I wanted to maximise the Cloudflare static caching and not fiddle too much with security, so it's 400-ish data chunks and each repo should have a stable hash. I only use everything down to 10 mentions since the long tail is long. Might revisit that later, but this gets me published.
Using Google BigQuery and GCP a lot, Jev for classification, Codex as an agent with GPT-6.1 Sol High.