Release · 2026 · 04 · 02 · 2 min read
v0.1.2 — Vision + image generation
Multi-model vision, a built-in image generation skill, and content-addressed memory snapshots.
Vision became pluggable: Gemini 2.5 Pro and GPT-5 are both wired up out of the box. Image generation lands as a default skill.
Memory snapshots moved to content-addressed storage, which makes restores deterministic and dedupes most of the disk footprint.
In this release
- —Vision tool now multi-model (Gemini, GPT-5)
- —Image generation skill (defaults to gemini-3-flash-preview)
- —Memory snapshots are now content-addressed