Updates December 22

PGrid

• New CLI command to see/clear stale classification pipeline rows for SSDs/CPUs
• Tweaked CPU classification prompts to handle eBay multi-listings
• Added SSD classification run file to be done as a worker task. Also added a debug email after classification runs
• Posted on Reddit about SSD and CPU tracking
• Fixed Computer Alliance SSD scraping not working (they moved their SSD URL)
• Fixed Amazon scraper to handle a location popup that happens if a proxy doesn’t appear to be from the US from Amazon’s point of view
• Extended specs search queue deduping to SSDs. Updated deduping functions to be able to be used across different categories
• Cleaned up old generic item Amazon scraping that was unused after moving to ‘sharded’ Amazon scraping
• Updated Amazon single page crawl to parse the specs section of the page. Goal is to use this to help classification later
• Fixed Amazon single page scraper where it incorrectly returned that the listing was ended. This was due to the canonical link being different to the current ASIN, but they point to the same product (and there is no redirect)
• Updated filter sliders to be able to select the same range but have the thumbs be spaced apart so they don’t overlap
• Got Google Vertex Gemini Flash 3 (search-grounded) working with BAML. This is to be used to verify attributes from specs search individually
• Work in progress for implementing classification for RAM
• Work in progress for migrating gpu classification to new architecture and deprecating old classification