feat: add image paste support to chat input

[morgmart]

Category: new-feature
User Impact: Users can now paste screenshots and images directly into the chat input with Cmd+V and send them to the AI alongside their message.

Problem: Pasting an image from the clipboard into the chat input did nothing — there was no way to share a screenshot or visual context without describing it in text.

Solution: Added clipboard paste handling to the chat textarea that captures image items, displays them as thumbnails above the input, and forwards them as base64 content blocks to the ACP agent. The ACP driver already supported images via (base64, mimeType) pairs — this PR wires that capability end-to-end from the UI through to the backend.

File changes

src/features/chat/ui/ChatInput.tsx
Added onPaste handler that intercepts clipboard image items and converts them to base64. Added PastedImageThumb local component that renders a thumbnail with a hover-reveal remove button and click-to-expand lightbox. Images appear above the persona tag so they're always at the top of the input area. canSend updated to allow sending with images even when text is empty.

src/features/chat/ui/ChatView.tsx
Updated handleSend (and the deferred-send ref used during persona switches) to accept and forward an optional images array to sendMessage.

src/features/chat/hooks/useChat.ts
Updated sendMessage to accept images?: { base64, mimeType }[], embed them as ImageContent blocks in the local user message for display in the timeline, and forward them to acpSendMessage.

src/shared/api/acp.ts
Added images?: [string, string][] to AcpSendMessageOptions and passes it through to the Tauri invoke call.

src-tauri/src/commands/acp.rs
Added images: Vec<(String, String)> parameter to the acp_send_message Tauri command and passes it through to AcpService::send_prompt.

src-tauri/src/services/acp/mod.rs
Added images parameter to send_prompt and passes &images to driver.run() instead of the previous hardcoded &[].

src/shared/ui/dialog.tsx
New shadcn-style Dialog component (adapted from shadcn/ui GitHub, using @radix-ui/react-dialog to match the project's existing Radix pattern). Animation-free content to avoid the transform-origin issue that caused dialogs to slide in from a corner.

src/shared/ui/ImageLightbox.tsx
New reusable lightbox component built on Dialog. Sized to fit the viewport, closes on overlay click or Escape. Positioned in shared/ui since image viewing will be needed in the message timeline too.

package.json / pnpm-lock.yaml
Added @radix-ui/react-dialog dependency for the Dialog primitive.

Reproduction Steps

1. Run the app and open a chat session
2. Take a screenshot (Cmd+Shift+4 on macOS) or copy any image to the clipboard
3. Click into the chat input and press Cmd+V
4. A thumbnail should appear above the input — hover it to reveal the remove button, click it to open the lightbox
5. Type a message (optional) and send — the image is included with the prompt to the AI

Screenshots/Demos

Before: Cmd+V in the chat input did nothing.

After: Pasted images appear as thumbnails above the input with hover-to-remove and click-to-expand behavior. Images are forwarded to the AI with the message.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 89aff4de70

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".