mingyan

Gemini Native API

Gemini native interface for direct generateContent and streamGenerateContent integrations

POST/gw_llm_power/v1/models/{model}:generateContent

baseUrlhttp://apillm.globalaiopc.com/gw_llm_power

Compatible with the Gemini official documentation. You can integrate directly with the following reference:

Gemini generateContent: https://ai.google.dev/api/generate-content

Authentication

Get Key

Use the Authorization header for authentication.

cURL

Authorization: Bearer {{YOUR_API_KEY}}
Content-Type: application/json

Notes:

Authorization: API key in the format Bearer {{YOUR_API_KEY}}

Supported Models

The following models are supported, organized by model family and version type:

Gemini

Standard / Stable Version	Official Version	Low-cost Version
`gemini-2.5-flash-lite`	-	-
`gemini-2.5-pro`	`gemini-2.5-pro-official`	`gemini-2.5-pro-low`
`gemini-3-flash-preview`	`gemini-3-flash-preview-official`	`gemini-3-flash-preview-low`
`gemini-3.1-flash-lite-preview`	-	-
`gemini-3.1-pro-preview`	`gemini-3.1-pro-preview-official`	`gemini-3.1-pro-preview-low`

Version Notes

Suffix / Type	Description
no suffix	Standard / stable version
`-official`	Official version
`-low`	Low-cost version

Multimodal Support

Image Analysis

All models support image analysis.

Video Analysis

All models support video analysis.

Request Path

The following Gemini native paths are supported:

POST /v1/models/{model}:generateContent
POST /v1/models/{model}:streamGenerateContent

Example full URL:

Text

http://apillm.globalaiopc.com/gw_llm_power/v1/models/gemini-3.1-pro-preview:generateContent

Request Parameters

contentsarrayrequired

Gemini native message content array. Each item usually contains role and parts

contents[].rolestring

Message role, commonly user or model

contents[].partsarrayrequired

Message part array. Commonly text objects, also supports Gemini native multimodal structures

contents[].parts[].textstring

Text content

generationConfig.temperaturenumber

Sampling temperature

generationConfig.topPnumber

Nucleus sampling parameter

generationConfig.maxOutputTokensinteger

Maximum output token count

systemInstruction.parts[].textstring

Optional system prompt

Response

Returns a Gemini native-style response.

candidates[].content.parts[].textstring

Generated text content

candidates[].finishReasonstring

Generation stop reason

usageMetadata.promptTokenCountinteger

Input token count

usageMetadata.candidatesTokenCountinteger

Output token count

usageMetadata.totalTokenCountinteger

Total token count

Notes

Replace {model} in the path with the actual model name, for example gemini-3.1-pro-preview
For streaming responses, use :streamGenerateContent
Pass the API key using the Authorization header