8.3.1
Patch Changes
- #4071
182434fThanks @DerGeraetK! - fix(ai): declare$ai_cache_reporting_exclusive: falseon OpenAI wrapper events so ingestion no longer double-bills cached input tokens for Claude models served through OpenAI-compatible hosts (e.g. OpenRouter). The flag stays unset when callers pass their own input or cache token counts throughposthogProperties, so existing passthrough workarounds keep reporting correctly.
(2026-07-16)