Multi-turn Chat with Gemini AI on iOS with SwiftUI

Watch the video or follow along with the post below.

In the previous post, we saw how we could send some text-only input & get a response from Gemini AI’s gemini-pro model. This is alright for simple tasks but to have a more conversational dialogue with the AI, this is not sufficient. Thankfully, Google has a solution for that.

According to the documentation

Using Gemini, you can build freeform conversations across multiple turns. The SDK simplifies the process by managing the state of the conversation, so unlike with generateContent, you don’t have to store the conversation history yourself.

So, all is well? Actually, no. For starters, according to the documentation:

Note: The gemini-pro-vision model (for text-and-image input) is not yet optimized for multi-turn conversations. Make sure to use gemini-pro and text-only input for chat use cases.

This means we cannot have conversations with the AI that involve image inputs, that’s what the gemini-pro-vision model is for. So for now, we can build a text-only exchange with the AI using the gemini-pro model.

The SDK simplifies this with the Chat object.

An object that represents a back-and-forth chat with a model, capturing the history and saving the context in memory between each message sent.

Initialising an instance of Chat is as simple as:

var chat = GenerativeModel(name: "gemini-pro", apiKey: APIKey.default).startChat(history: history)

Here, history is any existing history you’d like to pass to the client of type ModelContent that is comprised of the role (whether user or the ai model) and parts (the actual content). The Chat client then maintains this history in memory so you could actually refer previous inputs & AI will be able to go back & process it for the duration of the current session.

Looking at the documentation, Google’s sample projects implement a rudimentary chat storage for display purposes. The easiest way to do this would be to first define the role for each message, whether sent by the user(input) or received from the model(output)

enum ChatRole {
    case user
    case model
}

We can then construct a ChatMessage like so:

struct ChatMessage: Identifiable, Equatable {
    let id = UUID().uuidString
    var role: ChatRole
    var message: String
}

The id here will come in handy with a ScrollViewProxy to scroll to the most recent message in the chat when we receive a response from the AI & add it to our chat storage. Finally, we create the Chat instance with the gemini-pro model, start a chat and send messages.

chat = GenerativeModel(name: "gemini-pro", apiKey: APIKey.default).startChat(history: history)

let response = try await chat?.sendMessage(message)

Notice that we call the sendMessage function here compared to the generateContent function we were calling on the model in the previous post.

With some cosmetic detailing applied to give it a messaging look & feel, we can have something like this:

And that’s it! The complete code can be found here.

Leave a comment if you have any questions and share this article if you found it useful !