Check out the latest news from Firebase at Cloud Next 2025. Learn more.

Configure Gemini in Firebase within workspaces

Firebase Studio facilitates your development workflows with the following AI-assisted code features:

Suggested code completion as you type.
AI assistance with chat, which is workspace-aware and fully-integrated with your code. It can generate, translate, and explain code. And, with your review and approval, Gemini in Firebase can directly interact with your workspace to update files, run terminal commands, interpret command output, and determine next steps. Learn more at Try chat with Gemini.
Inline actions that you can take on selected pieces of code. For example, you can ask Gemini to make the selected code more readable.
Inline code assistance.

You can customize how Gemini in Firebase helps you by adjusting its settings and adding AI rules files:

Adjust code completion settings.
Adjust your codebase indexing settings.
Customize instructions for Gemini with an AI rules file.
Exclude files from Gemini with .aiexclude files.
Bring your own key: Use other Gemini models in chat

Use Gemini in Firebase in your workspace

Use Gemini in Firebase to boost your coding productivity through the chat panel or inline code assistance.

Use either chat or inline code assistance in your workspace:
- To use chat: In your open workspace, click spark Gemini at the bottom of the workspace.
- To use inline code assistance: Start typing your code and press Tab to accept suggestions.
Be aware that the following two options are enabled by default:
- Suggestions as you type, which provides inline code completion.
- Codebase indexing, which provides better customization and more helpful responses.
To change these selections for your workspace settings in the future:
- To update code completion settings, see Adjust your code completion settings.
- To update code indexing settings, see Adjust your code indexing settings.
You can also exclude specific files and directories from AI indexing. See Exclude files from Gemini with .aiexclude files.

Gemini in Firebase shortcuts

To quickly open chat with Gemini: press Ctrl+Shift+Space (or Cmd+Shift+Space on MacOS).

To view Gemini commands from the command palette:

Open the command palette by pressing Ctrl+Shift+P (or Cmd+Shift+P on MacOS).
Search for Gemini.

A list of Gemini commands appears.

Adjust your code completion settings

To help you write code, Firebase Studio provides AI code completion that predicts and autofills code in any open file as soon as you begin to type.

Be aware that code completion is turned on by default.

To toggle code completion on or off, adjust your code completion settings using one of the following methods:

If you use a settings.json file, set "IDX.aI.enableInlineCompletion" to true or false.
To update settings in the Firebase Studio workspace:
1. Click Manage (located at the bottom left of the workspace), then choose Settings, or press Ctrl+, (Cmd+, on Mac).
  
  If you're using the App Prototyping agent in Prototyper view, click Switch to Code to open Code view.
2. Select the Workspace tab, then search for the Firebase Studio > AI > Enable Inline Completion setting.
3. To turn off code completion, deselect the Enable inline code completion as you type option.

Adjust your codebase indexing settings

You can control whether Gemini indexes your code. Indexing your code provides more helpful results when using chat or inline AI assistance.

Be aware that codebase indexing is turned on by default.

To toggle code indexing on or off, adjust your codebase indexing settings using one of the following methods:

If you use a settings.json file, set "IDX.aI.enableCodebaseIndexing" to true or false.
To update settings in the Firebase Studio workspace:
1. Click Manage (located at the bottom left of the workspace), then choose Settings, or pres Ctrl+, (Cmd+, on Mac).
  
  If you're using the App Prototyping agent in Prototyper view, click Switch to Code to open Code view.
2. Select the Workspace tab, then search for the Firebase Studio > AI > Enable Inline Completion setting.
3. Select Firebase Studio > AI: Enable Codebase Indexing.
4. To turn off code indexing, deselect AI: Enable Codebase Indexing. You must update code indexing settings for each of your workspaces.

Customize instructions for Gemini in Firebase with an AI rules file

You can add context and system prompt information by creating an AI rules file (.idx/airules.md). Gemini in Firebase uses your rules as system instructions and context, ensuring that its responses are customized for your use case.

Use the AI rules file to share custom prompts, best practices, and even important context about your project with Gemini to achieve goals like:

Influencing Gemini's persona and specializing its expertise.
Applying project-wide standards, like coding style, conventions, and technology preferences.
Reducing the amount of information you need to share explicitly in code or chat by providing essential context about your project.

The rules you configure are used by Gemini in chat.

Create and test your AI rules file

To create and test your AI rules file:

Create a new file at .idx/airules.md in your Firebase Studio workspace (in the same directory as your dev.nix file). You can use one of the following options:
- From Explorer (Ctrl+Shift+E), right-click on .idx and select New file. Name the file airules.md and press Enter.
- From the terminal, use your preferred text editor to open .idx/airules.md.
Add content to the file. You may want to add information about the persona Gemini should use (like "You are an expert developer and helpful assistant who knows everything about Next.js"), coding and conversation standards, and context about the project. See the following Example for an example AI rules file.
Save the file and open Gemini in Firebase.
To start using your AI rules, you can do one of the following:
- Rebuild the workspace by refreshing the page. After you rebuild, Gemini in Firebase will use the rules file within chat. Changes to the AI rules file should be reflected in chat immediately.
- If you don't want to rebuild your workspace, you can ask Gemini using chat to load airules.md. If you make changes to the file during the current session, you may need to re-prompt Gemini to load the rules file again.
Ask questions about your code. Gemini responds using the information you included in the rules file as context.

Example

The following is a basic example of a rules file that you might use for a casual game developed with Next.js:

# Persona

You are an expert developer proficient in both front- and back-end development
with a deep understanding of Node.js, Next.js, React, and Tailwind CSS. You
create clear, concise, documented, and readable TypeScript code.

You are very experienced with Google Cloud and Firebase services and how
you might integrate them effectively.

# Coding-specific guidelines

- Prefer TypeScript and its conventions.
- Ensure code is accessible (for example, alt tags in HTML).
- You are an excellent troubleshooter. When analyzing errors, consider them
  thoroughly and in context of the code they affect.
- Do not add boilerplate or placeholder code. If valid code requires more
  information from the user, ask for it before proceeding.
- After adding dependencies, run `npm i` to install them.
- Enforce browser compatibility. Do not use frameworks/code that are not
  supported by the following browsers: Chrome, Safari, Firefox.
- When creating user documentation (README files, user guides), adhere to the
  Google developer documentation style guide
  (https://developers.google.com/style).

# Overall guidelines

- Assume that the user is a junior developer.
- Always think through problems step-by-step.

# Project context

- This product is a web-based strategy game with a marine life theme.
- Intended audience: casual game players between the ages of 17 and 100.

Exclude files from Gemini with `.aiexclude` files

You can control which files in your codebase should be kept hidden from Gemini by including .aiexclude files in your project. This lets you granularly control the project context you share with Gemini.

Much like a .gitignore file, an .aiexclude file tracks files that shouldn't be shared with Gemini, including the chat experience as well as AI features that operate in the editor. An .aiexclude file operates on files at or below the directory that contains it.

Files covered by .aiexclude won't be indexed by Gemini when Codebase Indexing is enabled. Additionally, .aiexclude will affect inline assistance for covered files in the following ways:

Chat assistance: Gemini won't be able to answer questions or offer suggestions about files covered by .aiexclude.
Code completion: Suggested code completions will not be available when editing covered files.
Inline assistance: You'll be able to generate new code, but not modify existing code when editing covered files.

Other development environments such as Android Studio may also honor .aiexclude files.

How to write `.aiexclude` files

An .aiexclude file follows the same syntax as a .gitignore file, with these following differences:

An empty .aiexclude file blocks all files in its directory and all sub-directories. This is the same as a file that contains **/*.
.aiexclude files don't support negation (prefixing patterns with !).

Examples

Here are some example .aiexclude file configurations:

Block all files named apikeys.txt at or below the directory that contains the .aiexclude file:
```
apikeys.txt
```
Block all files with the .key file extension at or below the directory that contains the .aiexclude file:
```
*.key
```
Block only the apikeys.txt file at the in the same directory as the .aiexclude, but not any subdirectories:
```
/apikeys.txt
```
Block all files in the directory my/sensitive/dir and all subdirectories. The path should be relative to the directory that contains the .aiexclude file:
```
my/sensitive/dir/
```

Bring your own key: Use other Gemini models in chat

You can configure the Gemini model that Gemini in Firebase chat uses. You have a choice of the built-in model, models configured in the chat window (including Gemini 2.5 models), or any Gemini model to which you have access.

For a list of all available models, see Gemini models.

To configure your key and select a different Gemini model:

In your open workspace, click spark Gemini at the bottom of the workspace (or the Gemini tab).
In the Gemini in Firebase chat window, click the model name drop-down, then click the Gemini API key link. User Settings appear.
In the IDX > AI: Gemini Api Key field, enter your Gemini API key.

Note: This field may already be populated if you used the App Prototyping agent and it created a Firebase project and Gemini API key for you.

You can now select any of the pre-configured Gemini models in chat.

To configure a Gemini model that isn't in the drop-down:

Identify the Gemini model you want to use in chat from the list at Gemini models. For example, you'd enter gemini-2.0-flash-lite to use the latest stable Gemini 2.0 Flash‑Lite model.
From the Gemini in Firebase chat window, click the model selector, then choose Custom model ID. User Settings opens.
Copy the model name you selected into the IDX > AI: Gemini Model field.
Close the chat window, then re-open it by clicking spark Gemini at the bottom of the workspace to refresh the model list.

Next steps

Try Gemini in Firebase.