Multimodal generation needs to be realized in one of the following two ways:
1. OpenAI-compatible interfaces::
existchat.completions.createThe request adds thefilesParameters:{
"model": "gemini-2.5-pro",
"messages": [{"role": "user", "content": "描述图片内容"}],
"files": ["image.jpg"]
}
2. Native Gemini API::
existgenerateContentEndpoints construct multi-part requests:"parts": [
{"text": "描述这张图片"},
{"file_data": {"mime_type": "image/jpeg", "file_uri": "image.jpg"}}
]
Technical Details:
- Support for JPEG/PNG and other common formats
- Upload up to 10MB of content in a single request
- The image will be encoded as base64 before being transferred
This answer comes from the articlegeminicli2api: Proxy tool to convert Gemini CLI to OpenAI-compatible APIsThe
