DocsAPI referenceGemini API

Gemini API

Google Gemini-compatible endpoints — generateContent, streamGenerateContent, countTokens.

POST/v1beta/models/{model}:generateContent

A chat endpoint compatible with the Google Gemini API request body. The model name and action live in the path ({model}:generateContent); authenticate with the x-goog-api-key header.

Google's official ?key= query parameter is not supported — the key must be in the x-goog-api-key header (the Google GenAI SDK uses that header by default).

Request body

Field	Type	Required	Description
`contents`	array	Yes	Conversation content: each item has a `role` (`user`/`model`) and `parts` (text or inline media).
`systemInstruction`	object	No	System prompt (`{ "parts": [{ "text": … }] }`).
`generationConfig`	object	No	Generation parameters: `temperature`, `maxOutputTokens`, etc.

request.sh

curl "https://soleapi.com/v1beta/models/gemini-3.7-flash:generateContent" \
  -H "x-goog-api-key: $SOLEAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "Write a haiku about the sea"}]}]
  }'

Response

response.json

{
  "candidates": [
    {
      "content": {
        "parts": [{ "text": "Waves fold into foam..." }],
        "role": "model"
      },
      "finishReason": "STOP"
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 14,
    "candidatesTokenCount": 23,
    "totalTokenCount": 37
  }
}

Streaming generation

POST/v1beta/models/{model}:streamGenerateContent

Append ?alt=sse to stream incrementally over SSE — each data: line is a response fragment; without alt=sse you get one complete JSON array. See Streaming.

Count tokens

POST/v1beta/models/{model}:countTokens

count_tokens.sh

curl "https://soleapi.com/v1beta/models/gemini-3.7-flash:countTokens" \
  -H "x-goog-api-key: $SOLEAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "Write a haiku about the sea"}]}]
  }'

# => { "totalTokens": 14 }

Image generation (Nano Banana)

Gemini image models (gemini-2.5-flash-image, gemini-3-pro-image, gemini-3.1-flash-image, gemini-3.1-flash-lite-image) have no separate image endpoint — they use the same generateContent / streamGenerateContent calls shown above. The request body is identical to a text conversation; just declare that you want images in generationConfig.responseModalities. Generated images come back as inlineData (base64) inside parts.

Field	Type	Description
`generationConfig.responseModalities`	array	`["TEXT", "IMAGE"]`; use `["IMAGE"]` for image-only output.
`generationConfig.imageConfig.aspectRatio`	string	`1:1`, `3:2`, `2:3`, `3:4`, `4:3`, `4:5`, `5:4`, `9:16`, `16:9`, `21:9`.
`generationConfig.imageConfig.imageSize`	string	`1K` (default), `2K`, `4K`; `gemini-3.1-flash-image` also accepts `0.5K`, `gemini-3.1-flash-lite-image` only `1K`, and `gemini-2.5-flash-image` does not support this field.

generate_image.sh

curl "https://soleapi.com/v1beta/models/gemini-3.1-flash-image:generateContent" \
  -H "x-goog-api-key: $SOLEAPI_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "contents": [{"parts": [{"text": "A watercolor lighthouse at dawn"}]}],
    "generationConfig": {
      "responseModalities": ["TEXT", "IMAGE"],
      "imageConfig": {"aspectRatio": "16:9", "imageSize": "2K"}
    }
  }'

response.json

{
  "candidates": [
    {
      "content": {
        "parts": [
          { "text": "Here is the lighthouse." },
          { "inlineData": { "mimeType": "image/png", "data": "iVBORw0KGgo..." } }
        ],
        "role": "model"
      },
      "finishReason": "STOP"
    }
  ],
  "usageMetadata": {
    "promptTokenCount": 9,
    "candidatesTokenCount": 1694,
    "candidatesTokensDetails": [
      { "modality": "TEXT", "tokenCount": 14 },
      { "modality": "IMAGE", "tokenCount": 1680 }
    ],
    "totalTokenCount": 1703
  }
}

Editing an image: put the reference image(s) as inlineData in contents[].parts (up to 14) together with the text instruction; the rest of the request is unchanged.

edit_request.json

{
  "contents": [{
    "parts": [
      { "text": "Turn the sky into a starry night" },
      { "inlineData": { "mimeType": "image/jpeg", "data": "<base64>" } }
    ]
  }],
  "generationConfig": { "responseModalities": ["TEXT", "IMAGE"] }
}

Images are billed by output tokens (the modality = IMAGE entry in usageMetadata.candidatesTokensDetails); see the model page for rates. A 1K image is roughly 1,120–1,290 tokens, 2K about 1,680, 4K about 2,000–2,520. Streaming works exactly as for text (streamGenerateContent?alt=sse); the image arrives whole in the final chunk.