2.7 KiB
Profiles
Profiles allow you to maintain separate configurations for different workflows, tasks, or personas. Each profile can define its own set of allowed extensions, a default model, and a custom system instruction (persona).
Profiles are particularly useful for:
- Switching between domain-specific setups (e.g., frontend development vs. system administration).
- Defining specialized AI personas for different tasks.
- Restricting available tools/extensions for specific projects.
Profile structure
Profiles are Markdown files (.md) that use YAML frontmatter for configuration
and the file body for system instructions.
Location: ~/.gemini/profiles/[profile-name].md
Example profile (coder.md)
---
name: Web Developer
default_model: gemini-2.0-flash
extensions:
- github
- web-search
---
You are a world-class Web Developer. Focus on React, TypeScript, and modern CSS.
Always prefer functional components and explain your reasoning.
Configuration fields
| Field | Description |
|---|---|
name |
(Optional) A display name for the profile. |
default_model |
(Optional) The model to use when this profile is active. |
extensions |
(Optional) A list of allowed extension IDs. If specified, only these extensions will be loaded. |
The body of the Markdown file is injected into the system prompt, allowing you to define a specific persona or set of instructions that the AI will always follow when the profile is active.
Usage
Single session usage
Use the --profiles flag (or -P) to use a profile for a single session:
gemini --profiles coder
Persistent default
You can set a profile as your default for all sessions:
gemini profiles enable coder
To return to the standard configuration:
gemini profiles disable
Commands
| Command | Description |
|---|---|
gemini profiles list |
List all available profiles and see which one is active. |
gemini profiles enable <name> |
Set a profile as the persistent default. |
gemini profiles disable |
Clear the persistent default profile. |
gemini profiles uninstall <name> |
Delete a profile from your local storage. |