feat: support multiple image uploads for AI event generation 🖼️

- Updated OpenRouter integration to accept an array of image URLs
- Updated ImagePicker to use the `multiple` attribute natively
- Added `appendImagesDeduped` for handling client-side image deduplication
- Enhanced clipboard pasting to extract multiple images at once
- Rendered multiple images in a horizontal thumbnail strip in the AIToolbar
- Added tests to cover multi-image logic and AI request mapping

src/lib/ai-event-messages.ts

@@ -0,0 +1,47 @@
+/**
+ * Pure helper that builds the OpenRouter chat messages array for a multimodal
+ * AI-event request.
+ *
+ * Extracted from the API route so it can be unit-tested without mocking HTTP.
+ */
+
+type TextPart = { type: "text"; text: string };
+type ImageUrlPart = { type: "image_url"; imageUrl: { url: string } };
+type ContentPart = TextPart | ImageUrlPart;
+
+type Message =
+	| { role: "system"; content: string }
+	| { role: "user"; content: ContentPart[] };
+
+/**
+ * Builds a 2-message array:
+ *   [0] system  → the system prompt string
+ *   [1] user    → [text part, ...image_url parts (one per image)]
+ *
+ * @param systemPrompt  Instruction string for the model
+ * @param prompt        Optional text from the user
+ * @param images        Array of base64 data URLs
+ */
+export function buildMultimodalMessages(
+	systemPrompt: string,
+	prompt: string | undefined,
+	images: string[],
+): Message[] {
+	const userContent: ContentPart[] = [
+		{
+			type: "text",
+			text: prompt || "Extract all calendar events from these images.",
+		},
+		...images.map(
+			(url): ImageUrlPart => ({
+				type: "image_url",
+				imageUrl: { url },
+			}),
+		),
+	];
+
+	return [
+		{ role: "system", content: systemPrompt },
+		{ role: "user", content: userContent },
+	];
+}