What's Changed
Features
- feat(evals): add payload-backed evaluation rule filters by @wochinge in #16484
- feat(evals): refresh all-caps, exact-match, and keyword-overlap templates by @annabellscha in #16515
- feat(otel): log encoded request body sizes by @sumerman in #16517
- feat(in-app-agent): record tool approval source in the event stream by @LysanderKie in #16372
Fixes
- fix(pricing): price GPT-5.4 reasoning tokens by @hassiebp in #16480
- fix(shared): apply score pagination when offset is 0 by @niklassemmler in #16405
- fix(experiments): keep evaluator search modal mounted by @wochinge in #16482
- fix: reject filters on pair-expanded dimensions by @sumerman in #16481
- fix(automations): require fresh tokens for dispatch URL changes by @hassiebp in #16479
- fix: enforce OTLP ingestion request body limits by @sumerman in #16487
- fix(evals): preserve LLM judge BullMQ retries by @wochinge in #16510
- fix(evals): render boolean test results as text by @wochinge in #16514
- fix(web): contain verbose media labels by @wochinge in #16488
- fix(assistant): Optimize sandbox & silent mcp tool prompt guidance by @bezbac in #16411
Chores
- chore(pricing): cut gpt-5.6-sol price across all tiers by @langfuse-bot in #16436
- chore(deps): bump next to 16.3.2 by @nimarb in #16485
- refactor(web): clean up
CreateProjectMemberButtonby @bezbac in #16483 - refactor(web): Create more feature dirs in storybook by @bezbac in #16521
Full Changelog: v4.17.0...v4.18.0