Move
Exercise
Stand
Reminder to try: ffmpeg checking for media file integrity (the output is Google AI-generated, so verify)
Checks for corrupt stream:
ffmpeg -v error -i input.mp4 -f null -
Fast check (checks media wrapper only):
ffmpeg -v error -i input.mp4 -c copy -f null -
Trip Planner is complete! It is kind of pathetic that it took me 11 days to think and implement the backend (but I have been busy)
I was told that I could not merge different OneID (Ontario Clinical Viewer) accounts into one - I don't know if it's the bad service (they are all assigned to one email address + one account can hold more than one pharmacy) or technical incapability (dumb), but I don't have the energy to probe further. This + Shoppers' Workday implementation = terrible, terrible enterprise solutions.
Playing around with models available for two Hugging Face tasks:
- Question-Answering - apparently this is no longer available in Hugging Face interface > v5
- Summarization
Some models were too large to be loaded into memory - I upgraded my Linode instance to include more memory and I think the result is better?
I think the planner is done!
- I spent two days thinking about data synchronization - how to detect the data insertion/deletion/corruption
- I was going to go for overkill and set up WebSocket - but going for a lot less efficient but simpler way - writing into a file
- The script fetches overall file every 3 seconds
- Very wasteful, not scalable
- The script posts to update file only when changes are made
- Hypothetically, if one submits with outdated data, then some newest change (less than 3 seconds old) will be lost
- Statistically, this is unlikely
- The script fetches overall file every 3 seconds
- I was going to go for overkill and set up WebSocket - but going for a lot less efficient but simpler way - writing into a file
Testing tomorrow!
Autonomous Intern 2: Text It. Talk to It. It's Handled.
It actually is kind of cool - not the functionality, but the concept