Merge 8d621d4d5a into cfa6c7cb17

packages/tools/src/vercel/middleware.ts

@@ -104,11 +104,41 @@ export const convertToConversationMessages = (
 					type: "text",
 					text: content.text,
 				})
-			} else if (
-				content.type === "file" &&
-				content.mediaType.startsWith("image/")
-			) {
-				const url = toConversationImageUrl(content.data, content.mediaType)
+			} else if (content.type === "file") {
+				const fileContent = content as {
+					mediaType?: unknown
+					mimeType?: unknown
+					data?: unknown
+				}
+				const mediaType =
+					typeof fileContent.mediaType === "string"
+						? fileContent.mediaType
+						: typeof fileContent.mimeType === "string"
+							? fileContent.mimeType
+							: ""
+				if (mediaType.startsWith("image/")) {
+					const url = toConversationImageUrl(fileContent.data, mediaType)
+					if (url) {
+						contentParts.push({ type: "image_url", imageUrl: { url } })
+					}
+				}
+			} else if (content.type === "image") {
+				const imageContent = content as {
+					image?: unknown
+					data?: unknown
+					mimeType?: unknown
+					mediaType?: unknown
+				}
+				const mediaType =
+					typeof imageContent.mimeType === "string"
+						? imageContent.mimeType
+						: typeof imageContent.mediaType === "string"
+							? imageContent.mediaType
+							: "image/jpeg"
+				const url = toConversationImageUrl(
+					imageContent.image ?? imageContent.data,
+					mediaType,
+				)
 				if (url) {
 					contentParts.push({ type: "image_url", imageUrl: { url } })
 				}

packages/tools/src/vercel/util.ts

@@ -10,7 +10,7 @@ import { toConversationImageUrl } from "../conversations-client"
 export type {
 	ProfileStructure,
 	ProfileMarkdownData,
-} from "../shared"
+} from "../shared/types"
 
 // Provider v2 does not export V3 names, so keep the public declaration on the
 // common V2 surface and structurally accept V3 models at the wrapper boundary.
@@ -98,12 +98,30 @@ export const hasPersistableUserContent = (
 			if (part.type === "text" && typeof part.text === "string") {
 				return Boolean(part.text.trim())
 			}
-			return (
-				part.type === "file" &&
-				typeof part.mediaType === "string" &&
-				part.mediaType.startsWith("image/") &&
-				toConversationImageUrl(part.data, part.mediaType) !== null
-			)
+			if (part.type === "file") {
+				const mediaType =
+					typeof part.mediaType === "string"
+						? part.mediaType
+						: typeof part.mimeType === "string"
+							? part.mimeType
+							: ""
+				return (
+					mediaType.startsWith("image/") &&
+					toConversationImageUrl(part.data, mediaType) !== null
+				)
+			}
+			if (part.type === "image") {
+				const mediaType =
+					typeof part.mimeType === "string"
+						? part.mimeType
+						: typeof part.mediaType === "string"
+							? part.mediaType
+							: "image/jpeg"
+				return (
+					toConversationImageUrl(part.image ?? part.data, mediaType) !== null
+				)
+			}
+			return false
 		})
 	})
 }

packages/tools/test/with-supermemory/conversation-conversion.test.ts

@@ -355,4 +355,36 @@ describe("convertToConversationMessages", () => {
 			{ role: "assistant", content: "Found it" },
 		])
 	})
+
+	it("converts standard Vercel AI SDK type: 'image' parts to conversation image_url", async () => {
+		const params: LanguageModelV2CallOptions = {
+			prompt: [
+				{
+					role: "user",
+					content: [
+						{ type: "text", text: "Look at this screenshot" },
+						{
+							type: "image",
+							image: "https://example.com/screenshot.png",
+							mimeType: "image/png",
+						},
+					],
+				} as unknown as LanguageModelV2Message,
+			],
+		}
+
+		expect(await persistMessages(params, "I see it")).toEqual([
+			{
+				role: "user",
+				content: [
+					{ type: "text", text: "Look at this screenshot" },
+					{
+						type: "image_url",
+						imageUrl: { url: "https://example.com/screenshot.png" },
+					},
+				],
+			},
+			{ role: "assistant", content: "I see it" },
+		])
+	})
 })