feat: input_image support

Sources/Core/Models/Item.swift

@@ -37,12 +37,14 @@ import MetaCodable
 			case text(String)
 			case audio(Audio)
 			case inputText(String)
+			case inputImage(String)
 			case inputAudio(Audio)
 
 			public var text: String? {
 				switch self {
 					case let .text(text): text
 					case let .inputText(text): text
+					case let .inputImage(image): image
 					case let .audio(audio): audio.transcript
 					case let .inputAudio(audio): audio.transcript
 				}
@@ -419,6 +421,7 @@ extension Item.Message.Content: Codable {
 		case text
 		case audio
 		case transcript
+        case image_url
 	}
 
 	private struct Text: Codable {
@@ -440,7 +443,10 @@ extension Item.Message.Content: Codable {
 			case "input_text":
 				let container = try decoder.container(keyedBy: Text.CodingKeys.self)
 				self = try .inputText(container.decode(String.self, forKey: .text))
-			case "output_audio":
+            case "input_image":
+                let inner = try decoder.container(keyedBy: CodingKeys.self)
+                self = try .inputImage(inner.decodeIfPresent(String.self, forKey: .image_url) ?? "")
+            case "output_audio":
 				self = try .audio(Item.Audio(from: decoder))
 			case "input_audio":
 				self = try .inputAudio(Item.Audio(from: decoder))
@@ -459,7 +465,10 @@ extension Item.Message.Content: Codable {
 			case let .inputText(text):
 				try container.encode(text, forKey: .text)
 				try container.encode("input_text", forKey: .type)
-			case let .audio(audio):
+            case let .inputImage(imageURL):
+                try container.encode(imageURL, forKey: .image_url)
+                try container.encode("input_image", forKey: .type)
+            case let .audio(audio):
 				try container.encode("output_audio", forKey: .type)
 				try container.encode(audio.audio, forKey: .audio)
 				try container.encode(audio.transcript, forKey: .transcript)

Sources/UI/Conversation.swift

@@ -13,7 +13,7 @@ public enum ConversationError: Error {
 public final class Conversation: @unchecked Sendable {
 	public typealias SessionUpdateCallback = (inout Session) -> Void
 
-	private let client: WebRTCConnector
+    private let client: WebRTCConnector
 	private var task: Task<Void, Error>!
 	private let sessionUpdateCallback: SessionUpdateCallback?
 	private let errorStream: AsyncStream<ServerError>.Continuation
@@ -152,11 +152,25 @@ public final class Conversation: @unchecked Sendable {
 	/// Send a text message and wait for a response.
 	/// Optionally, you can provide a response configuration to customize the model's behavior.
 	public func send(from role: Item.Message.Role, text: String, response: Response.Config? = nil) throws {
-		try send(event: .createConversationItem(.message(Item.Message(id: String(randomLength: 32), role: role, content: [.inputText(text)]))))
+		let id = UUID().uuidString.replacingOccurrences(of: "-", with: "") // random 32 character string
+		try send(event: .createConversationItem(.message(Item.Message(id: id, role: role, content: [.inputText(text)]))))
 		try send(event: .createResponse(using: response))
 	}
 
-	/// Send the response of a function call.
+	/// Send an image + text message and wait for a response.
+    public func send(from role: Item.Message.Role, image: Data, response: Response.Config? = nil) throws {
+        let dataURI = "data:image/jpeg;base64,\(image.base64EncodedString())"
+        let id = UUID().uuidString.replacingOccurrences(of: "-", with: "") // random 32 character string
+        let message = Item.Message(
+            id: id,
+            role: role,
+            content: [.inputImage(dataURI)]
+        )
+        try send(event: .createConversationItem(.message(message)))
+        try send(event: .createResponse(using: response))
+    }
+
+    /// Send the response of a function call.
 	public func send(result output: Item.FunctionCallOutput) throws {
 		try send(event: .createConversationItem(.functionCallOutput(output)))
 	}
@@ -176,7 +190,8 @@ private extension Conversation {
 				if let sessionUpdateCallback { try updateSession(withChanges: sessionUpdateCallback) }
 			case let .sessionUpdated(_, session):
 				self.session = session
-			case let .conversationItemCreated(_, item, _):
+			case let .conversationItemCreated(_, item, _),
+				let .conversationItemAdded(_, item, _):
 				entries.append(item)
 			case let .conversationItemDeleted(_, itemId):
 				entries.removeAll { $0.id == itemId }