151–160 of 1001 posts
It was way back before gpt and twilio, so I bent one IRC bot and wrote collection of warm messages that was send via SMS on a pseudorandom timeframe. We broke up when she finds out ... Luckly for both of us I guess.
My townhome complex had one of those call boxes at the front gate. When Doordash/FedEx/the cleaners/the in-laws/etc arrived they would have to call me from the call box and I'd have to answer it and listen to garbled audio to figure out who it was and press 9 to open the gate. It was kind of a pain, so I made a Twilio app to answer calls from the call box. I set up custom entry codes that I could hand out to anyone.…
And an app that records audio retroactively. Wanted to have recorded the last 2 minutes? just tell it "-2m". It has a ring buffer that buffers the last 2h, but also forgets all the time. (at least that was the plan, but audio streams were so much more difficult so I just dumped a full day of audio and timestamps and cut it later. That makes it much more "eavesdropping" again though which I wanted to avoid with the auto delete though....)
I reverse engineered a flash application that showed a map and provided address search. I scraped the map tiles and address to location database. Reimplemented the viewer application as a Java applet and preloaded the tiles and address database to a microSD card connected to the phone. So essentially I built my own offline maps for my not internet connected phone.
Address search required prefix tree because IO was too slow to use binary search on the phone.
Anyway this was done just before I went to a new city to attend university and it was really helpful to find out where I am and where to go. There was no navigation, but it showed the map, gps location and the location where I needed to get to.
So that was my personal project that really had great utility for me.
- auto detect and convert column types
- save as a parquet file in a folder
- then autogenerate a sqlalchemy table/metadata file in python for all tables with sensible defaults for column types (e.g. 2x the longest string in a column for varchar)
- build the db and all tables
- load data from the files into the tables
this makes it really easy to bootstrap the entire db from a folder of parquet files for testing with sqlite and then makes it easy to move to prod on postgres/sqlserver etc. Before I go to prod i still have to add constraints and keys and indexes but that doesn't take too long. and for dev/testing the data's not too big so performance doesn't really suffer from lack of keys/constraints then we can use something like alembic on the big sqlalchemy tables definition file to do db migrations.
it's kind of like this: https://github.com/agronholm/sqlacodegen but solving an inverse problem.
basically it bootstraps the db and schemas and gets me like 95% of the way there. my quality of life is better with it.
I've since built a couple more for my other tanks, and I rewrote the firmware for ESP8266/ESPHome. Now my family can ask Alexa to feed the fish, because of course the world needs that.
