Back to Details

gemini

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Gemini CLI

Gemini CLI

Commands

命令

Use the wrapper script for streamlined interaction:
bash
undefined
使用包装脚本实现流畅交互：
bash
undefined

New conversation

新对话

~/.agents/skills/gemini/gemini-chat.sh "your prompt here"
~/.agents/skills/gemini/gemini-chat.sh "your prompt here"

Continue conversation (session_id from previous output)

继续对话（会话ID来自上一次输出）

~/.agents/skills/gemini/gemini-chat.sh --resume <session_id> "follow-up prompt"

The script:
- Streams output in real-time (no more waiting blindly)
- Shows retry/error info (429 등)
- Saves raw response to tmpfile for debugging
- Extracts session_id automatically
~/.agents/skills/gemini/gemini-chat.sh --resume <session_id> "follow-up prompt"

该脚本：
- 实时流式输出（无需盲目等待）
- 显示重试/错误信息（如429等）
- 将原始响应保存到临时文件以便调试
- 自动提取session_id

Examples

示例

bash
undefined
bash
undefined

Ask Gemini for code review

请求Gemini进行代码审查

~/.agents/skills/gemini/gemini-chat.sh "Review this function for potential bugs: $(cat src/utils.ts)"
~/.agents/skills/gemini/gemini-chat.sh "Review this function for potential bugs: $(cat src/utils.ts)"

Follow up (use session_id from previous output)

跟进对话（使用上一次输出的session_id）

~/.agents/skills/gemini/gemini-chat.sh --resume abc-123 "How would you refactor it?"
~/.agents/skills/gemini/gemini-chat.sh --resume abc-123 "How would you refactor it?"

Leverage Google Search grounding

利用Google Search grounding功能

~/.agents/skills/gemini/gemini-chat.sh "What are the latest changes in TypeScript 5.8?"
undefined
~/.agents/skills/gemini/gemini-chat.sh "What are the latest changes in TypeScript 5.8?"
undefined

Raw CLI Usage (if needed)

原始CLI使用（如有需要）

bash
tmpfile=$(mktemp)
echo "Response saved to: $tmpfile"
gemini --model auto-gemini-3 --approval-mode yolo "prompt" --output-format stream-json | tee "$tmpfile"
session_id=$(jq -rs 'last | .session_id' "$tmpfile")
bash
tmpfile=$(mktemp)
echo "Response saved to: $tmpfile"
gemini --model auto-gemini-3 --approval-mode yolo "prompt" --output-format stream-json | tee "$tmpfile"
session_id=$(jq -rs 'last | .session_id' "$tmpfile")

Model Selection

模型选择

Available 
--model
options:
ValueDescription
auto-gemini-3
기본값. gemini-3-pro, gemini-3-flash 중 자동 선택
auto-gemini-2.5
gemini-2.5-pro, gemini-2.5-flash 중 자동 선택
gemini-2.5-flash
직접 지정 (가장 안정적)
gemini-2.5-pro
직접 지정
可用的
--model
选项：
值描述
auto-gemini-3
默认值。自动选择gemini-3-pro或gemini-3-flash
auto-gemini-2.5
自动选择gemini-2.5-pro或gemini-2.5-flash
gemini-2.5-flash
手动指定（最稳定）
gemini-2.5-pro
手动指定

Quota Exhausted (429 에러) 대처

配额耗尽（429错误）应对

Gemini 3 quota 초과 시 (429, 
MODEL_CAPACITY_EXHAUSTED
, 
RESOURCE_EXHAUSTED
):
bash
undefined
当Gemini 3配额耗尽时（错误码429，
MODEL_CAPACITY_EXHAUSTED
、
RESOURCE_EXHAUSTED
）：
bash
undefined

auto-gemini-2.5로 fallback (2.5-pro 또는 2.5-flash 자동 선택)

回退到auto-gemini-2.5（自动选择2.5-pro或2.5-flash）

~/.agents/skills/gemini/gemini-chat.sh --model auto-gemini-2.5 "your prompt"

**Fallback 순서:**
1. `auto-gemini-3` (기본) → Gemini 3 계열 사용
2. `auto-gemini-2.5` (fallback) → Gemini 2.5-pro/flash 중 자동 선택
~/.agents/skills/gemini/gemini-chat.sh --model auto-gemini-2.5 "your prompt"

**回退顺序：**
1. `auto-gemini-3`（默认）→ 使用Gemini 3系列模型
2. `auto-gemini-2.5`（回退选项）→ 自动选择Gemini 2.5-pro/flash

Notes

注意事项

  • Gemini automatically uses 
    google_web_search
    when needed
  • Image analysis: include file path in prompt, Gemini reads via 
    read_file
  • Raw response file remains after execution for debugging
  • Gemini会在需要时自动使用
    google_web_search
    功能
  • 图片分析：在提示词中包含文件路径，Gemini将通过
    read_file
    读取
  • 原始响应文件在执行后会保留，用于调试