Updates Feb 24

PGrid

• Minor fixes to gpu sales discord notification logic to not double notify of the same item
• added a toggle for GPU and CPU list to show prices for out of stock items to get an idea of market for low availability items
• Added feature flags to control stopping emails based on price decrease percent and high value items
• Fixed issue with some workers getting stuck on job never being able to start a new job. Added timeout that will kill the worker process after 60 mins.
• Wired up the ssd classification ai reviewing with the frontend so the review updates and reasoning is visible
• Did a little more performance updates related to all the price snapshots. Added a new archive gpu price snapshot table where price snapshots older than 2 days would go. This makes the current table more active and reduced build times from 3 mins to 2.5 mins for the frontend.
• New CPU Added: 5700X3D
• Reorganised ssd classification pipeline by splitting up the main file into different files, deleting old versions of prompts and functions. Original file was over 1k lines
• New shadcn based frontend for the internal app. Wanting to have react based internal app instead of ai generated html directly in the server code for both maintenance and to have a more complicated internal app tracking worker progress.
• New workers page with early version of task history
• Further investigating issues with some workers timing out at 60 minutes, most of them seem be from the generic item search on ebay which was different to other jobs as it reused a browser instance between scraping different pages. Tried adding an error handler which would restart browsers on errors but that didn’t work so removed the usage of shared browsers for now.
• Added a UI to verify ssd specs + some updates to specs search

Brisbane Support

• More updates, added switching from list to map view after clicking on an address in mobile view. added ability to locate the user. Fixed some marker display issues. Launched.