Updates June 30
PGrid
- Added url slugs to ssd variants which was missing and caused ssds to be different from other categories. This was causing issues with clickhouse queries being shared between categories. Added migration for existing ssd variants
- Refactoring clickhouse ssd queries to both use variant url slugs as well as regions
- Added details pages for ssds to view listings and price history
- There was a bug with clickhouse queries for getting listings of an item. There was a cpu that changed price and the change in price was visible in the graph but the specific item listing was shown as out of stock. Spent some time investigating. Had a potential fix but due to changing base queries wanted to set up automated testing.
- Added new gpu: RTX 5050. updated classification notification to let me know if theres lots of gpus waiting to be classified which could indicate theres a new gpu model.
- Update gpu product description to have an option to only use perplexity instead of dual perplexity and chatgpt. chatgpt scraping is currently broken
- Minor fixes with pnpm lockfiles and how root and worker folders are related to each other
- Got an initial clickhouse integration test working that used test containers, set up migrations and set up fixture data which was exported using a script for specific test gpus, cpus and ssds
- Started using claude code to create thorough tests for each of the clickhouse queries.

- Spent some time updating eslint config for workers service so claude can do linting but claude ran into issues. Ended up using tsc for claude to get lint info.
- Investigated bug with frontend build where the sync postgres to clickhouse task would run a docker container but that container wouldnt exit after finishing. This caused the entire server to kill the github action workers after multiple containers didnt exit. Potential fix by adding exit calls to prisma and clickhouse clients.