Updates Feb 3

PGrid

• Fixed bug: browser lifecycle bug with doing generic ebay searches
• New PGrid discord server to try build a community
• Created a discord bot that will post in stock status for 5080 and 5090 gpus
• added ability to ignore emails from email review page
• added feature flags with first flag being disabling email sending
• posted on reddit about new discord server and feature to get notified of stock. decent interest and good to have new users in the server
• fixed bug with new chromium browser not using vpn proxy
• fixed issue where if scrapers fail 3 times it sets all gpus from a source as failed
• updated gpu price snapshot cleanup function and set it to run everyday. forgot to set it up to run on workers and ran it once, removed over 1m rows.
• fixed issue with PLE showing a 5080 card available in the listings page but not available in the individual page. wrote a hack that double checks for 50 series
• added feature flag to disable discord server message sending
• added some new AMD cpus to json database. its partial manual and partial automated process which i would like to make mostly automated (+ handle different categories) in the future
• New channel on discord to post same price alerts i get internally
• Added a metrics table to the database. Initial use is to track hourly levels of email alerts set and graph it.
• Cool feedback from discord, nearly 200 users in a few days. Not making any money from this but might be good to have a base community when new features are built
• Refreshed the description generating scripts for gpus to now use a 3 step process that makes use of searched grounded ai from 2 sources in the first step and a second step to combine that information into a specific format. Seems to give more useful and accurate information. Also uses browser scrapers for grok and perplexity so don’t have to pay for api usage. 3rd step is to normalise the combined output into a markdown file, removing ai preambles, auto creating links.
• Progress using search grounded ai to populate the database for ssd variants and specs. This time my approach is to use ai to find variants of a given ssd then do some review quality control steps.