๐๐ฒ๐บ๐ถ๐ป๐ถ ๐๐๐ ๐ง๐ฒ๐๐ ๐ฅ๐ฒ๐ฝ๐ผ๐ฟ๐
I tested the Gemini CLI tool on both macOS and Ubuntu. I ran 12 different tests to see if the tool works as promised.
Here are the results from my testing:
- Installation: Works perfectly using npm. If you lack sudo rights on remote servers, use a user-level npm prefix.
- Headless Mode: The -p flag works for direct prompts.
- Data Piping: You can pipe error logs or git diffs directly into the CLI. This is great for quick analysis.
- Commit Messages: The tool generates high-quality Conventional Commits from git diffs.
- JSON Output: Use the --output-format json flag to get structured data. This works well with jq.
- Sandbox Mode: The tool restricts shell commands but allows file reading. This keeps your system safe.
- Project Context: The CLI automatically reads GEMINI.md files. It uses this to understand your project rules.
- Batch Processing: The @file syntax allows you to generate documentation for multiple files at once.
Important notes on API limits:
The free tier limits depend on how you authenticate.
- OAuth users get higher limits.
- API Key users face strict limits.
- Gemini-2.5-flash via API key allows only 20 requests per day.
- Other models like Gemini-3.1-pro have zero quota on the free API key tier.
Always specify the model with -m gemini-2.5-flash to avoid errors.
Summary: The tool is reliable for engineering tasks. It handles context and automation well. Just watch your daily API quota.
Source: https://dev.to/jh5_pulse/wei-lai-ying-gai-gai-yong-google-magika-lai-pan-duan-dang-an-lei-xing--1fd3
Optional learning community: https://t.me/GyaanSetuAi