107th Monthly Technical Session
Hi there! I would like to share about MTS(Monthly Technical Session) 107th.
It’s a small internal tech conference, and we share our knowledge with other Hennge members.
Let’s jump in!
Glamorous TUIs
The first presentation was by Camille, and it was about “Glamorous TUIs”.
She did some demonstrations of some TUI tools she had played with, which was interesting to see.
Surprisingly, she showed fireworks with the tool Confetty in the terminal in the beginning.
The next demo she did was the tool “draw.” She actually drew lines and also typed some words into the terminal.
She also showed how to build a simple terminal app using Golang-based tools called Bubble Tea and Lip Gloss. The first one is to set the logic, and the second one is for the styling. We saw the simple timer app and the code base she coded.
Finally, the Kanban board appeared in the terminal. It’s the tool called Kan-cli. It was really interesting to see some TUI libraries since I haven’t discovered that kind of thing a lot.
4/4 Incident Enabling TTL on DynamoDB
The next presenter was Hans, and he talked about an incident that occurred previously in his team.
In April, he tried to enable TTL (Time To Live) feature on his team’s DynamoDB and experienced some incidents.
TTL is a feature provided by AWS DynamoDB to check the timestamp and delete old items automatically. After he enabled the feature, suddenly, the app’s login feature was dead 😱
The issue was caused because DynamoDB consumed the provisioned capacity when deleting the outdated items and the table started rejecting all the write requests to the table.
However, that wasn’t what he expected. He knew it wouldn’t consume the capacity when deleting items with the TTL feature.
But the thing is that his team was using the global tables feature and there was a replicated table, and for those tables, AWS consumes the capacity.
He eventually turned the feature off and re-enabled it after he deleted the records.
HENNGE One Business Process Overview
Toshi’s talk was about our product, HENNGE One. In the beginning, he pointed out some points that can be improved.
We manage our customers’ information via our internal portal, and we operate updates from there manually.
Every time we change customers’ data we have to be careful since it might cause a big problem for their business. To avoid incidents, we double-check in the online meeting.
It’s necessary for us to do so in the current situation to operate reliable products, but Toshi said we can consider letting customers update some of the updates from the “customer portal” that they can use.
However, he emphasized that it’s just an idea right now, and we are going to decide whether we’ll do it or not by comparing it with other tasks’ priority.
Looking for something new to watch?
The next presenter was our intern, Chentduren. He talked about recommender systems used in some video streaming services, e.g. Netfilx and the main focus about the method called “Collaborative filtering”.
“Collective filtering” is an idea to calculate users’ preferences by using features e.g. action, humor, romance, etc.
In the algorithm, it uses the matrix that holds feature data for each movie and feature data for each user and puts them together.
He showed an example (which I couldn’t understand 😅) and also referred to the Python library called fast.ai to easily implement that kind of system.
He demonstrated how to train the system and actually get some recommendations from the data collector.
Easy Tsugaruben for dummies
We also had a lightning talk by Gunji-san, originally from Aomori.
To prepare for the workation in Aomori, she lectured some dialect called “Tsugaruben”.
I’m pretty sure that Tsugaruben is one of the most difficult dialects in Japan.
Eating unripe fruits by tracking and trusting
Next comes our CEO, Ogura-san! His lightning talk was about how we can try out some AI-related technologies.
He came up with using a new proxy temporarily called “ManyAIProxy”.
It keeps the auth-related keys inside and provides them to SDKs.
By building this, our company may have usage quota or some other features to “track” while letting us use the AI APIs.
Conclusion
We want to thank all the speakers for the wonderful presentations!
Of course, we had a beer bash, as always, after the session. See you in the next MTS recap!
