Skip to main content

Gemini - Google AI Studio

前提条件​

Gemini-Pro

示例用法​

from litellm import completion
import os

os.environ['GEMINI_API_KEY'] = ""
response = completion(
    model="gemini/gemini-pro", 
    messages=[{"role": "user", "content": "写代码以从 LiteLLM 打招呼"}]
)

支持的 OpenAI 参数​

  • temperature
  • top_p
  • max_tokens
  • stream
  • tools
  • tool_choice
  • response_format
  • n
  • stop

查看更新列表

传递 Gemini 特定参数​

响应架构​

LiteLLM 支持将 response_schema 作为参数发送给 Google AI Studio 上的 Gemini-1.5-Pro。

响应架构

from litellm import completion 
import json 
import os 

os.environ['GEMINI_API_KEY'] = ""

messages = [
    {
        "role": "user",
        "content": "列出 5 个流行的饼干食谱。"
    }
]

response_schema = {
        "type": "array",
        "items": {
            "type": "object",
            "properties": {
                "recipe_name": {
                    "type": "string",
                },
            },
            "required": ["recipe_name"],
        },
    }


completion(
    model="gemini/gemini-1.5-pro", 
    messages=messages, 
    response_format={"type": "json_object", "response_schema": response_schema} # 👈 关键更改
    )

print(json.loads(completion.choices[0].message.content))

验证架构

要验证 response_schema，请设置 enforce_validation: true。

from litellm import completion, JSONSchemaValidationError
try: 
    completion(
    model="gemini/gemini-1.5-pro", 
    messages=messages, 
    response_format={
        "type": "json_object", 
        "response_schema": response_schema,
        "enforce_validation": true # 👈 关键更改
    }
    )
except JSONSchemaValidationError as e: 
    print("原始响应: {}".format(e.raw_response))
    raise e

LiteLLM 将根据架构验证响应，如果响应不符合架构，将引发 JSONSchemaValidationError。

JSONSchemaValidationError 继承自 openai.APIError

使用 e.raw_response 访问原始响应

GenerationConfig 参数​

要传递额外的 GenerationConfig 参数 - 例如 topK，只需将其作为键值对传递到请求体中，LiteLLM 将直接通过请求体传递。

查看 Gemini GenerationConfigParams

from litellm import completion 
import json 
import os 

os.environ['GEMINI_API_KEY'] = ""

messages = [
    {
        "role": "user",
        "content": "列出5种受欢迎的饼干食谱。"
    }
]

completion(
    model="gemini/gemini-1.5-pro", 
    messages=messages, 
    topK=1 # 👈 关键变化
)

print(json.loads(completion.choices[0].message.content))

验证模式

要验证response_schema，设置 enforce_validation: true。

from litellm import completion, JSONSchemaValidationError
try: 
completion(
    model="gemini/gemini-1.5-pro", 
    messages=messages, 
    response_format={
        "type": "json_object", 
        "response_schema": response_schema,
        "enforce_validation": true # 👈 关键更改
    }
)
except JSONSchemaValidationError as e: 
print("原始响应: {}".format(e.raw_response))
raise e

指定安全设置​

在某些用例中，您可能需要调用模型并传递与默认值不同的安全设置。为此，只需将 safety_settings 参数传递给 completion 或 acompletion。例如：

response = completion(
    model="gemini/gemini-pro", 
    messages=[{"role": "user", "content": "为LiteLLM写代码以打招呼"}],
    safety_settings=[
        {
            "category": "HARM_CATEGORY_HARASSMENT",
            "threshold": "BLOCK_NONE",
        },
        {
            "category": "HARM_CATEGORY_HATE_SPEECH",
            "threshold": "BLOCK_NONE",
        },
        {
            "category": "HARM_CATEGORY_SEXUALLY_EXPLICIT",
            "threshold": "BLOCK_NONE",
        },
        {
            "category": "HARM_CATEGORY_DANGEROUS_CONTENT",
            "threshold": "BLOCK_NONE",
        },
    ]
)

工具调用​

from litellm import completion
import os
# 设置环境变量
os.environ["GEMINI_API_KEY"] = ".."

tools = [
    {
        "type": "function",
        "function": {
            "name": "get_current_weather",
            "description": "获取给定位置的当前天气",
            "parameters": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "城市和州，例如：San Francisco, CA",
                    },
                    "unit": {"type": "string", "enum": ["celsius", "fahrenheit"]},
                },
                "required": ["location"],
            },
        },
    }
]
messages = [{"role": "user", "content": "今天波士顿的天气怎么样？"}]

response = completion(
    model="gemini/gemini-1.5-flash",
    messages=messages,
    tools=tools,
)
# 添加任何断言，这里检查响应参数
print(response)
assert isinstance(response.choices[0].message.tool_calls[0].function.name, str)
assert isinstance(
    response.choices[0].message.tool_calls[0].function.arguments, str
)

JSON模式​

from litellm import completion 
import json 
import os 

os.environ['GEMINI_API_KEY'] = ""

messages = [
    {
        "role": "user",
        "content": "列出5个受欢迎的曲奇饼干食谱。"
    }
]



completion(
    model="gemini/gemini-1.5-pro", 
    messages=messages, 
    response_format={"type": "json_object"} # 👈 关键更改
)

print(json.loads(completion.choices[0].message.content))
# Gemini-Pro-Vision LiteLLM 支持以下通过 `url` 传递的图像类型 - 直接链接的图像 - https://storage.googleapis.com/github-repo/img/gemini/intro/landmark3.jpg - 本地存储的图像 - ./localimage.jpeg

示例用法​

import os
import litellm
from dotenv import load_dotenv

# 从 .env 文件加载环境变量
load_dotenv()
os.environ["GEMINI_API_KEY"] = os.getenv('GEMINI_API_KEY')

prompt = '用几句话描述这张图片。'
# 注意：你可以直接传递图像的URL或路径。
image_url = 'https://storage.googleapis.com/github-repo/img/gemini/intro/landmark3.jpg'

# 根据文档创建消息负载
messages = [
    {
        "role": "user",
        "content": [
            {
                "type": "text",
                "text": prompt
            },
            {
                "type": "image_url",
                "image_url": {"url": image_url}
            }
        ]
    }
]

# 向Gemini模型发起API调用
response = litellm.completion(
    model="gemini/gemini-pro-vision",
    messages=messages,
)

# 提取响应内容
content = response.get('choices', [{}])[0].get('message', {}).get('content')

# 打印结果
print(content)

上下文缓存​

使用Google AI Studio上下文缓存，支持在消息内容块中添加以下内容：

{
    ...,
    "cache_control": {"type": "ephemeral"}
}
note

Gemini上下文缓存仅允许缓存1个连续的消息块。

向Gemini发起的原始请求如下所示：

curl -X POST "https://generativelanguage.googleapis.com/v1beta/models/gemini-1.5-flash-001:generateContent?key=$GOOGLE_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
      "contents": [
        {
          "parts":[{
            "text": "请总结这段文字"
          }],
          "role": "user"
        },
      ],
      "cachedContent": "'$CACHE_NAME'"
    }'

from litellm import completion 

for _ in range(2): 
    resp = completion(
        model="gemini/gemini-1.5-pro",
        messages=[
        # 系统消息
            {
                "role": "system",
                "content": [
                    {
                        "type": "text",
                        "text": "这里是复杂法律协议的全文" * 4000,
                        "cache_control": {"type": "ephemeral"}, # 👈 关键变化
                    }
                ],
            },
            # 使用cache_control参数标记为缓存，以便此检查点可以读取之前的缓存。
            {
                "role": "user",
                "content": [
                    {
                        "type": "text",
                        "text": "这个协议中的关键条款和条件是什么？",
                        "cache_control": {"type": "ephemeral"},
                    }
                ],
            }]
    )

    print(resp.usage) # 👈 第二个使用块将减少，因为使用了缓存令牌

使用 - PDF / 视频 / 等文件​

内联数据（例如音频流）​

LiteLLM遵循OpenAI的格式，接受发送内联数据作为base64编码的字符串。

遵循的格式为

data:<mime_type>;base64,<encoded_data>

LITELLM调用 

import litellm
from pathlib import Path
import base64
import os

os.environ["GEMINI_API_KEY"] = "" 

litellm.set_verbose = True # 👈 查看原始调用

audio_bytes = Path("speech_vertex.mp3").read_bytes()
encoded_data = base64.b64encode(audio_bytes).decode("utf-8")
print("Audio Bytes = {}".format(audio_bytes))
model = "gemini/gemini-1.5-flash"
response = litellm.completion(
    model=model,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Please summarize the audio."},
                {
                    "type": "image_url",
                    "image_url": "data:audio/mp3;base64,{}".format(encoded_data), # 👈 设置MIME类型+数据
                },
            ],
        }
    ],
)

等效的GOOGLE API调用 

# 初始化适合您用例的Gemini模型。
model = genai.GenerativeModel('models/gemini-1.5-flash')

# 创建提示。
prompt = "Please summarize the audio."

# 将samplesmall.mp3文件加载到包含音频文件字节的Python Blob对象中，然后将提示和音频传递给Gemini。
response = model.generate_content([
    prompt,
    {
        "mime_type": "audio/mp3",
        "data": pathlib.Path('samplesmall.mp3').read_bytes()
    }
])

# 输出Gemini对提示和内联音频的响应。
print(response.text)

https:// 文件​

import litellm
import os

os.environ["GEMINI_API_KEY"] = "" 

litellm.set_verbose = True # 👈 查看原始调用

model = "gemini/gemini-1.5-flash"
response = litellm.completion(
    model=model,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Please summarize the file."},
                {
                    "type": "image_url",
                    "image_url": "https://storage..." # 👈 设置图片URL
                },
            ],
        }
    ],
)

gs:// 文件​

import litellm
import os

os.environ["GEMINI_API_KEY"] = "" 

litellm.set_verbose = True # 👈 查看原始调用

model = "gemini/gemini-1.5-flash"
response = litellm.completion(
    model=model,
    messages=[
        {
            "role": "user",
            "content": [
                {"type": "text", "text": "Please summarize the file."},
                {
                    "type": "image_url",
                    "image_url": "gs://..." # 👈 设置云存储桶URL
                },
            ],
        }
    ],
)

聊天模型​

tip

我们支持所有Gemini模型，只需在发送litellm请求时将model=gemini/<任何Gemini上的模型>作为前缀

模型名称函数调用必需的操作系统变量
gemini-procompletion(model='gemini/gemini-pro', messages)os.environ['GEMINI_API_KEY']
gemini-1.5-pro-latestcompletion(model='gemini/gemini-1.5-pro-latest', messages)os.environ['GEMINI_API_KEY']
gemini-pro-visioncompletion(model='gemini/gemini-pro-vision', messages)os.environ['GEMINI_API_KEY']