Merge pull request #2371 from code-yeongyu/fix/issue-2323

fix: respect multimodal provider vision capabilities

src/shared/vision-capable-models-cache.ts

@@ -0,0 +1,17 @@
+import type { VisionCapableModel } from "../plugin-state"
+
+let visionCapableModelsCache = new Map<string, VisionCapableModel>()
+
+export function setVisionCapableModelsCache(
+  cache: Map<string, VisionCapableModel>,
+): void {
+  visionCapableModelsCache = cache
+}
+
+export function readVisionCapableModelsCache(): VisionCapableModel[] {
+  return Array.from(visionCapableModelsCache.values())
+}
+
+export function clearVisionCapableModelsCache(): void {
+  visionCapableModelsCache = new Map<string, VisionCapableModel>()
+}