Exécuter Gemma avec l'API Gemini

L'API Gemini fournit un accès hébergé à Gemma en tant qu'API de programmation que vous pouvez utiliser dans le développement ou le prototypage d'applications. Cette API est une alternative pratique à la configuration de votre propre instance locale de Gemma et d'un service Web pour gérer les tâches d'IA générative.

Modèles compatibles

L'API Gemini est compatible avec les modèles Gemma 4 suivants :

  • gemma-4-31b-it
  • gemma-4-26b-a4b-it

L'exemple suivant montre comment utiliser Gemma avec l'API Gemini :

Python

from google import genai

client = genai.Client()

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    contents="Roses are red...",
)

print(response.text)

JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI();

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: "Roses are red...",
});
console.log(response.text);

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "Roses are red..."}]
    }]
   }'

Obtenir une clé API

Vous pouvez accéder à l'API Gemini sur de nombreuses plates-formes, telles que les services mobiles, Web et cloud, et avec plusieurs langages de programmation. Pour en savoir plus sur les packages SDK de l'API Gemini, consultez la page Téléchargements du SDK de l'API Gemini. Pour une présentation générale de l'API Gemini, consultez le guide de démarrage rapide de l'API Gemini.

Raisonnement

Gemma 4 utilise un "processus de réflexion" interne qui optimise son raisonnement en plusieurs étapes, offrant des performances supérieures dans les domaines logiquement exigeants tels que le codage algorithmique et les preuves mathématiques avancées.

Bien que Gemma 4 soit compatible avec l'activation ou la désactivation de cette fonctionnalité, vous pouvez l'activer dans l'API en définissant le niveau de réflexion sur "high".

L'exemple suivant montre comment activer le processus de réflexion :

Python

from google import genai
from google.genai import types

client = genai.Client()

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    contents="What is the water formula?",
    config=types.GenerateContentConfig(
        thinking_config=types.ThinkingConfig(thinking_level="high")
    ),
)

print(response.text)

JavaScript

import { GoogleGenAI, ThinkingLevel } from "@google/genai";

const ai = new GoogleGenAI();

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: "What is the water formula?",
  config: {
    thinkingConfig: {
      thinkingLevel: ThinkingLevel.HIGH,
    },
  },
});
console.log(response.text);

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "What is the water formula?"}]
    }],
    "generationConfig": {
      "thinkingConfig": {
            "thinkingLevel": "high"
      }
    }
   }'

En savoir plus sur le raisonnement :

Compréhension des images

Les modèles Gemma 4 peuvent traiter des images, ce qui permet de nombreux cas d'utilisation pour les développeurs qui auraient historiquement nécessité des modèles spécifiques à un domaine.

L'exemple suivant montre comment utiliser les entrées d'image Gemma avec l'API Gemini :

Python

from google import genai

client = genai.Client()

my_file = client.files.upload(file="path/to/sample.jpg")

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    contents=[my_file, "Caption this image."],
)

print(response.text)

JavaScript

import {
  GoogleGenAI,
  createUserContent,
  createPartFromUri,
} from "@google/genai";

const ai = new GoogleGenAI();

const myfile = await ai.files.upload({
  file: "path/to/sample.jpg",
  config: { mimeType: "image/jpeg" },
});

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: createUserContent([
    createPartFromUri(myfile.uri, myfile.mimeType),
    "Caption this image.",
  ]),
});
console.log(response.text);
 ```

REST

IMAGE_PATH="cats-and-dogs.jpg"
MIME_TYPE=$(file -b --mime-type "${IMAGE_PATH}")
NUM_BYTES=$(wc -c < "${IMAGE_PATH}")
DISPLAY_NAME=IMAGE

tmp_header_file=upload-header.tmp

# Initial resumable request defining metadata.
curl "https://generativelanguage.googleapis.com/upload/v1beta/files" \
  -D upload-header.tmp \
  -H "X-Goog-Upload-Protocol: resumable" \
  -H "X-Goog-Upload-Command: start" \
  -H "X-Goog-Upload-Header-Content-Length: ${NUM_BYTES}" \
  -H "X-Goog-Upload-Header-Content-Type: ${MIME_TYPE}" \
  -H "Content-Type: application/json" \
  -d "{'file': {'display_name': '${DISPLAY_NAME}'}}" 2> /dev/null

upload_url=$(grep -i "x-goog-upload-url: " "${tmp_header_file}" | cut -d" " -f2 | tr -d "\r")
rm "${tmp_header_file}"

# Upload the actual bytes.
curl "${upload_url}" \
  -H "Content-Length: ${NUM_BYTES}" \
  -H "X-Goog-Upload-Offset: 0" \
  -H "X-Goog-Upload-Command: upload, finalize" \
  --data-binary "@${IMAGE_PATH}" 2> /dev/null > file_info.json

file_uri=$(jq -r ".file.uri" file_info.json)
echo file_uri=$file_uri

# Now generate content using that file
curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
    -H 'Content-Type: application/json' \
    -X POST \
    -d '{
      "contents": [{
        "parts":[
          {"file_data":{"mime_type": "'"${MIME_TYPE}"'", "file_uri": "'"${file_uri}"'"}},
          {"text": "Caption this image."}]
        }]
      }' 2> /dev/null > response.json

cat response.json
echo

jq -r ".candidates[].content.parts[].text" response.json

En savoir plus sur la compréhension des images :

Instructions système

Vous pouvez transmettre une instruction système pour définir le comportement du modèle :

Python

from google import genai
from google.genai import types

client = genai.Client()

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    config=types.GenerateContentConfig(
        system_instruction="You are a wise Kyoto tea master. Speak calmly and poetically, using nature metaphors. Keep answers under 3 sentences."
    ),
    contents="What is the purpose of the tea ceremony?"
)
print(response.text)

JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI();

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: "What is the purpose of the tea ceremony?",
  config: {
    systemInstruction: "You are a wise Kyoto tea master. Speak calmly and poetically, using nature metaphors. Keep answers under 3 sentences."
  }
});
console.log(response.text);

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "What is the purpose of the tea ceremony?"}]
  }],
  "systemInstruction": {
    "parts": [{"text": "You are a wise Kyoto tea master. Speak calmly and poetically, using nature metaphors. Keep answers under 3 sentences."}]
  }
}'

Conversations multitours

Le SDK fournit une interface de chat qui suit automatiquement l'historique des conversations :

Python

from google import genai

client = genai.Client()
chat = client.chats.create(model="gemma-4-26b-a4b-it")

response = chat.send_message("What are the three most famous castles in Japan?")
print(response.text)

response = chat.send_message("Which one should I visit in spring for cherry blossoms?")
print(response.text)

JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI();
const chat = ai.chats.create({ model: "gemma-4-26b-a4b-it" });

let response = await chat.sendMessage({ message: "What are the three most famous castles in Japan?" });
console.log(response.text);

response = await chat.sendMessage({ message: "Which one should I visit in spring for cherry blossoms?" });
console.log(response.text);

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [
    {
      "role": "user",
      "parts": [{ "text": "What are the three most famous castles in Japan?" }]
    },
    {
      "role": "model",
      "parts": [{ "text": "Himeji Castle, Matsumoto Castle, and Kumamoto Castle are often considered the top three." }]
    },
    {
      "role": "user",
      "parts": [{ "text": "Which one should I visit in spring for cherry blossoms?" }]
    }
  ]
}'

Appel de fonction

Définissez les outils comme des déclarations de fonction. Le modèle décide quand les appeler :

Python

from google import genai
from google.genai import types

# Define the function declaration
get_weather = {
    "name": "get_weather",
    "description": "Get current weather for a given location.",
    "parameters": {
        "type": "object",
        "properties": {
            "location": {
                "type": "string",
                "description": "City and state, e.g. 'San Francisco, CA'",
            },
        },
        "required": ["location"],
    },
}

client = genai.Client()
tools = types.Tool(function_declarations=[get_weather])
config = types.GenerateContentConfig(tools=[tools])

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    contents="Should I bring an umbrella to Kyoto today?",
    config=config,
)

# The model returns a function call instead of text
if response.function_calls:
    for fc in response.function_calls:
        print(f"Function to call: {fc.name}")
        print(f"ID: {fc.id}")
        print(f"Arguments: {fc.args}")
else:
    print("No function call found in the response.")
    print(response.text)

JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI();

const get_weather = {
    name: "get_weather",
    description: "Get current weather for a given location.",
    parameters: {
        type: "object",
        properties: {
            location: {
                type: "string",
                description: "City and state, e.g. 'San Francisco, CA'",
            },
        },
        required: ["location"],
    },
};

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: "Should I bring an umbrella to Kyoto today?",
  config: {
    tools: [{ functionDeclarations: [get_weather] }]
  }
});

if (response.functionCalls) {
    for (const fc of response.functionCalls) {
        console.log(`Function to call: ${fc.name}`);
        console.log(`Arguments: ${JSON.stringify(fc.args)}`);
    }
} else {
    console.log("No function call found in the response.");
    console.log(response.text);
}

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "Should I bring an umbrella to Kyoto today?"}]
  }],
  "tools": [{
    "functionDeclarations": [{
      "name": "get_weather",
      "description": "Get current weather for a given location.",
      "parameters": {
        "type": "OBJECT",
        "properties": {
          "location": {
            "type": "STRING",
            "description": "City and state, e.g. 'San Francisco, CA'"
          }
        },
        "required": ["location"]
      }
    }]
  }]
}'

Ancrez les réponses de Gemma 4 dans des données Web en temps réel avec la recherche Google :

Python

from google import genai
from google.genai import types

client = genai.Client()

response = client.models.generate_content(
    model="gemma-4-26b-a4b-it",
    contents="What are the dates for cherry blossom season in Tokyo this year?",
    config=types.GenerateContentConfig(
        tools=[{"google_search":{}}]
    ),
)

print(response.text)

# Access grounding metadata for citations
for chunk in response.candidates[0].grounding_metadata.grounding_chunks:
    print(f"Source: {chunk.web.title} — {chunk.web.uri}")

JavaScript

import { GoogleGenAI } from "@google/genai";

const ai = new GoogleGenAI();

const response = await ai.models.generateContent({
  model: "gemma-4-26b-a4b-it",
  contents: "What are the dates for cherry blossom season in Tokyo this year?",
  config: {
    tools: [{ googleSearch: {} }]
  }
});

console.log(response.text);

if (response.candidates?.[0]?.groundingMetadata?.groundingChunks) {
    for (const chunk of response.candidates[0].groundingMetadata.groundingChunks) {
        if (chunk.web) {
            console.log(`Source: ${chunk.web.title} — ${chunk.web.uri}`);
        }
    }
}

REST

curl "https://generativelanguage.googleapis.com/v1beta/models/gemma-4-26b-a4b-it:generateContent" \
-H 'Content-Type: application/json' \
-X POST \
-d '{
  "contents": [{
    "parts":[{"text": "What are the dates for cherry blossom season in Tokyo this year?"}]
  }],
  "tools": [{"googleSearch": {}}]
}'