Google AI Studio Provisions Cloud Firestore for Prompt-Built Apps
In a Google for Developers guide, Nikita Namjoshi shows how AI Studio turns a prompt-built app from a static page into a Firebase-backed application when the requested workflow requires stored data. Using a book-tracking app as the example, she argues that AI Studio can detect the need for persistence, prompt the user to enable Firebase, wire in authentication, and save extracted book metadata to Cloud Firestore. The guide also shows that publishing the app publicly requires configuring a Gemini API key, with an optional spend cap to limit usage costs.
AI Studio treats persistence as part of the app, not an afterthought
Nikita Namjoshi frames the problem simply: most apps people build are not just static web pages, and AI Studio can respond by prompting for Firebase, provisioning persistence and authentication, and writing app data into Cloud Firestore. An event site may need to collect RSVPs. A shared registry or gift list needs to remember entries. A nutrition tracker needs records over time. In those cases, the missing piece is a database.
Most apps that you build won't be just static web pages, you're going to need to store information.
Her example uses a book tracker rather than a generic form. The prompt asks AI Studio to build an app where the user can take a picture of a book, store the name, author, and date the picture was taken, and display every logged book in a table. The prompt also specifies a “dark academia” style, making the request both visual and stateful: the app needs a front end, camera interaction, extraction of book metadata, a registry, authentication, and storage.
After the prompt, AI Studio generates several design directions. Namjoshi selects one with an archivist-like aesthetic. AI Studio then prompts her to enable a database. The interface describes the step as enabling Firebase, with data saved over time, a free tier, and a selectable hosting location. Namjoshi says enabling Firebase allows Gemini to set up provisioning, authentication, and database integration so that information entered through the app is automatically stored.
The point is not that the user hand-builds a schema first. In this flow, the database requirement emerges from the app request itself: a book tracker that remembers captured books cannot work as a static page.
The generated app captures a book and writes structured data
Once the app is built, Nikita Namjoshi opens it full screen, authenticates, activates the camera lens, holds up a book, and commits the capture. The app’s interface is titled “THE ARCHIVIST” and shows an acquisition registry. Before capture, the registry has no entries. After capture, the book appears in the registry with the title, author, and acquisition date.
The first captured book is Small Is Beautiful. Namjoshi notes that the app also kept the picture, which she says was not necessary for her use case but was captured anyway. The registry shows the title, author E. F. Schumacher, and the archived date.
This matters because the example is doing more than saving a typed form submission. In the visible AI Studio response, Gemini extracts the book title and author from the captured image. The app also stores a timestamp, a base64 representation of the cover photo, and a user ID tied to authentication.
| Stored field | What the visible AI Studio response says it represents |
|---|---|
| title | The name of the book, extracted by Gemini AI |
| author | The author or authors, extracted by Gemini AI |
| imageUrl | A base64 representation of the captured cover photo |
| capturedAt | A server-side timestamp of when the entry was created |
| userId | The user’s unique authentication ID, used to keep the library private |
Namjoshi then asks the model, in the chat, to show where the data is stored. AI Studio returns a link to the Firestore console. In Firebase, the Cloud Firestore view shows a books collection and a document containing the captured record: author “E. F. Schumacher,” a captured timestamp, an image URL beginning with a base64 JPEG string, the title “Small is Beautiful: Economics as if People Mattered,” and a user ID.
The Firestore view shows the registry data persisted in Cloud Firestore rather than only in the browser. The generated app can also point the user back to the underlying storage.
Publishing requires an API key and can be constrained with a spend cap
Deployment adds an operational constraint. Nikita Namjoshi wants to try the app on her phone, so she goes to publish it. AI Studio first requires configuration of the Gemini API key under secrets. The interface shows a GEMINI_API_KEY secret initially set to the AI Studio Free Tier.
Namjoshi says that if she wants to deploy the app to Google Cloud and make it available at a public URL she can share, she needs to use a paid API key. She sets one, then returns to publishing.
She also points to a monthly spend cap. The paid API key is the setup requirement for publishing the app publicly; the spend cap is a separate cost-control setting. In the demo, the cap is set at five dollars. The UI says Gemini API usage will pause when the cap is reached, while also noting it may be subject to overages during a 10-minute processing latency. Namjoshi describes the cap as useful if she shares the app with friends, because it prevents spending from going out of control.
After saving the secret and publishing, AI Studio reports that the app is published, shows that the Gemini API spend cap has been set, and provides an app URL on run.app. Namjoshi opens the deployed app, authenticates again, and sees the previously captured book under her account.
The published app therefore depends on more than generated interface code. AI Studio can generate and provision the app, but making it publicly available involves API-key configuration and, in this example, an added usage limit.
The phone test shows the database is shared across clients
To test the deployed version, Nikita Namjoshi copies the app URL and opens it on her phone. She authenticates again, enables the camera, takes a picture of another book, and waits while the app analyzes the image.
The desktop app then updates to show the new entry: The Embodied Mind: Cognitive Science and Human Experience, by Francisco J. Varela, Evan Thompson, and Eleanor Rosch. The registry now shows two entries: the newly captured book and Small Is Beautiful. Both are dated April 22, 2026.
The workflow ends with the same app running on a phone and on a desktop, authenticated to the same user, with records appearing in the shared registry after capture. Namjoshi says that now, whenever she finishes reading a book, she can take a picture with her phone and have it logged in the app so she can track her reading goals for the year.
Her conclusion is modest but clear: adding a database in AI Studio expands what a prompt-built app can do beyond a static page. The book tracker is the example, but the underlying pattern is broader: when an app needs remembered user data, AI Studio can prompt for Firebase, create the backing store, wire in authentication, and expose the data in Firestore.