mirror of
https://github.com/run-llama/llama-cloud-py.git
synced 2026-08-24 10:52:54 -04:00
4f0c201b2f
Testing showed cost_effective parse produces oversized/missing bounding boxes (3 of 7 fields broken). Agentic parse fixes all of them while keeping extraction on cost_effective tier. Tested all 4 tier combos: - CE extract + default parse: filing_date 363x331, primary_examiner NO BBOX - CE extract + agentic parse: filing_date 118x10, primary_examiner 210x20 - Agentic extract + default parse: same as above (parse tier is what matters) - Agentic extract + agentic parse: same as above Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>