ReMe/reme/steps
WQS 6125fc197d
feat(auto-memory): add session image support (#532)
* feat(auto-memory): add opt-in image input

* feat(auto-memory): caption session images into source-linked notes

* fix(auto-memory): version Pillow 10 compatible image preparation

* fix(auto-memory): harden image evidence and retry boundaries

* fix(auto-memory): preserve image evidence across replay and concurrent writes

Keep persisted image positions through disabled history backfills and transcript filtering. Merge note links atomically, map 16-bit grayscale without clipping, recheck restored caption owners, and reuse unchanged identity metadata during batch publication. Add regressions and document conservative custom-rename behavior.

* refactor(auto-memory): restore main baseline for image modes v2

* refactor(images): share resource caption preprocessing and model calls

* feat(watch): support scoped exclusions for managed session images

* feat(auto-memory): add opt-in resource and caption-only image input

* refactor(auto-memory): keep caption-only mode with text fallback

* refactor(auto-memory): make caption-only mode dispatch explicit

* docs(auto-memory): focus image guide on caption-only mode

* feat(auto-memory): add direct multimodal image extraction

* refactor(auto-memory): route direct images through the vision model

* test(auto-memory): consolidate overlapping image regressions

* feat(auto-memory): interleave direct images with conversation text

* docs(auto-memory): clarify history rendering scope for direct inputs

* feat(auto-memory): require vision declaration for direct-only images

* refactor(auto-memory): keep bound model and native image helpers

* refactor(auto-memory): keep image inputs native and opt-in

* refactor(auto-memory): simplify image docs and tests

* refactor(auto-memory): confine image adaptation to image requests

* refactor(auto-memory): remove image switch type validation
2026-09-20 15:35:00 +08:00
..
benchmark feat(benchmark): extract BEAM and LongMemEval into standalone plugins (#512) 2026-09-03 14:06:39 +08:00
common fix: skip dependency bindings in status memory walk (#510) 2026-08-31 16:24:41 +08:00
evolve feat(auto-memory): add session image support (#532) 2026-09-20 15:35:00 +08:00
file_io refactor(auto-fin): one note per topic, merged by a separate digest step (#558) 2026-09-20 14:52:56 +08:00
index refactor(auto-fin): one note per topic, merged by a separate digest step (#558) 2026-09-20 14:52:56 +08:00
transfer feat(benchmark): add lme benchmark steps (#326) 2026-07-07 18:53:39 +09:00
__init__.py feat(plugins): extract DingTalk integration (#551) 2026-09-15 20:05:52 +08:00
base_step.py feat: auto-tag plugin-generated reports (#534) 2026-09-11 11:52:52 +08:00