> ## Documentation Index
> Fetch the complete documentation index at: https://mixroute.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.
# Gemini 原生（圖片，串流）
> Gemini 原生圖片流式請求與響應處理。
使用 Gemini 原生 streamGenerateContent 端點流式獲取圖片響應，請求欄位與 generateContent 相同。
`POST https://api.mixroute.ai/v1/models/{model}:streamGenerateContent?alt=sse`
## 支援模型
| 模型 ID | 能力與限制 |
| - | - |
| `gemini-nano-banana-2.1` | Nano Banana 2.1；1K/2K/4K 輸出；MINIMAL/MEDIUM/HIGH 思考，預設 MEDIUM。 |
| `gemini-3.1-flash-image` | Nano Banana 2，支援 512、1K、2K、4K 輸出及 MINIMAL/HIGH 思考等級。 |
| `gemini-3.1-flash-image-preview` | Nano Banana 2，支援 512、1K、2K、4K 輸出及 MINIMAL/HIGH 思考等級。 |
| `gemini-3-pro-image` | Nano Banana Pro，支援 1K、2K、4K 輸出及圖文推理。 |
| `gemini-3-pro-image-preview` | Nano Banana Pro，支援 1K、2K、4K 輸出及圖文推理。 |
| `gemini-3.1-flash-lite-image` | Nano Banana 2 Lite，僅 1K 輸出，不支援 Google Search 搜尋增強。 |
| `gemini-2.5-flash-image` | 初代 Nano Banana，約 1K 輸出，不傳 imageSize。 |
請求體欄位與模型限制參見 [Gemini 圖片生成](/docs/zh-hant/api-reference/endpoint/nano-banana)。在查詢字串中使用 `alt=sse`，不新增請求體 stream 參數。
## 呼叫示例
將 MixRoute Key 設定到環境變數 `MIXROUTE_API_KEY`，通過 `Authorization: Bearer ...` 認證。編輯圖片前，將素材佔位值替換為可訪問的圖片或本地檔案。
```bash theme={null}
curl --no-buffer --request POST "https://api.mixroute.ai/v1/models/gemini-nano-banana-2.1:streamGenerateContent?alt=sse" \
--header "Authorization: Bearer $MIXROUTE_API_KEY" \
--header "Content-Type: application/json" \
--data '{
"contents": [
{
"role": "user",
"parts": [
{
"text": "A clean product photograph of a red ceramic mug on a white background."
}
]
}
],
"generationConfig": {
"responseModalities": [
"TEXT",
"IMAGE"
],
"imageConfig": {
"aspectRatio": "1:1",
"imageSize": "1K"
}
}
}'
```
### Python
```python theme={null}
import base64
import json
import os
from pathlib import Path
import requests
payload = json.loads(r'''
{
"contents": [
{
"role": "user",
"parts": [
{
"text": "A clean product photograph of a red ceramic mug on a white background."
}
]
}
],
"generationConfig": {
"responseModalities": [
"TEXT",
"IMAGE"
],
"imageConfig": {
"aspectRatio": "1:1",
"imageSize": "1K"
}
}
}
''')
def events(lines):
data = []
for line in lines:
if line == "":
if data:
yield "\n".join(data)
data.clear()
elif line.startswith("data:"):
value = line[5:]
data.append(value[1:] if value.startswith(" ") else value)
if data:
yield "\n".join(data)
url = "https://api.mixroute.ai/v1/models/gemini-nano-banana-2.1:streamGenerateContent?alt=sse"
image_index = 0
with requests.post(url, headers={"Authorization": "Bearer " + os.environ["MIXROUTE_API_KEY"]}, json=payload, stream=True, timeout=(30, 180)) as response:
response.raise_for_status()
response.encoding = "utf-8"
for data in events(response.iter_lines(decode_unicode=True)):
if not data.strip():
continue
if data.strip() == "[DONE]":
break
chunk = json.loads(data)
if "error" in chunk:
raise RuntimeError(chunk["error"])
for candidate in chunk.get("candidates", []):
for part in candidate.get("content", {}).get("parts", []):
if part.get("thought"):
continue
if "inlineData" in part:
inline = part["inlineData"]
if inline.get("data"):
extension = {"image/png": "png", "image/jpeg": "jpg", "image/webp": "webp"}.get(inline.get("mimeType"), "bin")
Path(f"image_{image_index}.{extension}").write_bytes(base64.b64decode(inline["data"], validate=True))
image_index += 1
elif "text" in part:
print(part["text"], end="")
```
## 響應
從 candidates\[].content.parts\[] 讀取文本和圖片。圖片通常使用 inlineData 中的 mimeType 與 Base64 data，多輪複用模型內容時保留 thoughtSignature。沒有圖片時檢查 promptFeedback 與 finishReason；usageMetadata 是用量統計，不是圖片生成參數。
[Gemini 同步圖片請求](/docs/zh-hant/api-reference/endpoint/nano-banana)
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.