happy gemini

Start Gemini agent powered by Google AI models

The happy gemini command starts a Gemini-powered AI agent using Google’s Gemini models for coding assistance and task automation.

Syntax

happy gemini [options]
happy gemini model <subcommand>
happy gemini project <subcommand>

Description

Gemini provides AI-powered coding assistance using Google’s advanced language models. Features include:

• Code generation and refactoring
• Natural language task execution
• Mobile control and remote sessions
• Google Workspace integration
• Configurable models and projects

You need to authenticate with Google first using happy connect gemini.

Commands

Start Gemini Agent

happy gemini [options]

Starts an interactive Gemini agent session.

--started-by · string

Session origin indicator. Options: daemon or terminal. Internal use

Model Management

Set Model

happy gemini model set <model>

Configure which Gemini model to use for future sessions.

model · string · required

Model name to use. Available options:

• gemini-2.5-pro (default, most capable)
• gemini-2.5-flash (faster, efficient)
• gemini-2.5-flash-lite (lightweight)

Get Model

happy gemini model get

Display the currently configured model.

Project Management

Set Google Cloud Project

happy gemini project set <project-id>

Configure Google Cloud Project ID for Google Workspace accounts.

project-id · string · required

Your Google Cloud Project ID (required for Workspace accounts)

Google Workspace accounts require a Google Cloud Project ID. Consumer Gmail accounts do not need this.

Get Project

happy gemini project get

Display the currently configured Google Cloud Project ID.

Setup

1. Connect Google Account

happy connect gemini

This initiates OAuth authentication with Google and securely stores your credentials.

2. Configure Model (Optional)

happy gemini model set gemini-2.5-flash

Set your preferred model. Defaults to gemini-2.5-pro if not configured.

3. Set Project ID (Workspace Only)

If you’re using a Google Workspace account:

happy gemini project set your-project-id

Find your project ID in the Google Cloud Console. See the Workspace authentication guide for details.

Examples

Start Gemini Session

happy gemini

Launches Gemini with your configured settings.

Configure Model

Set to Flash (Faster)

happy gemini model set gemini-2.5-flash

Setup for Workspace Account

# 1. Connect account
happy connect gemini

# 2. Set your GCP project
happy gemini project set my-workspace-project

# 3. Verify configuration
happy gemini project get

# 4. Start Gemini
happy gemini

Check Configuration

happy gemini model get

Output:

Current model: gemini-2.5-flash

happy gemini project get

Output:

Current Google Cloud Project: my-project-123
  Linked to account: [e-mail]

Output

When you start Gemini, you’ll see an interactive terminal interface:

Starting Gemini agent...

[Gemini Display]
╭─ Messages ─────────────────────────────────────╮
│                                                 │
│ User: Write a Python script to process CSV     │
│ Assistant: I'll create that script for you...   │
│                                                 │
╰─────────────────────────────────────────────────╯

Model: gemini-2.5-pro
Status: Ready

Configuration Files

Model Configuration

Model settings are stored in:

• ~/.gemini/config.json
• ~/.config/gemini/config.json (alternative location)

{
  "model": "gemini-2.5-flash"
}

Project Configuration

Google Cloud Project settings:

{
  "googleCloudProject": "my-project-123",
  "googleCloudProjectEmail": "[e-mail]"
}

OAuth Credentials

Google OAuth tokens are stored in:

• ~/.gemini/oauth_creds.json

These are automatically managed by happy connect gemini.

Environment Variables

GEMINI_MODEL · string

Override the model configuration

GEMINI_MODEL=gemini-2.5-flash happy gemini

GOOGLE_CLOUD_PROJECT · string

Override the project configuration

GOOGLE_CLOUD_PROJECT=my-project happy gemini

Features

Permission Modes

Gemini supports the same permission modes as other agents:

• default: Ask before executing
• effective: Auto-approve safe operations
• bypassPermissions: Auto-approve all

Reasoning Display

Gemini shows its thought process:

[Thinking] Analyzing the requirements...
[Tool] Running: cat example.py
[Result] File contents retrieved

Diff Tracking

File modifications are tracked and displayed:

--- example.py
+++ example.py
@@ -1,3 +1,4 @@
 import sys
+import json
 
 def main():

Session Resume

Gemini automatically maintains conversation context across turns.

Troubleshooting

Authentication Required Error

If you see “Authentication required”:

# For Workspace accounts, set project ID
happy gemini project set your-project-id

# Re-authenticate
happy connect gemini

See the Workspace setup guide.

Invalid Model Error

If you specify an invalid model:

happy gemini model set invalid-model

Output:

Invalid model: invalid-model
Available models: gemini-2.5-pro, gemini-2.5-flash, gemini-2.5-flash-lite

Connection Status

Check if Gemini is properly connected:

happy connect status

Output:

  ✓  Google Gemini: connected ([e-mail])

Token Expired

If your OAuth token expires:

happy connect gemini

Re-authenticate to refresh the token.

View Logs

happy doctor

Check logs in ~/.happy/logs/ for detailed error information.

Model Comparison

Model	Speed	Capability	Best For
gemini-2.5-pro	Slower	Highest	Complex tasks, reasoning
gemini-2.5-flash	Fast	High	Most coding tasks
gemini-2.5-flash-lite	Fastest	Good	Simple queries, quick edits

Related Commands

• happy connect - Connect Google account
• happy acp - Alternative way to run Gemini
• happy codex - OpenAI-powered agent
• happy sandbox - Configure sandboxing
• happy doctor - System diagnostics

Gemini uses your Google account for authentication. Usage may be subject to Google’s API quotas and pricing.