25 lines (19 loc) · 968 Bytes

CHANGELOG.md

[1.2] - 2025-09-22

Added

Image “Lite” representation: summary + labels + scene graph + micro Q/A + OCR (all text, token-lean).
Export: context command (JSON/Markdown) with optional target_tokens cap.
Estimated savings report after each compression run.

Changed

Defaults: Images → Compact representation=lite; Conversation → 400 tokens; Notes → JSON.
Batch behavior: process all images in the current conversation by default.

Notes

“full” image representation (embeddings/pHash/ANN) remains offline-only.
Prefer Compact+Lite for day-to-day—~82–85% token savings vs. raw image encodings.

[1.1] - 2025-09-21

Added

Initial Export: context command and snapshot schema.

[1.0] - 2025-09-21

Added

Base Context Compression Menu: images, conversation, code, notes, document, all.
Ultra/Compact/Verbose image summaries and 400-token conversation rollups.