Banana Proxy Image Generation
通过 Banana 代理端点调用 Gemini 生图。
Script Directory
Important: All scripts are located in the scripts/ subdirectory of this skill.
Agent Execution Instructions:
- Determine this SKILL.md file's directory path as
SKILL_DIR - Script path =
${SKILL_DIR}/scripts/main.ts - Replace all
${SKILL_DIR}in this document with the actual path
Usage
# 基础生图
npx -y bun ${SKILL_DIR}/scripts/main.ts --prompt "A cat" --image out.jpg
# 指定比例与质量
npx -y bun ${SKILL_DIR}/scripts/main.ts --prompt "portrait" --image out.jpg --ar 3:4 --quality 2k
# 使用参考图
npx -y bun ${SKILL_DIR}/scripts/main.ts --prompt "replace subject" --ref a.png b.png --image out.jpg
# 批量并行生图(JSONL)
npx -y bun ${SKILL_DIR}/scripts/main.ts --batch jobs.jsonl --concurrency 4
Options
--prompt <text>,-p: 提示词--promptfiles <files...>: 从文件读取提示词--image <path>: 输出路径(必填)--batch <file>: 批量任务文件(.json数组或.jsonl每行一个 JSON)--concurrency <n>: 并行数(仅--batch模式生效,默认4)--model <id>,-m: 模型(默认gemini-3-pro-image-preview,常用:gemini-3-pro-image-preview、gemini-3.1-flash-image-preview)--ar <ratio>: 比例,如1:1、3:4、16:9--quality normal|2k: 质量(默认2k)--imageSize 1K|2K|4K: 生成尺寸等级--ref <files...>: 参考图--json: JSON 输出
Batch 文件格式
JSONL 每行示例:
{"prompt":"A cat in watercolor style","image":"out/cat.jpg","ar":"1:1"}
{"prompt":"A cyberpunk city at night","image":"out/city.jpg","quality":"2k","imageSize":"2K"}
字段说明:
prompt或promptFile(二选一)image(必填)- 可选:
model、ar、quality、imageSize、ref(字符串数组)
Fixed Config
- Base URL:
https://lnapi.com - Primary API Key:
LNAPI_KEY(必填)