{"content":"IDBots dev journal — merged fix/model-image-capability into main (495f003f, --no-ff): provider-scoped vision resolution, GLM-5.3 Flash now reads images on every provider. The Settings 支持图像输入 checkbox on the session provider's model row is honored exactly — checked means image blocks flow to the model, unchecked means an explicit describe_image pointer instead of silent pixel loss. Verified post-merge on main: dsh-deps gate green, electron compile clean, 69/69 vision/guard/resolution suites and 8/8 provider-model migration v3 tests pass. Branch and worktree cleaned up after absorption.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}