跳轉到

結構化輸出

你可以設定 Gemini 模型,讓它產生的回應遵循你提供的 JSON Schema。這能確保結果可預測且具備型別安全,也讓從非結構化文字中 擷取結構化資料變得更簡單。

結構化輸出特別適合用於:

  • 資料擷取:從文字中提取姓名、日期等特定資訊。
  • 結構化分類:將文字歸入預先定義的類別。
  • 代理工作流程:為工具或 API 產生結構化輸入。

除了 REST API 支援 JSON Schema 之外,Google GenAI SDK 也 讓你使用下列函式庫定義結構描述: Pydantic (Python) and Zod (JavaScript).

如果你需要是/否機率、單選選項,或附帶 校準後信賴度且不耗用輸出權杖的評分量表分數,請使用 Gemini Decision 1 模型搭配 Decisions API.

結構化輸出範例

食譜擷取器

這個範例示範如何運用基本的 JSON Schema 型別,例如 object、array、 string 與 integer,從文字中擷取結構化資料。

from google import genai
from pydantic import BaseModel, Field
from typing import List, Optional

class Ingredient(BaseModel):
    name: str = Field(description="Name of the ingredient.")
    quantity: str = Field(description="Quantity of the ingredient, including units.")

class Recipe(BaseModel):
    recipe_name: str = Field(description="The name of the recipe.")
    prep_time_minutes: Optional[int] = Field(description="Optional time in minutes to prepare the recipe.")
    ingredients: List[Ingredient]
    instructions: List[str]

client = genai.Client()

prompt = """
Please extract the recipe from the following text.
The user wants to make delicious chocolate chip cookies.
They need 2 and 1/4 cups of all-purpose flour, 1 teaspoon of baking soda,
1 teaspoon of salt, 1 cup of unsalted butter (softened), 3/4 cup of granulated sugar,
3/4 cup of packed brown sugar, 1 teaspoon of vanilla extract, and 2 large eggs.
For the best part, they'll need 2 cups of semisweet chocolate chips.
First, preheat the oven to 375°F (190°C). Then, in a small bowl, whisk together the flour,
baking soda, and salt. In a large bowl, cream together the butter, granulated sugar, and brown sugar
until light and fluffy. Beat in the vanilla and eggs, one at a time. Gradually beat in the dry
ingredients until just combined. Finally, stir in the chocolate chips. Drop by rounded tablespoons
onto ungreased baking sheets and bake for 9 to 11 minutes.
"""

interaction = client.interactions.create(
    model="gemini-4-hydrogen",
    input=prompt,
    response_format={
        "type": "text",
        "mime_type": "application/json",
        "schema": Recipe.model_json_schema()
    },
)

recipe = Recipe.model_validate_json(interaction.output_text)
print(recipe)
import { GoogleGenAI } from "@google/genai";
import * as z from "zod";

const recipeJsonSchema = {
  type: "object",
  properties: {
    recipe_name: {
      type: "string",
      description: "The name of the recipe."
    },
    prep_time_minutes: {
        type: "integer",
        description: "Optional time in minutes to prepare the recipe."
    },
    ingredients: {
      type: "array",
      items: {
        type: "object",
        properties: {
          name: { type: "string", description: "Name of the ingredient."},
          quantity: { type: "string", description: "Quantity of the ingredient, including units."}
        },
        required: ["name", "quantity"]
      }
    },
    instructions: {
      type: "array",
      items: { type: "string" }
    }
  },
  required: ["recipe_name", "ingredients", "instructions"]
};

const recipeSchema = z.fromJSONSchema(recipeJsonSchema);

const client = new GoogleGenAI({});

const prompt = `
Please extract the recipe from the following text.
The user wants to make delicious chocolate chip cookies.
They need 2 and 1/4 cups of all-purpose flour, 1 teaspoon of baking soda,
1 teaspoon of salt, 1 cup of unsalted butter (softened), 3/4 cup of granulated sugar,
3/4 cup of packed brown sugar, 1 teaspoon of vanilla extract, and 2 large eggs.
For the best part, they'll need 2 cups of semisweet chocolate chips.
First, preheat the oven to 375°F (190°C). Then, in a small bowl, whisk together the flour,
baking soda, and salt. In a large bowl, cream together the butter, granulated sugar, and brown sugar
until light and fluffy. Beat in the vanilla and eggs, one at a time. Gradually beat in the dry
ingredients until just combined. Finally, stir in the chocolate chips. Drop by rounded tablespoons
onto ungreased baking sheets and bake for 9 to 11 minutes.
`;

const interaction = await client.interactions.create({
  model: "gemini-4-hydrogen",
  input: prompt,
  response_format: {
    type: 'text',
    mime_type: 'application/json',
    schema: recipeJsonSchema
  },
});

const recipe = recipeSchema.parse(JSON.parse(interaction.output_text));
console.log(recipe);
import com.google.genai.Client;
import com.google.genai.gaos.models.interactions.CreateModelInteraction;
import com.google.genai.gaos.models.interactions.CreateModelInteractionResponseFormat;
import com.google.genai.gaos.models.interactions.Interaction;
import com.google.genai.gaos.models.interactions.InteractionsInput;
import com.google.genai.gaos.models.interactions.Model;
import com.google.genai.gaos.models.interactions.ResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormatMimeType;
import com.google.genai.gaos.models.operations.CreateInteractionRequestBody;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;

Client client = new Client();

Map<String, Object> ingredientProps = new HashMap<>();
Map<String, Object> nameProp = new HashMap<>();
nameProp.put("type", "string");
nameProp.put("description", "Name of the ingredient.");
ingredientProps.put("name", nameProp);

Map<String, Object> quantityProp = new HashMap<>();
quantityProp.put("type", "string");
quantityProp.put("description", "Quantity of the ingredient, including units.");
ingredientProps.put("quantity", quantityProp);

Map<String, Object> ingredientItemSchema = new HashMap<>();
ingredientItemSchema.put("type", "object");
ingredientItemSchema.put("properties", ingredientProps);
ingredientItemSchema.put("required", Arrays.asList("name", "quantity"));

Map<String, Object> properties = new HashMap<>();

Map<String, Object> recipeNameProp = new HashMap<>();
recipeNameProp.put("type", "string");
recipeNameProp.put("description", "The name of the recipe.");
properties.put("recipe_name", recipeNameProp);

Map<String, Object> prepTimeProp = new HashMap<>();
prepTimeProp.put("type", "integer");
prepTimeProp.put("description", "Optional time in minutes to prepare the recipe.");
properties.put("prep_time_minutes", prepTimeProp);

Map<String, Object> ingredientsProp = new HashMap<>();
ingredientsProp.put("type", "array");
ingredientsProp.put("items", ingredientItemSchema);
properties.put("ingredients", ingredientsProp);

Map<String, Object> instructionsProp = new HashMap<>();
instructionsProp.put("type", "array");
Map<String, Object> stringItem = new HashMap<>();
stringItem.put("type", "string");
instructionsProp.put("items", stringItem);
properties.put("instructions", instructionsProp);

Map<String, Object> recipeJsonSchema = new HashMap<>();
recipeJsonSchema.put("type", "object");
recipeJsonSchema.put("properties", properties);
recipeJsonSchema.put("required", Arrays.asList("recipe_name", "ingredients", "instructions"));

String prompt =
    "Please extract the recipe from the following text.\n"
        + "The user wants to make delicious chocolate chip cookies.\n"
        + "They need 2 and 1/4 cups of all-purpose flour, 1 teaspoon of baking soda,\n"
        + "1 teaspoon of salt, 1 cup of unsalted butter (softened), 3/4 cup of granulated sugar,\n"
        + "3/4 cup of packed brown sugar, 1 teaspoon of vanilla extract, and 2 large eggs.\n"
        + "For the best part, they'll need 2 cups of semisweet chocolate chips.\n"
        + "First, preheat the oven to 375°F (190°C). Then, in a small bowl, whisk together the flour,\n"
        + "baking soda, and salt. In a large bowl, cream together the butter, granulated sugar, and brown sugar\n"
        + "until light and fluffy. Beat in the vanilla and eggs, one at a time. Gradually beat in the dry\n"
        + "ingredients until just combined. Finally, stir in the chocolate chips. Drop by rounded tablespoons\n"
        + "onto ungreased baking sheets and bake for 9 to 11 minutes.";

CreateModelInteractionResponseFormat format =
    CreateModelInteractionResponseFormat.of(
        ResponseFormat.of(
            TextResponseFormat.builder()
                .mimeType(TextResponseFormatMimeType.APPLICATION_JSON)
                .schema(recipeJsonSchema)
                .build()));

CreateModelInteraction params =
    CreateModelInteraction.builder()
        .model(Model.of("gemini-4-hydrogen"))
        .input(InteractionsInput.of(prompt))
        .responseFormat(format)
        .build();

Interaction interaction =
    client.interactions.create(CreateInteractionRequestBody.of(params)).interaction().get();

System.out.println(interaction.outputText().orElse(""));
package main

import (
    "context"
    "fmt"
    "log"

    "google.golang.org/genai"
    "google.golang.org/genai/interactions/models/interactions"
    "google.golang.org/genai/interactions/models/operations"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil {
        log.Fatal(err)
    }

    recipeJsonSchema := map[string]any{
        "type": "object",
        "properties": map[string]any{
            "recipe_name": map[string]any{
                "type":        "string",
                "description": "The name of the recipe.",
            },
            "prep_time_minutes": map[string]any{
                "type":        "integer",
                "description": "Optional time in minutes to prepare the recipe.",
            },
            "ingredients": map[string]any{
                "type": "array",
                "items": map[string]any{
                    "type": "object",
                    "properties": map[string]any{
                        "name": map[string]any{
                            "type":        "string",
                            "description": "Name of the ingredient.",
                        },
                        "quantity": map[string]any{
                            "type":        "string",
                            "description": "Quantity of the ingredient, including units.",
                        },
                    },
                    "required": []string{"name", "quantity"},
                },
            },
            "instructions": map[string]any{
                "type": "array",
                "items": map[string]any{
                    "type": "string",
                },
            },
        },
        "required": []string{"recipe_name", "ingredients", "instructions"},
    }

    prompt := `Please extract the recipe from the following text.
The user wants to make delicious chocolate chip cookies.
They need 2 and 1/4 cups of all-purpose flour, 1 teaspoon of baking soda,
1 teaspoon of salt, 1 cup of unsalted butter (softened), 3/4 cup of granulated sugar,
3/4 cup of packed brown sugar, 1 teaspoon of vanilla extract, and 2 large eggs.
For the best part, they'll need 2 cups of semisweet chocolate chips.
First, preheat the oven to 375°F (190°C). Then, in a small bowl, whisk together the flour,
baking soda, and salt. In a large bowl, cream together the butter, granulated sugar, and brown sugar
until light and fluffy. Beat in the vanilla and eggs, one at a time. Gradually beat in the dry
ingredients until just combined. Finally, stir in the chocolate chips. Drop by rounded tablespoons
onto ungreased baking sheets and bake for 9 to 11 minutes.`

    format := interactions.NewCreateModelInteractionResponseFormat(
        interactions.NewResponseFormat(interactions.TextResponseFormat{
            MimeType: interactions.TextResponseFormatMimeTypeApplicationJSON.ToPointer(),
            Schema:   recipeJsonSchema,
        }),
    )

    resp, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
        Body: operations.NewCreateInteractionRequestBody(
            interactions.CreateModelInteraction{
                Model:          interactions.Model("gemini-4-hydrogen"),
                Input:          interactions.NewInteractionsInput(prompt),
                ResponseFormat: &format,
            },
        ),
    })
    if err != nil {
        log.Fatal(err)
    }

    fmt.Println(resp.Interaction.GetOutputText())
}
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
    -H "x-goog-api-key: $GEMINI_API_KEY" \
    -H 'Content-Type: application/json' \
    -H "Api-Revision: 2026-05-20" \
    -d '{
      "model": "gemini-4-hydrogen",
      "input": "Please extract the recipe from the following text.\nThe user wants to make delicious chocolate chip cookies.\nThey need 2 and 1/4 cups of all-purpose flour, 1 teaspoon of baking soda,\n1 teaspoon of salt, 1 cup of unsalted butter (softened), 3/4 cup of granulated sugar,\n3/4 cup of packed brown sugar, 1 teaspoon of vanilla extract, and 2 large eggs.\nFor the best part, they will need 2 cups of semisweet chocolate chips.\nFirst, preheat the oven to 375°F (190°C). Then, in a small bowl, whisk together the flour,\nbaking soda, and salt. In a large bowl, cream together the butter, granulated sugar, and brown sugar\nuntil light and fluffy. Beat in the vanilla and eggs, one at a time. Gradually beat in the dry\ningredients until just combined. Finally, stir in the chocolate chips. Drop by rounded tablespoons\nonto ungreased baking sheets and bake for 9 to 11 minutes.",
      "response_format": {
        "type": "text",
        "mime_type": "application/json",
        "schema": {
          "type": "object",
          "properties": {
            "recipe_name": {
              "type": "string",
              "description": "The name of the recipe."
            },
            "prep_time_minutes": {
                "type": "integer",
                "description": "Optional time in minutes to prepare the recipe."
            },
            "ingredients": {
              "type": "array",
              "items": {
                "type": "object",
                "properties": {
                  "name": { "type": "string", "description": "Name of the ingredient."},
                  "quantity": { "type": "string", "description": "Quantity of the ingredient, including units."}
                },
                "required": ["name", "quantity"]
              }
            },
            "instructions": {
              "type": "array",
              "items": { "type": "string" }
            }
          },
          "required": ["recipe_name", "ingredients", "instructions"]
        }
      }
      }
    }'

範例回應:

{
  "recipe_name": "Delicious Chocolate Chip Cookies",
  "ingredients": [
    { "name": "all-purpose flour", "quantity": "2 and 1/4 cups" },
    { "name": "baking soda", "quantity": "1 teaspoon" },
    { "name": "salt", "quantity": "1 teaspoon" },
    { "name": "unsalted butter (softened)", "quantity": "1 cup" },
    { "name": "granulated sugar", "quantity": "3/4 cup" },
    { "name": "packed brown sugar", "quantity": "3/4 cup" },
    { "name": "vanilla extract", "quantity": "1 teaspoon" },
    { "name": "large eggs", "quantity": "2" },
    { "name": "semisweet chocolate chips", "quantity": "2 cups" }
  ],
  "instructions": [
    "Preheat the oven to 375°F (190°C).",
    "In a small bowl, whisk together the flour, baking soda, and salt.",
    "In a large bowl, cream together the butter, granulated sugar, and brown sugar until light and fluffy.",
    "Beat in the vanilla and eggs, one at a time.",
    "Gradually beat in the dry ingredients until just combined.",
    "Stir in the chocolate chips.",
    "Drop by rounded tablespoons onto ungreased baking sheets and bake for 9 to 11 minutes."
  ]
}

內容審核

這個範例展示如何以 anyOf 建立條件式結構描述、以 enum 進行 分類,讓輸出結構能依內容而變化。

from google import genai
from pydantic import BaseModel, Field
from typing import Union, Literal

class SpamDetails(BaseModel):
    reason: str = Field(description="The reason why the content is considered spam.")
    spam_type: Literal["phishing", "scam", "unsolicited promotion", "other"] = Field(description="The type of spam.")

class NotSpamDetails(BaseModel):
    summary: str = Field(description="A brief summary of the content.")
    is_safe: bool = Field(description="Whether the content is safe for all audiences.")

class ModerationResult(BaseModel):
    decision: Union[SpamDetails, NotSpamDetails]

client = genai.Client()

prompt = """
Please moderate the following content and provide a decision.
Content: 'Congratulations! You''ve won a free cruise to the Bahamas. Click here to claim your prize: www.definitely-not-a-scam.com'
"""

interaction = client.interactions.create(
    model="gemini-4-hydrogen",
    input=prompt,
    response_format={
        "type": "text",
        "mime_type": "application/json",
        "schema": ModerationResult.model_json_schema()
    },
)

result = ModerationResult.model_validate_json(interaction.output_text)
print(result)
import { GoogleGenAI } from "@google/genai";
import * as z from "zod";

const moderationResultJsonSchema = {
  type: "object",
  properties: {
    decision: {
      anyOf: [
        {
          type: "object",
          title: "SpamDetails",
          description: "Details for content classified as spam.",
          properties: {
            reason: { type: "string", description: "The reason why the content is considered spam." },
            spam_type: { type: "string", enum: ["phishing", "scam", "unsolicited promotion", "other"], description: "The type of spam." }
          },
          required: ["reason", "spam_type"]
        },
        {
          type: "object",
          title: "NotSpamDetails",
          description: "Details for content classified as not spam.",
          properties: {
            summary: { type: "string", description: "A brief summary of the content." },
            is_safe: { type: "boolean", description: "Whether the content is safe for all audiences." }
          },
          required: ["summary", "is_safe"]
        }
      ]
    }
  },
  required: ["decision"]
};

const moderationResultSchema = z.fromJSONSchema(moderationResultJsonSchema);

const client = new GoogleGenAI({});

const prompt = `
Please moderate the following content and provide a decision.
Content: 'Congratulations! You''ve won a free cruise to the Bahamas. Click here to claim your prize: www.definitely-not-a-scam.com'
`;

const interaction = await client.interactions.create({
  model: "gemini-4-hydrogen",
  input: prompt,
  response_format: {
    type: 'text',
    mime_type: 'application/json',
    schema: moderationResultJsonSchema
  },
});

const result = moderationResultSchema.parse(JSON.parse(interaction.output_text));
console.log(result);
import com.google.genai.Client;
import com.google.genai.gaos.models.interactions.CreateModelInteraction;
import com.google.genai.gaos.models.interactions.CreateModelInteractionResponseFormat;
import com.google.genai.gaos.models.interactions.Interaction;
import com.google.genai.gaos.models.interactions.InteractionsInput;
import com.google.genai.gaos.models.interactions.Model;
import com.google.genai.gaos.models.interactions.ResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormatMimeType;
import com.google.genai.gaos.models.operations.CreateInteractionRequestBody;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;

Client client = new Client();

Map<String, Object> spamProps = new HashMap<>();
Map<String, Object> reasonProp = new HashMap<>();
reasonProp.put("type", "string");
reasonProp.put("description", "The reason why the content is considered spam.");
spamProps.put("reason", reasonProp);

Map<String, Object> spamTypeProp = new HashMap<>();
spamTypeProp.put("type", "string");
spamTypeProp.put("enum", Arrays.asList("phishing", "scam", "unsolicited promotion", "other"));
spamTypeProp.put("description", "The type of spam.");
spamProps.put("spam_type", spamTypeProp);

Map<String, Object> spamDetailsSchema = new HashMap<>();
spamDetailsSchema.put("type", "object");
spamDetailsSchema.put("title", "SpamDetails");
spamDetailsSchema.put("properties", spamProps);
spamDetailsSchema.put("required", Arrays.asList("reason", "spam_type"));

Map<String, Object> notSpamProps = new HashMap<>();
Map<String, Object> summaryProp = new HashMap<>();
summaryProp.put("type", "string");
summaryProp.put("description", "A brief summary of the content.");
notSpamProps.put("summary", summaryProp);

Map<String, Object> isSafeProp = new HashMap<>();
isSafeProp.put("type", "boolean");
isSafeProp.put("description", "Whether the content is safe for all audiences.");
notSpamProps.put("is_safe", isSafeProp);

Map<String, Object> notSpamDetailsSchema = new HashMap<>();
notSpamDetailsSchema.put("type", "object");
notSpamDetailsSchema.put("title", "NotSpamDetails");
notSpamDetailsSchema.put("properties", notSpamProps);
notSpamDetailsSchema.put("required", Arrays.asList("summary", "is_safe"));

Map<String, Object> decisionProp = new HashMap<>();
decisionProp.put("anyOf", Arrays.asList(spamDetailsSchema, notSpamDetailsSchema));

Map<String, Object> properties = new HashMap<>();
properties.put("decision", decisionProp);

Map<String, Object> moderationResultJsonSchema = new HashMap<>();
moderationResultJsonSchema.put("type", "object");
moderationResultJsonSchema.put("properties", properties);
moderationResultJsonSchema.put("required", Arrays.asList("decision"));

String prompt =
    "Please moderate the following content and provide a decision.\n"
        + "Content: 'Congratulations! You''ve won a free cruise to the Bahamas. Click here to claim your prize: www.definitely-not-a-scam.com'";

CreateModelInteractionResponseFormat format =
    CreateModelInteractionResponseFormat.of(
        ResponseFormat.of(
            TextResponseFormat.builder()
                .mimeType(TextResponseFormatMimeType.APPLICATION_JSON)
                .schema(moderationResultJsonSchema)
                .build()));

CreateModelInteraction params =
    CreateModelInteraction.builder()
        .model(Model.of("gemini-4-hydrogen"))
        .input(InteractionsInput.of(prompt))
        .responseFormat(format)
        .build();

Interaction interaction =
    client.interactions.create(CreateInteractionRequestBody.of(params)).interaction().get();

System.out.println(interaction.outputText().orElse(""));
package main

import (
    "context"
    "fmt"
    "log"

    "google.golang.org/genai"
    "google.golang.org/genai/interactions/models/interactions"
    "google.golang.org/genai/interactions/models/operations"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil {
        log.Fatal(err)
    }

    spamDetailsSchema := map[string]any{
        "type":  "object",
        "title": "SpamDetails",
        "properties": map[string]any{
            "reason": map[string]any{
                "type":        "string",
                "description": "The reason why the content is considered spam.",
            },
            "spam_type": map[string]any{
                "type":        "string",
                "enum":        []string{"phishing", "scam", "unsolicited promotion", "other"},
                "description": "The type of spam.",
            },
        },
        "required": []string{"reason", "spam_type"},
    }

    notSpamDetailsSchema := map[string]any{
        "type":  "object",
        "title": "NotSpamDetails",
        "properties": map[string]any{
            "summary": map[string]any{
                "type":        "string",
                "description": "A brief summary of the content.",
            },
            "is_safe": map[string]any{
                "type":        "boolean",
                "description": "Whether the content is safe for all audiences.",
            },
        },
        "required": []string{"summary", "is_safe"},
    }

    moderationResultJsonSchema := map[string]any{
        "type": "object",
        "properties": map[string]any{
            "decision": map[string]any{
                "anyOf": []any{spamDetailsSchema, notSpamDetailsSchema},
            },
        },
        "required": []string{"decision"},
    }

    prompt := "Please moderate the following content and provide a decision.\n" +
        "Content: 'Congratulations! You've won a free cruise to the Bahamas. Click here to claim your prize: www.definitely-not-a-scam.com'"

    format := interactions.NewCreateModelInteractionResponseFormat(
        interactions.NewResponseFormat(interactions.TextResponseFormat{
            MimeType: interactions.TextResponseFormatMimeTypeApplicationJSON.ToPointer(),
            Schema:   moderationResultJsonSchema,
        }),
    )

    resp, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
        Body: operations.NewCreateInteractionRequestBody(
            interactions.CreateModelInteraction{
                Model:          interactions.Model("gemini-4-hydrogen"),
                Input:          interactions.NewInteractionsInput(prompt),
                ResponseFormat: &format,
            },
        ),
    })
    if err != nil {
        log.Fatal(err)
    }

    fmt.Println(resp.Interaction.GetOutputText())
}
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
    -H "x-goog-api-key: $GEMINI_API_KEY" \
    -H 'Content-Type: application/json' \
    -H "Api-Revision: 2026-05-20" \
    -d '{
      "model": "gemini-4-hydrogen",
      "input": "Please moderate the following content and provide a decision.\nContent: '\''Congratulations! You have won a free cruise to the Bahamas. Click here to claim your prize: www.definitely-not-a-scam.com'\''",
      "response_format": {
        "type": "text",
        "mime_type": "application/json",
        "schema": {
          "type": "object",
          "properties": {
            "decision": {
              "anyOf": [
                {
                  "type": "object",
                  "title": "SpamDetails",
                  "description": "Details for content classified as spam.",
                  "properties": {
                    "reason": { "type": "string", "description": "The reason why the content is considered spam." },
                    "spam_type": { "type": "string", "enum": ["phishing", "scam", "unsolicited promotion", "other"], "description": "The type of spam." }
                  },
                  "required": ["reason", "spam_type"]
                },
                {
                  "type": "object",
                  "title": "NotSpamDetails",
                  "description": "Details for content classified as not spam.",
                  "properties": {
                    "summary": { "type": "string", "description": "A brief summary of the content." },
                    "is_safe": { "type": "boolean", "description": "Whether the content is safe for all audiences." }
                  },
                  "required": ["summary", "is_safe"]
                }
              ]
            }
          },
          "required": ["decision"]
        }
      }
      }
    }'

範例回應:

{
  "decision": {
    "reason": "The content is an unsolicited prize notification attempting to trick the user into clicking a suspicious link.",
    "spam_type": "scam"
  }
}

遞迴結構

這個範例說明如何定義遞迴結構描述,例如 組織架構圖。

from google import genai
from pydantic import BaseModel, Field
from typing import List

class Employee(BaseModel):
    """Represents an employee in an organization."""
    name: str
    employee_id: int
    reports: List["Employee"] = Field(
        default_factory=list,
        description="A list of employees reporting to this employee."
    )

client = genai.Client()

prompt = """
Generate an organization chart for a small team.
The manager is Alice, who manages Bob and Charlie. Bob manages David.
"""

interaction = client.interactions.create(
    model="gemini-4-hydrogen",
    input=prompt,
    response_format={
        "type": "text",
        "mime_type": "application/json",
        "schema": Employee.model_json_schema()
    },
)

employee = Employee.model_validate_json(interaction.output_text)
print(employee)
import { GoogleGenAI } from "@google/genai";
import * as z from "zod";

const employeeJsonSchema = {
  type: "object",
  properties: {
    name: { type: "string" },
    employee_id: { type: "integer" },
    reports: {
      type: "array",
      description: "A list of employees reporting to this employee.",
      items: {
        "$ref": "#"
      }
    }
  },
  required: ["name", "employee_id", "reports"]
};

const employeeSchema = z.fromJSONSchema(employeeJsonSchema);

const client = new GoogleGenAI({});

const prompt = `
Generate an organization chart for a small team.
The manager is Alice, who manages Bob and Charlie. Bob manages David.
`;

const interaction = await client.interactions.create({
  model: "gemini-4-hydrogen",
  input: prompt,
  response_format: {
    type: 'text',
    mime_type: 'application/json',
    schema: employeeJsonSchema
  },
});

const employee = employeeSchema.parse(JSON.parse(interaction.output_text));
console.log(employee);
import com.google.genai.Client;
import com.google.genai.gaos.models.interactions.CreateModelInteraction;
import com.google.genai.gaos.models.interactions.CreateModelInteractionResponseFormat;
import com.google.genai.gaos.models.interactions.Interaction;
import com.google.genai.gaos.models.interactions.InteractionsInput;
import com.google.genai.gaos.models.interactions.Model;
import com.google.genai.gaos.models.interactions.ResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormatMimeType;
import com.google.genai.gaos.models.operations.CreateInteractionRequestBody;
import java.util.Arrays;
import java.util.Collections;
import java.util.HashMap;
import java.util.Map;

Client client = new Client();

Map<String, Object> properties = new HashMap<>();

Map<String, Object> nameProp = new HashMap<>();
nameProp.put("type", "string");
properties.put("name", nameProp);

Map<String, Object> idProp = new HashMap<>();
idProp.put("type", "integer");
properties.put("employee_id", idProp);

Map<String, Object> reportsProp = new HashMap<>();
reportsProp.put("type", "array");
reportsProp.put("description", "A list of employees reporting to this employee.");
reportsProp.put("items", Collections.singletonMap("$ref", "#"));
properties.put("reports", reportsProp);

Map<String, Object> employeeJsonSchema = new HashMap<>();
employeeJsonSchema.put("type", "object");
employeeJsonSchema.put("properties", properties);
employeeJsonSchema.put("required", Arrays.asList("name", "employee_id", "reports"));

String prompt =
    "Generate an organization chart for a small team.\n"
        + "The manager is Alice, who manages Bob and Charlie. Bob manages David.";

CreateModelInteractionResponseFormat format =
    CreateModelInteractionResponseFormat.of(
        ResponseFormat.of(
            TextResponseFormat.builder()
                .mimeType(TextResponseFormatMimeType.APPLICATION_JSON)
                .schema(employeeJsonSchema)
                .build()));

CreateModelInteraction params =
    CreateModelInteraction.builder()
        .model(Model.of("gemini-4-hydrogen"))
        .input(InteractionsInput.of(prompt))
        .responseFormat(format)
        .build();

Interaction interaction =
    client.interactions.create(CreateInteractionRequestBody.of(params)).interaction().get();

System.out.println(interaction.outputText().orElse(""));
package main

import (
    "context"
    "fmt"
    "log"

    "google.golang.org/genai"
    "google.golang.org/genai/interactions/models/interactions"
    "google.golang.org/genai/interactions/models/operations"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil {
        log.Fatal(err)
    }

    employeeJsonSchema := map[string]any{
        "type": "object",
        "properties": map[string]any{
            "name": map[string]any{
                "type": "string",
            },
            "employee_id": map[string]any{
                "type": "integer",
            },
            "reports": map[string]any{
                "type":        "array",
                "description": "A list of employees reporting to this employee.",
                "items": map[string]any{
                    "$ref": "#",
                },
            },
        },
        "required": []string{"name", "employee_id", "reports"},
    }

    prompt := "Generate an organization chart for a small team.\n" +
        "The manager is Alice, who manages Bob and Charlie. Bob manages David."

    format := interactions.NewCreateModelInteractionResponseFormat(
        interactions.NewResponseFormat(interactions.TextResponseFormat{
            MimeType: interactions.TextResponseFormatMimeTypeApplicationJSON.ToPointer(),
            Schema:   employeeJsonSchema,
        }),
    )

    resp, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
        Body: operations.NewCreateInteractionRequestBody(
            interactions.CreateModelInteraction{
                Model:          interactions.Model("gemini-4-hydrogen"),
                Input:          interactions.NewInteractionsInput(prompt),
                ResponseFormat: &format,
            },
        ),
    })
    if err != nil {
        log.Fatal(err)
    }

    fmt.Println(resp.Interaction.GetOutputText())
}
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
    -H "x-goog-api-key: $GEMINI_API_KEY" \
    -H 'Content-Type: application/json' \
    -H "Api-Revision: 2026-05-20" \
    -d '{
      "model": "gemini-4-hydrogen",
      "input": "Generate an organization chart for a small team.\nThe manager is Alice, who manages Bob and Charlie. Bob manages David.",
      "response_format": {
        "type": "text",
        "mime_type": "application/json",
        "schema": {
          "type": "object",
          "properties": {
            "name": { "type": "string" },
            "employee_id": { "type": "integer" },
            "reports": {
              "type": "array",
              "description": "A list of employees reporting to this employee.",
              "items": {
                "$ref": "#"
              }
            }
          },
          "required": ["name", "employee_id", "reports"]
        }
      }
      }
    }'

範例回應:

{
  "name": "Alice",
  "employee_id": 101,
  "reports": [
    {
      "name": "Bob",
      "employee_id": 102,
      "reports": [
        {
          "name": "David",
          "employee_id": 104,
          "reports": []
        }
      ]
    },
    {
      "name": "Charlie",
      "employee_id": 103,
      "reports": []
    }
  ]
}

串流結果

你可以串流結構化輸出,在回應產生的同時就開始處理。 串流傳回的區塊是有效的部分 JSON 字串, 串接起來即可組成最終的 JSON 物件。

from google import genai
from pydantic import BaseModel
from typing import Literal

class Feedback(BaseModel):
    sentiment: Literal["positive", "neutral", "negative"]
    summary: str

client = genai.Client()
prompt = "The new UI is incredibly intuitive. Add a very long summary to test streaming!"

stream = client.interactions.create(
    model="gemini-4-hydrogen",
    input=prompt,
    response_format={
        "type": "text",
        "mime_type": "application/json",
        "schema": Feedback.model_json_schema()
    },
    stream=True
)
for event in stream:
    if event.event_type == "step.delta" and event.delta.text:
        print(event.delta.text, end="")
import { GoogleGenAI } from "@google/genai";
import * as z from "zod";

const feedbackJsonSchema = {
  type: "object",
  properties: {
    sentiment: { type: "string", enum: ["positive", "neutral", "negative"] },
    summary: { type: "string" }
  },
  required: ["sentiment", "summary"]
};

const feedbackSchema = z.fromJSONSchema(feedbackJsonSchema);

const client = new GoogleGenAI({});

const stream = await client.interactions.create({
  model: "gemini-4-hydrogen",
  input: "The new UI is incredibly intuitive. Add a very long summary!",
  response_format: {
    type: 'text',
    mime_type: 'application/json',
    schema: feedbackJsonSchema
  },
  stream: true,
});

for await (const event of stream) {
  if (event.event_type === "step.delta" && event.delta?.type === "text") {
    process.stdout.write(event.delta.text);
  }
}
import com.google.genai.Client;
import com.google.genai.gaos.models.interactions.CreateModelInteraction;
import com.google.genai.gaos.models.interactions.CreateModelInteractionResponseFormat;
import com.google.genai.gaos.models.interactions.InteractionSSEEvent;
import com.google.genai.gaos.models.interactions.InteractionSSEStreamEvent;
import com.google.genai.gaos.models.interactions.InteractionsInput;
import com.google.genai.gaos.models.interactions.Model;
import com.google.genai.gaos.models.interactions.ResponseFormat;
import com.google.genai.gaos.models.interactions.StepDelta;
import com.google.genai.gaos.models.interactions.StepDeltaData;
import com.google.genai.gaos.models.interactions.TextDelta;
import com.google.genai.gaos.models.interactions.TextResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormatMimeType;
import com.google.genai.gaos.models.operations.CreateInteractionRequestBody;
import com.google.genai.gaos.models.operations.CreateInteractionResponse;
import com.google.genai.gaos.utils.EventStream;
import java.util.Arrays;
import java.util.HashMap;
import java.util.Map;

Client client = new Client();

Map<String, Object> properties = new HashMap<>();

Map<String, Object> sentimentProp = new HashMap<>();
sentimentProp.put("type", "string");
sentimentProp.put("enum", Arrays.asList("positive", "neutral", "negative"));
properties.put("sentiment", sentimentProp);

Map<String, Object> summaryProp = new HashMap<>();
summaryProp.put("type", "string");
properties.put("summary", summaryProp);

Map<String, Object> feedbackJsonSchema = new HashMap<>();
feedbackJsonSchema.put("type", "object");
feedbackJsonSchema.put("properties", properties);
feedbackJsonSchema.put("required", Arrays.asList("sentiment", "summary"));

String prompt = "The new UI is incredibly intuitive. Add a very long summary to test streaming!";

CreateModelInteractionResponseFormat format =
    CreateModelInteractionResponseFormat.of(
        ResponseFormat.of(
            TextResponseFormat.builder()
                .mimeType(TextResponseFormatMimeType.APPLICATION_JSON)
                .schema(feedbackJsonSchema)
                .build()));

CreateModelInteraction params =
    CreateModelInteraction.builder()
        .model(Model.of("gemini-4-hydrogen"))
        .input(InteractionsInput.of(prompt))
        .responseFormat(format)
        .stream(true)
        .build();

CreateInteractionResponse response =
    client.interactions.create(CreateInteractionRequestBody.of(params));

try (EventStream<InteractionSSEStreamEvent> events = response.events()) {
  for (InteractionSSEStreamEvent streamEvent : events) {
    InteractionSSEEvent event = streamEvent.data().orElse(null);
    if (event instanceof StepDelta) {
      StepDeltaData data = ((StepDelta) event).delta().orElse(null);
      if (data instanceof TextDelta) {
        ((TextDelta) data).text().ifPresent(System.out::print);
      }
    }
  }
}
package main

import (
    "context"
    "fmt"
    "log"

    "google.golang.org/genai"
    "google.golang.org/genai/interactions/models/interactions"
    "google.golang.org/genai/interactions/models/operations"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil {
        log.Fatal(err)
    }

    feedbackJsonSchema := map[string]any{
        "type": "object",
        "properties": map[string]any{
            "sentiment": map[string]any{
                "type": "string",
                "enum": []string{"positive", "neutral", "negative"},
            },
            "summary": map[string]any{
                "type": "string",
            },
        },
        "required": []string{"sentiment", "summary"},
    }

    prompt := "The new UI is incredibly intuitive. Add a very long summary to test streaming!"

    format := interactions.NewCreateModelInteractionResponseFormat(
        interactions.NewResponseFormat(interactions.TextResponseFormat{
            MimeType: interactions.TextResponseFormatMimeTypeApplicationJSON.ToPointer(),
            Schema:   feedbackJsonSchema,
        }),
    )

    resp, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
        Body: operations.NewCreateInteractionRequestBody(
            interactions.CreateModelInteraction{
                Model:          interactions.Model("gemini-4-hydrogen"),
                Input:          interactions.NewInteractionsInput(prompt),
                ResponseFormat: &format,
                Stream:         genai.Ptr(true),
            },
        ),
    })
    if err != nil {
        log.Fatal(err)
    }
    defer resp.InteractionSSEStreamEvent.Close()

    for resp.InteractionSSEStreamEvent.Next() {
        event := resp.InteractionSSEStreamEvent.Value()
        if stepDelta := event.GetDataStepDelta(); stepDelta != nil {
            if textDelta := stepDelta.GetDeltaText(); textDelta != nil {
                fmt.Print(textDelta.GetText())
            }
        }
    }
}

搭配工具使用結構化輸出

預覽:這項功能僅適用於 Gemini 4 與 Gemini 3 系列模型, 例如 gemini-4-carbon、gemini-4-argon、gemini-4-hydrogen 與 gemini-3.8-flash。

Gemini 4 與 Gemini 3 讓你能將結構化輸出與內建工具結合使用,包括 使用 Google 搜尋接地、 URL Context、 程式碼執行、 File Search,以及 函式呼叫。

from google import genai
from pydantic import BaseModel, Field
from typing import List

class MatchResult(BaseModel):
    winner: str = Field(description="The name of the winner.")
    final_match_score: str = Field(description="The final match score.")
    scorers: List[str] = Field(description="The name of the scorer.")

client = genai.Client()

interaction = client.interactions.create(
    model="gemini-4-carbon",
    input="Search for all details for the latest Euro.",
    tools=[{"type": "google_search"}, {"type": "url_context"}],
    response_format={
        "type": "text",
        "mime_type": "application/json",
        "schema": MatchResult.model_json_schema()
    },
)

result = MatchResult.model_validate_json(interaction.output_text)
print(result)
import { GoogleGenAI } from "@google/genai";
import * as z from "zod";

const matchJsonSchema = {
  type: "object",
  properties: {
    winner: { type: "string" },
    final_match_score: { type: "string" },
    scorers: { type: "array", items: { type: "string" } }
  },
  required: ["winner", "final_match_score", "scorers"]
};

const matchSchema = z.fromJSONSchema(matchJsonSchema);

const client = new GoogleGenAI({});

const interaction = await client.interactions.create({
  model: "gemini-4-carbon",
  input: "Search for all details for the latest Euro.",
  tools: [{type: "google_search"}, {type: "url_context"}],
  response_format: {
    type: 'text',
    mime_type: 'application/json',
    schema: matchJsonSchema
  },
});

const match = matchSchema.parse(JSON.parse(interaction.output_text));
console.log(match);
import com.google.genai.Client;
import com.google.genai.gaos.models.interactions.CreateModelInteraction;
import com.google.genai.gaos.models.interactions.CreateModelInteractionResponseFormat;
import com.google.genai.gaos.models.interactions.GoogleSearch;
import com.google.genai.gaos.models.interactions.Interaction;
import com.google.genai.gaos.models.interactions.InteractionsInput;
import com.google.genai.gaos.models.interactions.Model;
import com.google.genai.gaos.models.interactions.ResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormat;
import com.google.genai.gaos.models.interactions.TextResponseFormatMimeType;
import com.google.genai.gaos.models.interactions.URLContext;
import com.google.genai.gaos.models.operations.CreateInteractionRequestBody;
import java.util.Arrays;
import java.util.Collections;
import java.util.HashMap;
import java.util.Map;

Client client = new Client();

Map<String, Object> properties = new HashMap<>();

Map<String, Object> winnerProp = new HashMap<>();
winnerProp.put("type", "string");
winnerProp.put("description", "The name of the winner.");
properties.put("winner", winnerProp);

Map<String, Object> scoreProp = new HashMap<>();
scoreProp.put("type", "string");
scoreProp.put("description", "The final match score.");
properties.put("final_match_score", scoreProp);

Map<String, Object> scorersProp = new HashMap<>();
scorersProp.put("type", "array");
scorersProp.put("description", "The name of the scorer.");
scorersProp.put("items", Collections.singletonMap("type", "string"));
properties.put("scorers", scorersProp);

Map<String, Object> matchJsonSchema = new HashMap<>();
matchJsonSchema.put("type", "object");
matchJsonSchema.put("properties", properties);
matchJsonSchema.put("required", Arrays.asList("winner", "final_match_score", "scorers"));

CreateModelInteractionResponseFormat format =
    CreateModelInteractionResponseFormat.of(
        ResponseFormat.of(
            TextResponseFormat.builder()
                .mimeType(TextResponseFormatMimeType.APPLICATION_JSON)
                .schema(matchJsonSchema)
                .build()));

CreateModelInteraction params =
    CreateModelInteraction.builder()
        .model(Model.of("gemini-4-carbon"))
        .input(InteractionsInput.of("Search for all details for the latest Euro."))
        .tools(Arrays.asList(GoogleSearch.builder().build(), URLContext.builder().build()))
        .responseFormat(format)
        .build();

Interaction interaction =
    client.interactions.create(CreateInteractionRequestBody.of(params)).interaction().get();

System.out.println(interaction.outputText().orElse(""));
package main

import (
    "context"
    "fmt"
    "log"

    "google.golang.org/genai"
    "google.golang.org/genai/interactions/models/interactions"
    "google.golang.org/genai/interactions/models/operations"
)

func main() {
    ctx := context.Background()
    client, err := genai.NewClient(ctx, nil)
    if err != nil {
        log.Fatal(err)
    }

    matchJsonSchema := map[string]any{
        "type": "object",
        "properties": map[string]any{
            "winner": map[string]any{
                "type":        "string",
                "description": "The name of the winner.",
            },
            "final_match_score": map[string]any{
                "type":        "string",
                "description": "The final match score.",
            },
            "scorers": map[string]any{
                "type":        "array",
                "description": "The name of the scorer.",
                "items": map[string]any{
                    "type": "string",
                },
            },
        },
        "required": []string{"winner", "final_match_score", "scorers"},
    }

    format := interactions.NewCreateModelInteractionResponseFormat(
        interactions.NewResponseFormat(interactions.TextResponseFormat{
            MimeType: interactions.TextResponseFormatMimeTypeApplicationJSON.ToPointer(),
            Schema:   matchJsonSchema,
        }),
    )

    resp, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
        Body: operations.NewCreateInteractionRequestBody(
            interactions.CreateModelInteraction{
                Model: interactions.Model("gemini-4-carbon"),
                Input: interactions.NewInteractionsInput("Search for all details for the latest Euro."),
                Tools: []interactions.Tool{
                    interactions.NewTool(interactions.GoogleSearch{}),
                    interactions.NewTool(interactions.URLContext{}),
                },
                ResponseFormat: &format,
            },
        ),
    })
    if err != nil {
        log.Fatal(err)
    }

    fmt.Println(resp.Interaction.GetOutputText())
}
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
  -H "x-goog-api-key: $GEMINI_API_KEY" \
  -H 'Content-Type: application/json' \
  -H "Api-Revision: 2026-05-20" \
  -d '{
    "model": "gemini-4-carbon",
    "input": "Search for all details for the latest Euro.",
    "tools": [{"type": "google_search"}, {"type": "url_context"}],
    "response_format": {
      "type": "text",
      "mime_type": "application/json",
      "schema": {
        "type": "object",
        "properties": {
            "winner": {"type": "string"},
            "final_match_score": {"type": "string"},
            "scorers": {"type": "array", "items": {"type": "string"}}
        },
        "required": ["winner", "final_match_score", "scorers"]
      }
    }
  }'

JSON Schema 支援

若要產生 JSON 物件,請將 response_format 設定為型別 text 的物件(或包含 物件的陣列),並將其 mime_type 設為 application/json。結構描述應放在 schema 欄位中提供。

Gemini 的結構化輸出模式支援 JSON Schema 規格的 一部分子集。

以下是支援的 type 值:

  • string:文字。
  • number:浮點數。
  • integer:整數。
  • boolean:true 或 false 值。
  • object:以鍵值對表示的結構化資料。
  • array:項目清單。
  • null:若要允許某個屬性為 null,請在型別陣列中加入 "null" (e.g., {"type": ["string", "null"]}).

下列描述性屬性有助於引導模型:

  • title:屬性的簡短說明。
  • description:屬性較長且較詳細的說明。

依類型而異的屬性

object 型別的值:

  • properties:一個物件,其中每個鍵是屬性名稱,每個值 則是該屬性的結構描述。
  • required:字串陣列,列出哪些屬性為必填。
  • additionalProperties:控制是否允許未列在 properties 中的屬性。可以是布林值或結構描述。
  • 不支援的計數限制:minProperties 與 maxProperties 在 受限解碼期間不會強制執行,伺服器會忽略它們。 如有需要,請在應用程式程式碼中驗證屬性數量。

string 型別的值:

  • enum:列出分類任務可用的特定字串集合。
  • format:指定字串的語法格式,例如 date-time、date、 time.
  • 不支援的長度限制:minLength 與 maxLength 在受限解碼 期間不會強制執行,伺服器會忽略它們。 如有需要,請在應用程式程式碼中驗證字串長度。

number 與 integer 型別的值:

  • enum:列出可用的特定數值集合。
  • minimum:最小值(包含該值)。
  • maximum:最大值(包含該值)。

array 型別的值:

  • items:定義陣列中所有項目的結構描述。
  • prefixItems:定義前 N 個項目的結構描述清單,可用來 建立類似 tuple 的結構。
  • minItems:陣列中的最少項目數。
  • maxItems:陣列中的最多項目數。
  • 不支援的陣列限制:uniqueItems、contains、minContains 與 maxContains 在受限解碼期間不會強制執行,伺服器會 忽略它們。如有需要,請在應用程式程式碼中去除重複項目 或檢查陣列內容。

結構化輸出與函式呼叫的比較

功能 主要使用情境
結構化輸出 格式化最終回應:當你希望模型的回答採用特定格式時使用。
函式呼叫 在對話過程中採取行動:當模型需要先請你執行某項任務、再給出最終回答時使用。

最佳做法

  • 清楚的說明:使用 description 欄位引導模型。
  • 明確的型別:使用具體的型別(integer、string、enum)。
  • 提示詞設計:清楚說明你希望模型做什麼。
  • 驗證:輸出雖是語法正確的 JSON,仍應一律在應用程式中 驗證其值。
  • 錯誤處理:針對符合結構描述但語意不正確的輸出, 實作完善的錯誤處理。

限制

  • 結構描述子集:並非所有 JSON Schema 功能都受支援。
  • 結構描述複雜度:過大或巢狀過深的結構描述可能會被拒絕。
  • 不強制執行的限制:minLength、maxLength、minProperties、 maxProperties、uniqueItems、contains、minContains 與 maxContains (以及它們的 snake_case 變體)不會由受限解碼文法引擎 強制執行。如果你的結構描述中含有這些關鍵字(例如 Pydantic 的 Field(min_length=...) 或 Zod 的 .min() 自動產生的),API 會在要求正規化期間自動將其移除並 忽略,而不會回傳錯誤。minItems 與 maxItems 則會 強制執行。若要強制執行其他限制,請在用戶端應用程式中驗證 輸出(例如使用 Pydantic 的 model_validate_json() 或 Zod 的 .parse())。

簡化複雜的結構描述

當你要求結構化輸出時,API 會將你的 JSON Schema 編譯成 一個限制權杖產生的狀態機。這個狀態機會對總狀態數、 分支路徑與巢狀深度設下上限。如果你的結構描述 超出這些上限,API 會以 400 INVALID_ARGUMENT 錯誤拒絕要求。

若要修正結構描述被拒絕的錯誤,請降低結構描述的結構深度與 分支程度。你可以用以下幾種方式簡化複雜的結構描述:

  • 將非結構性規則移入說明:欄位的 description 屬性 與系統指令能引導模型,又不會為輸出文法增加狀態。 如果複雜的限制讓你的結構描述超出上限,請將這些 規則寫成簡潔的說明,並在應用程式程式碼中驗證 回應。
  • 攤平巢狀物件並限制遞迴:將深層巢狀的物件階層 改為較扁平的結構。如果你在遞迴型別中使用 $defs 與 $ref, 請將遞迴參照設為選用或可為 null。避免 透過必填屬性,或 minItems 大於 零的陣列進行遞迴。
  • 減少選用欄位、列舉與聯集造成的分支:選用 屬性、enum 值與 anyOf 分支會讓輸出文法的分支路徑倍增。 請盡可能將屬性標示為 required,並在欄位可為空時 允許 null 值。精簡過大的 enum 清單、簡化 anyOf 聯集,或將限制寫入提示詞中(例如搭配 Pydantic 結構描述),作為較寬鬆的限制。
  • 縮短屬性名稱並移除陣列或數值界限:屬性鍵 消耗的狀態數與其字元長度成正比。minItems 與 maxItems 等 陣列界限會為每個計數位置複製一份項目結構描述, 數值界限(minimum 與 maximum)也會擴大 狀態空間。請使用簡潔的屬性名稱、移除未使用的欄位,並將 界限寫入提示詞中(例如搭配 Pydantic 結構描述)作為 較寬鬆的限制,或在應用程式程式碼中驗證界限。