RaiInvocationModeration interface
声明请求/响应文本所在位置,以便内容安全保护栏能够提取。
属性
| input_content_type | 请求体是如何解析的。 省略时,服务默认为 |
| input_paths | 请求主体中通往用户文本的路径。 当input_content_type是 |
| output_content_type | 响应体是如何解析的。 省略时,服务默认为 |
| output_paths | 在 NON-STREAMING 响应体中访问代理文本的路径。 当response_mode是non_streaming/两者同时且output_content_type时 |
| response_mode | 作者声明的响应形状;决定哪个输出门运行以及哪些字段需要。 |
| stream_selectors | 每个事件类型一个带有文本的SSE事件>字段选择器。 当response_mode同时流媒体/同时使用且output_content_type是 |
属性详细信息
input_content_type
请求体是如何解析的。 省略时,服务默认为 json。
input_content_type?: RaiInvocationContentType
属性值
input_paths
请求主体中通往用户文本的路径。 当input_content_type是 json时的必修。
input_paths?: string[]
属性值
string[]
output_content_type
响应体是如何解析的。 省略时,服务默认为 json。
output_content_type?: RaiInvocationContentType
属性值
output_paths
在 NON-STREAMING 响应体中访问代理文本的路径。 当response_mode是non_streaming/两者同时且output_content_type时 json,必须。
output_paths?: string[]
属性值
string[]
response_mode
stream_selectors
每个事件类型一个带有文本的SSE事件>字段选择器。 当response_mode同时流媒体/同时使用且output_content_type是 json必需的。
stream_selectors?: RaiSseTextSelector[]