URL 上下文¶
URL 上下文工具可讓你以 URL 的形式,為模型提供額外的上下文。在請求中加入 URL 後,模型便會存取這些頁面的內容(前提是該 URL 類型未列在限制一節中),用以豐富並強化其回應。
URL 上下文工具適用於下列工作:
-
擷取資料:從多個 URL 中擷取特定資訊,例如價格、名稱或關鍵發現。
-
比較文件:分析多份報告、文章或 PDF,找出差異並追蹤趨勢。
-
彙整與建立內容:結合多個來源 URL 的資訊,產生正確的摘要、部落格文章或報告。
-
分析程式碼與文件:指向 GitHub 儲存庫或技術文件,用來解說程式碼、產生設定說明或回答問題。
比較兩個頁面¶
以下範例說明如何比較兩個不同網站上的食譜。
# This will only work for SDK newer than 2.0.0
from google import genai
client = genai.Client()
url1 = "https://www.foodnetwork.com/recipes/ina-garten/perfect-roast-chicken-recipe-1940592"
url2 = "https://www.allrecipes.com/recipe/21151/simple-whole-roast-chicken/"
interaction = client.interactions.create(
model="gemini-4-hydrogen",
input=f"Compare the ingredients and cooking times from the recipes at {url1} and {url2}",
tools=[{"type": "url_context"}]
)
# Print the model's text response and its source annotations
for step in interaction.steps:
if step.type == "model_output":
for content_block in step.content:
if content_block.type == "text":
print(content_block.text)
if content_block.annotations:
print("\nSources:")
for annotation in content_block.annotations:
if annotation.type == "url_citation":
print(f" - {annotation.title}: {annotation.url}")
// This will only work for SDK newer than 2.0.0
import { GoogleGenAI } from "@google/genai";
const client = new GoogleGenAI({});
async function main() {
const interaction = await client.interactions.create({
model: "gemini-4-hydrogen",
input: "Compare the ingredients and cooking times from the recipes at https://www.foodnetwork.com/recipes/ina-garten/perfect-roast-chicken-recipe-1940592 and https://www.allrecipes.com/recipe/21151/simple-whole-roast-chicken/",
tools: [{ type: "url_context" }]
});
// Print the model's text response and its source annotations
for (const step of interaction.steps) {
if (step.type === 'model_output') {
for (const contentBlock of step.content || []) {
if (contentBlock.type === 'text') {
console.log(contentBlock.text);
if (contentBlock.annotations) {
console.log("\nSources:");
for (const annotation of contentBlock.annotations) {
if (annotation.type === 'url_citation') {
console.log(` - ${annotation.title}: ${annotation.url}`);
}
}
}
}
}
}
}
}
await main();
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)
}
url1 := "https://www.foodnetwork.com/recipes/ina-garten/perfect-roast-chicken-recipe-1940592"
url2 := "https://www.allrecipes.com/recipe/21151/simple-whole-roast-chicken/"
res, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
Body: operations.NewCreateInteractionRequestBody(interactions.CreateModelInteraction{
Model: interactions.Model("gemini-4-hydrogen"),
Input: interactions.NewInteractionsInput(
fmt.Sprintf("Compare the ingredients and cooking times from the recipes at %s and %s", url1, url2),
),
Tools: []interactions.Tool{
interactions.NewTool(interactions.URLContext{}),
},
}),
})
if err != nil {
log.Fatal(err)
}
// Print the model's text response and its source annotations
for _, step := range res.Interaction.Steps {
if step.ModelOutputStep != nil {
for _, contentBlock := range step.ModelOutputStep.Content {
if contentBlock.TextContent != nil {
fmt.Println(contentBlock.TextContent.Text)
if len(contentBlock.TextContent.Annotations) > 0 {
fmt.Println("\nSources:")
for _, annotation := range contentBlock.TextContent.Annotations {
if annotation.URLCitation != nil {
fmt.Printf(" - %s: %s\n", annotation.URLCitation.GetTitle(), annotation.URLCitation.GetURL())
}
}
}
}
}
}
}
}
# Specifies the API revision to avoid breaking changes when they become default
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": "Compare the ingredients and cooking times from the recipes at https://www.foodnetwork.com/recipes/ina-garten/perfect-roast-chicken-recipe-1940592 and https://www.allrecipes.com/recipe/21151/simple-whole-roast-chicken/",
"tools": [{"type": "url_context"}]
}'
運作方式¶
URL 上下文工具採用兩步驟的擷取流程,兼顧速度、成本與新資料的取得。當你提供 URL 時,工具會先嘗試從內部索引快取中擷取內容,這層快取經過高度最佳化。若某個 URL 不在索引中(例如是很新的頁面),工具會自動改採即時擷取,直接存取該 URL,立即取得其內容。
與其他工具結合使用¶
你可以將 URL 上下文工具與其他工具結合,打造更強大的工作流程。
Gemini 3 模型支援將內建工具(如 URL 上下文)與自訂工具(函式呼叫)結合使用。詳情請參閱工具組合頁面。
使用搜尋接地¶
當 URL 上下文與使用 Google 搜尋接地同時啟用時,模型可以先運用其搜尋能力,在線上尋找相關資訊,再透過 URL 上下文工具,深入理解找到的頁面。這種做法對於需要廣泛搜尋,又需要深入分析特定頁面的提示詞特別有效。
# This will only work for SDK newer than 2.0.0
from google import genai
client = genai.Client()
interaction = client.interactions.create(
model="gemini-4-hydrogen",
input="Give me three day events schedule based on {{'YOUR_URL'}}. Also let me know what needs to taken care of considering weather and commute.",
tools=[
{"type": "url_context"},
{"type": "google_search"}
]
)
for step in interaction.steps:
if step.type == "model_output":
for content_block in step.content:
if content_block.type == "text":
print(content_block.text)
// This will only work for SDK newer than 2.0.0
import { GoogleGenAI } from "@google/genai";
const client = new GoogleGenAI({});
async function main() {
const interaction = await client.interactions.create({
model: "gemini-4-hydrogen",
input: "Give me three day events schedule based on {{'YOUR_URL'}}. Also let me know what needs to taken care of considering weather and commute.",
tools: [
{ type: "url_context" },
{ type: "google_search" }
]
});
for (const step of interaction.steps) {
if (step.type === 'model_output') {
for (const contentBlock of step.content || []) {
if (contentBlock.type === 'text') console.log(contentBlock.text);
}
}
}
}
await main();
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)
}
res, err := client.Interactions.Create(ctx, operations.CreateInteractionRequest{
Body: operations.NewCreateInteractionRequestBody(interactions.CreateModelInteraction{
Model: interactions.Model("gemini-4-hydrogen"),
Input: interactions.NewInteractionsInput("Give me three day events schedule based on YOUR_URL. Also let me know what needs to taken care of considering weather and commute."),
Tools: []interactions.Tool{
interactions.NewTool(interactions.URLContext{}),
interactions.NewTool(interactions.GoogleSearch{}),
},
}),
})
if err != nil {
log.Fatal(err)
}
for _, step := range res.Interaction.Steps {
if step.ModelOutputStep != nil {
for _, contentBlock := range step.ModelOutputStep.Content {
if contentBlock.TextContent != nil {
fmt.Println(contentBlock.TextContent.Text)
}
}
}
}
}
# Specifies the API revision to avoid breaking changes when they become default
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": "Give me three day events schedule based on {{'YOUR_URL'}}. Also let me know what needs to taken care of considering weather and commute.",
"tools": [
{"type": "url_context"},
{"type": "google_search"}
]
}'
理解回應¶
當模型使用 URL 上下文工具時,其文字回應會在文字內容區塊中加入行內的 url_citation 標註。每個標註都會將回應文字中的某個區段(透過 start_index 與 end_index)連結到其來源 URL。這是在你的應用程式中呈現引用來源的主要方式——擷取方法請參閱上方的主要範例。
回應中還包含一個 url_context_result 步驟,記錄每次 URL 擷取嘗試的中繼資料(狀態、擷取的 URL)。這主要用於除錯。
安全檢查¶
系統會對 URL 執行內容審查,確認其符合安全標準。若某個 URL 未通過這項檢查,對應的 url_context_result 步驟會顯示 status 為 "unsafe"。
權杖數量¶
從你在提示詞中指定的 URL 擷取到的內容,會計入輸入權杖。你可以在互動的 usage 物件中查看權杖數量,範例如下:
'usage': {
'output_tokens': 45,
'input_tokens': 27,
'input_tokens_details': [{'modality': 'TEXT', 'token_count': 27}],
'thoughts_tokens': 31,
'tool_use_input_tokens': 10309,
'tool_use_input_tokens_details': [{'modality': 'TEXT', 'token_count': 10309}],
'total_tokens': 10412
}
每個權杖的價格取決於所使用的模型,詳情請參閱定價頁面。
支援的模型¶
| Model | URL Context |
|---|---|
| Gemini 4 Carbon | ✔️ |
| Gemini 4 Argon | ✔️ |
| Gemini 4 Hydrogen | ✔️ |
| Gemini 3.8 Flash | ✔️ |
| Gemini 3.6 Flash | ✔️ |
| Gemini 3.5 Flash-Lite | ✔️ |
| Gemini 3.1 Pro Preview | ✔️ |
| Gemini 3.1 Flash-Lite | ✔️ |
| Gemini 3 Flash Preview | ✔️ |
| Gemini 2.5 Pro | ✔️ |
| Gemini 2.5 Flash | ✔️ |
| Gemini 2.5 Flash-Lite | ✔️ |
最佳做法¶
-
提供具體的 URL:為獲得最佳效果,請提供指向你希望模型分析之內容的直接 URL。模型只會擷取你提供的 URL 內容,不會擷取巢狀連結中的任何內容。
-
檢查可存取性:確認你提供的 URL 不會導向需要登入或設有付費牆的頁面。
-
使用完整的 URL:提供完整的 URL,包括通訊協定(例如 https://www.google.com,而非僅 google.com)。
限制¶
-
請求上限:此工具每次請求最多可處理 20 個 URL。
-
URL 內容大小:從單一 URL 擷取的內容大小上限為 34MB。
-
公開存取:URL 必須能在網路上公開存取。不支援 localhost 位址(例如 localhost、127.0.0.1)、私人網路與通道服務(例如 ngrok、pinggy)。
支援與不支援的內容類型¶
此工具可從下列內容類型的 URL 擷取內容:
-
文字(text/html、application/json、text/plain、text/xml、text/css、text/javascript、text/csv、text/rtf)
-
圖片(image/png、image/jpeg、image/bmp、image/webp)
-
PDF(application/pdf)
下列內容類型不受支援:
-
付費牆內容
-
YouTube 影片(如何處理 YouTube URL,請參閱影片理解)
-
Google Workspace 檔案,例如 Google 文件或試算表
-
影片與音訊檔案