Gemini CLI Cheat Sheet

Google's official CLI for Gemini models. Fast, scriptable, and designed for terminal workflows.

---

Installation

# npm
npm install -g @google/generative-ai-cli
yarn
yarn global add @google/generative-ai-cli
Set API key
export GEMINI_API_KEY="your-key-here"

---

Core Commands

Command	Description
gemini	Start interactive chat session
gemini "prompt"	One-shot prompt (prints response and exits)
gemini -f file.ts "explain this"	Include file as context
gemini -f *.ts "find bugs"	Include multiple files with glob
gemini chat	Explicit interactive mode
gemini models	List available models
gemini config	Show current configuration

---

Flags

Model Selection

Flag	Description
--model <name>	Choose model (gemini-pro, gemini-pro-vision, gemini-ultra)
--model-version <version>	Specific version (e.g., gemini-1.5-pro)

Input/Output

Flag	Description
-f, --file <path>	Include file(s) in context (supports globs)
--json	Output in JSON format
--stream	Stream response tokens (default)
--no-stream	Wait for complete response
--output <file>	Write response to file

Behavior

Flag	Description
--temperature <0-1>	Creativity control (0=focused, 1=creative)
--top-p <0-1>	Nucleus sampling threshold
--top-k <number>	Top-k sampling limit
--max-tokens <number>	Maximum response length
--system <message>	System instruction/persona

Advanced

Flag	Description
--api-key <key>	Override API key (instead of env var)
--timeout <ms>	Request timeout in milliseconds
--verbose	Show debug information
--no-color	Disable colored output
--help	Show help message

---

Slash Commands (Interactive Mode)

Command	Description
/help	Show available commands
/clear	Clear conversation history
/history	Show conversation history
/save <file>	Save conversation to file
/load <file>	Load conversation from file
/model <name>	Switch model mid-conversation
/temp <0-1>	Change temperature
/system <message>	Set system instruction
/exit	Exit interactive mode

---

Configuration

Config File Location

• Linux/macOS: ~/.config/gemini-cli/config.json
• Windows: %APPDATA%/gemini-cli/config.json

Example config.json

{
  "model": "gemini-1.5-pro",
  "temperature": 0.7,
  "maxTokens": 2048,
  "stream": true,
  "defaultSystemMessage": "You are a helpful coding assistant."
}

Set Config Values

gemini config set model gemini-1.5-pro
gemini config set temperature 0.8
gemini config set maxTokens 4096

---

Examples

One-Shot Prompts

# Simple question
gemini "Explain async/await in JavaScript"
Code generation
gemini "Write a Python function to merge two sorted arrays"
With file context
gemini -f app.ts "Explain this code"
gemini -f src/*/.ts "Find potential bugs"

Interactive Sessions

# Start interactive mode
gemini
Multi-turn conversation
> Write a REST API in Express
> Add authentication with JWT
> Add rate limiting
> /save api-conversation.json

Scripting & Automation

# JSON output for parsing
gemini --json "List 5 Linux commands" | jq '.response'
Generate code and save
gemini "Create a React component for a login form" --output Login.tsx
Batch processing
for file in src/*.ts; do
  gemini -f "$file" "Add JSDoc comments" --output "$file.new"
done

---

Extensions & Plugins

Vision Support (gemini-pro-vision)

# Analyze image
gemini --model gemini-pro-vision -f screenshot.png "What's in this image?"
Compare images
gemini --model gemini-pro-vision -f before.png -f after.png "What changed?"

Function Calling

// Define functions in config
{
  "functions": [
    {
      "name": "getCurrentWeather",
      "description": "Get current weather for a location",
      "parameters": {
        "type": "object",
        "properties": {
          "location": { "type": "string" }
        }
      }
    }
  ]
}

Code Execution

# Enable code execution (experimental)
gemini --enable-code-execution "Calculate fibonacci(50)"

---

Tips & Best Practices

Performance

• Use --no-stream for scripting to get complete responses
• Set --max-tokens to limit costs
• Use gemini-1.5-flash for faster, cheaper responses

Accuracy

• Lower temperature (0.2-0.4) for factual/code tasks
• Higher temperature (0.7-0.9) for creative writing
• Use --system to set persona/constraints

File Context

• Gemini CLI supports glob patterns: -f src/*/.ts
• Large files are automatically chunked
• Use --json output for parsing in scripts

Cost Management

# Count tokens before sending
wc -w prompt.txt  # Rough estimate
Use flash model for simple tasks
gemini --model gemini-1.5-flash "Quick question"

---

Troubleshooting

API Key Issues

# Verify key is set
echo $GEMINI_API_KEY
Test with explicit key
gemini --api-key "your-key" "Test prompt"

Rate Limiting

• Default: 60 requests/minute (gemini-pro)
• Use --timeout to handle retries
• Upgrade API tier for higher limits

File Not Found

# Check file path
ls -la file.ts
Use absolute paths if needed
gemini -f /absolute/path/to/file.ts "Explain"

---

Resources

• Official Docs: https://ai.google.dev/docs
• API Reference: https://ai.google.dev/api
• Pricing: https://ai.google.dev/pricing
• GitHub: https://github.com/google/generative-ai-cli