Posted by Thomas Ezan, Sr Developer Relation Engineer
With the appearance of Generative AI, Android builders now have entry to capabilities that have been beforehand out of attain. As an illustration, now you can simply add picture captioning to your app with none pc imaginative and prescient data.
With the upcoming launch of the steady model of VertexAI in Firebase in a couple of weeks (accessible in Beta Since Google I/O), you’ll effortlessly incorporate the capabilities of Gemini 1.5 Flash and Gemini 1.5 Professional into your app. The inference runs on Google’s servers, making these capabilities accessible to any gadget with an web connection.
A number of Android builders have already begun leveraging this expertise. Let’s discover a few of their creations.
Generate your meals for the week
The crew behind Meal Planner, a meal planner and buying record administration app, is leveraging Gemini 1.5 Flash to create authentic meal plans. Based mostly on the consumer’s weight loss program, the variety of individuals you’re cooking for and any meals allergic reactions or intolerances, the app routinely creates a meal plan for the chosen week.
For every dish, the mannequin lists elements and portions, considering the variety of parts. It additionally offers directions on methods to put together it. The app routinely generates a buying record for the week based mostly on the ingredient record for every meal.
To allow dependable processing of the mannequin’s response and to combine it within the app, the crew leveraged Gemini’s JSON mode. They specified responseMimeType = “software/json” within the mannequin configuration and outlined the anticipated JSON response schema within the immediate (see API documentation).
Following the launch of the meal technology function, Meal Planner obtained overwhelmingly optimistic suggestions. The function simplified meal planning for customers with dietary restrictions and helped scale back meals waste. Within the months after its introduction, Meal Planner skilled a 17% surge in premium customers.
Journaling whereas chatting with Leo
Just a few months in the past, the crew behind the journal app Life wished to offer an revolutionary approach to let their customers log entries. They created “Leo”, an AI diary assistant chatting with customers and changing conversations right into a journal entry.
To switch the habits of the mannequin and the tone of its responses, the crew used system directions to outline the chatbot persona. This enables the consumer to set the habits and tone of the assistant: Decide “Skilled and formal” and the mannequin will preserve the dialog strict, choose “Pleasant and cheerful” and it’ll loosen up the dialogue with a lot of emojis!
The crew noticed a rise of consumer engagement following the launch of the function.
And if you wish to know extra about how the Life builders used Gemini API of their app, we had an important dialog with Jomin from the crew. This dialog is a part of a brand new Android podcast collection referred to as Android Construct Time, that you could additionally watch on YouTube.
Create a nickname on the fly
The HiiKER app offers offline mountain climbing maps. The app additionally fosters a neighborhood by letting customers score trails and leaving feedback. However customers signing up don’t all the time add a username to their profile. To keep away from the danger of reducing the conversion price by making username choice obligatory at signup time, the crew determined to make use of the Gemini API to counsel distinctive usernames based mostly on the consumer’s nation or space.
To generate authentic usernames, the crew set a excessive temperature worth and performed with the top-Okay and top-P values to extend the creativity of the mannequin.
This AI-assisted function led to a big raise within the proportion of customers with “full” profiles contributing to a optimistic impression on engagement and retention.
It’s time to construct!
Generative AI remains to be a really new area and we’re simply beginning to have easy accessibility to those capabilities for Android. From enabling superior personalization, creating pleasant interactive experiences or simplifying signup, you might need distinctive challenges that you’re attempting to resolve as an Android app developer. It’s a nice time to begin taking a look at these challenges as alternatives that generative AI might help you sort out!
You’ll be able to be taught extra in regards to the superior options of the Gemini Cloud fashions, discover an introduction to generative AI for Android builders, and get began with Vertex AI in Firebase documentation.
To be taught extra about AI on Android, take a look at different assets we now have accessible throughout AI in Android Highlight Week.
Use #AndroidAI hashtag to share your creations or suggestions on social media, and be a part of us on the forefront of the AI revolution!