in again, not a specific unique detail to recreate or anything.
AWS Interface
Every time I start up with AWS again, even after just a day or two, it's like 'what is all of this?', 'where are my things??' etc etc..I decided not to terminate services so that I could start up again more easily each session, but even so.. Long story short - when I search for 'AWS Lex' the result is a screen has 'Amazon Lex' in a big splash screen, and it says 'get started'.. and I'm like - I've already started though?! So originally it was very perplexing. But now I just go ahead and 'New Start' or whatever the wording is, then I get to my existing bot(s). Whew!
I was able to go into my bucket that I'd set up, add a folder for images, and upload the images that I have in this readme and others - to use as slot images in the future. That part really confused me the first time around - getting those assigned, and working. Hopefully better this time (when I get around to it).
So now I've got a bucket with images in it, and two bots - the beginning of my reminder bot, and an example bot about buying flowers. I have the two bots in two different detached web windows, to contrast and compare. I have slots in the flower bot, but not in my reminder bot yet.
Class Presentation
I had vaguely been thinking I could perhaps do a short presentation for my classmates on my AWS work, particularly I was thinking just to refresh the basics of bots and in particular what can be accomplished simply via the intent/slot functionality, vs what requires a lambda hook. But my process ended up not being in ordnung (in order) sufficiently to demonstrate that. I mean with the flower shop one I could have, but I really wanted was focused on the response card functionality, and not having that... also the class time was quickly running out, and so on.
Acknowledgements
I would like to first acknowledge the guidance and teaching of our FinTech Boot Camp Instructor, Garth Mortensen, our T.A. Roberto Salazar Reyna; our Student Success Manager, Angelica Baraona; and the whole team behind the curriculum, the logistics, the whole program. I also found the collective Stack Overflow wisdom essential as ever. In addition, I'm starting to work through the 'AWS Certified Cloud Practitioner - Study Guide -Foundational (CLF-C01) Exam' text by Sybex (Wiley). And I have the 518 page Amazon Lex guide, so there's that.
Additional resources that have been provided or that I've come across in my process include the training resources from Deloitte's digital initiative, and these videos on AWS workings: AWS Free Tiers, SageMaker Setup, as well as AWS example tools and exercises.
I'm also more widely exploring readme formatting etc.. And also: I'm very glad I saved as many notes and path markers etc.. as I did at the time. At this point, I wish I had saved more of course. But every little bit helps.
This project is for the purpose of my own ongoing learning, it is not a grading instance.