Updates December 15
PGrid
- Added URL query params for filtering so filtered table state can be shared
- Finished implementation of CPU model specs search.
- Deleted separate table we had for CPU generation which only had a single metadata of release date for each CPU generation. Refactored everything to deal with this. Some issues with tests.
- Added multiple steps to search the DB before doing specs search, and DB search afterwards to normalise values
- Added a verification step which uses individual Perplexity API searches for the most important attributes to double-check they are correct
- Updated specs search deduping to include a ‘popularity score’ so the most common models in the specs search after deduping are prioritised for specs search
- Updated CPU description generation to take input of the CPU row from the DB. This fixes inconsistent release dates
- Updated CPU description cleaning to be able to reject input markdown file from Perplexity. Some files were error files and had no information
- Updated ClickHouse schema to have daily item prices materialised views for all sources and another table excluding Amazon which gets used for price history
- Updated CLI with deals and OpenAI usage commands

- Updated eBay scraper to better handle ended listings
- Added filtering for CPU table including saving filter state with URL query params