fix(enhance): include note images in summaries

[ComputelessComputer]

Read referenced image attachments for multimodal summary models and send them as AI SDK image parts. Add fs-sync attachment reading support and coverage.

---

Note

Medium Risk
Adds local attachment reads and sends image bytes to external LLM APIs when vision models are selected; mitigated by model gating and size/count limits but still affects cost, latency, and note content exposure.

Overview
Enhance can now use note images as multimodal context when the selected LLM supports vision.

When the current provider/model passes modelSupportsImageInput, the transform step gathers images from pre-meeting transcript memos and post-meeting session raw_md. References are parsed from TipTap JSON and markdown (asset:// / local attachment paths and inline data: URLs). Referenced session attachments are loaded via a new attachment_read fs-sync API (Rust core, Tauri command, JS bindings, default permissions), with caps of 10 images and 8MB per image.

The enhance workflow attaches those images as AI SDK ImageParts alongside the text prompt (template discovery and summary streaming) and adds a short instruction that attached images are visual context. Text-only models skip collection entirely.

Tests cover model gating, image reference parsing, attachment reads, and transform wiring.

^{Reviewed by Cursor Bugbot for commit 5a040b2. Bugbot is set up for automated code reviews on this repo. Configure here.}

[netlify]

✅ Deploy Preview for old-char canceled.

Name	Link
🔨 Latest commit	5a040b2
🔍 Latest deploy log	https://app.netlify.com/projects/old-char/deploys/6a1d0a8bd176940008dea827

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 9d66acf. Configure here.}