1.3.0

Highlights

• We're automating spam identification and topic classification of forum questions. This is an AI-based feature built using the latest LLM models. Looking forward, the spam moderation queue will be populated only with questions that couldn't be automatically verified as spam. The topic moderation queue will be retired, and its current backlog will be automatically categorized in the near future. Contributors can continue using the Edit details button on forum posts to adjust the product, topic, or other info when needed.

What's Changed

• Bump django from 4.2.20 to 4.2.21 by @dependabot in #6650
• Playwright: extend test coverage to kb article helpfulness voting by @emilghittasv in #6648
• Playwright: Extend coverage to kb helpfulness widget feedback canceled state & helpfulness metadata percentage information by @emilghittasv in #6653
• initial pieces for llm-based question classification by @escattone in #6659
• add spam confidence by @escattone in #6660
• add initial topic metadata by @escattone in #6661
• Prompt optimizations by @akatsoulas in #6664
• Catch MultipleObjectsReturned exception by @akatsoulas in #6666
• Classify questions by @akatsoulas in #6663
• remove obsolete file by @escattone in #6670
• data migration to rename topic title by @escattone in #6669
• only select from active and visible topics by @escattone in #6671
• include title of question when classifying by @escattone in #6672
• ensure consistent topic and topic tag when classifying by @escattone in #6673
• ensure question_classifier task is idempotent by @escattone in #6674
• Don't load djang-silk in shell_plus by @akatsoulas in #6679
• Playwright: Update AAQ test question content by @emilghittasv in #6678
• helper function that provides product info by @escattone in #6681

Full Changelog: 1.2.16...1.3.0