最終更新日: 2025-04-19
利用上の注意
このポストは Microsoft 社の Azure 公式ドキュメント(CC BY 4.0 または MIT ライセンス) をもとに生成AIを用いて翻案・要約した派生作品です。 元の文書は MicrosoftDocs/azure-ai-docs にホストされています。
生成AIの性能には限界があり、誤訳や誤解釈が含まれる可能性があります。 本ポストはあくまで参考情報として用い、正確な情報は必ず元の文書を参照してください。
このポストで使用されている商標はそれぞれの所有者に帰属します。これらの商標は技術的な説明のために使用されており、商標権者からの公式な承認や推奨を示すものではありません。
# Highlights
このコード差分レビューは、Azure AI Searchに関連する一連のドキュメントのマイナーな更新に焦点を当てています。主要なアップデートとしては、特定の文書の日付の更新、モデルカタログの明確化、スキルセットの詳細化、地域サポート情報の正確性の向上、ならびにベクトル検索および量子化に関する情報の充実が含まれます。
特に大きな破壊的変更はありませんが、ドキュメントの用語や表現の改訂が行われ、ユーザーが誤解しないように配慮されています。
このコード差分によるドキュメント更新は、主に利用者の理解を深め、情報の正確性を向上させることを目的としています。これまで以上にユーザーはAMLスキルやAIサービスの能力、地域サポート、量子化手法などに関する詳細な情報を容易に得られるようになりました。
たとえば、「Azure AI Foundryモデルカタログ」という用語の追加は、どのカタログを指しているのかをより明確に示すことで、ユーザーが適切にサービスを利用できるよう配慮されています。また、スキルセットの作成時に参照するべき詳細情報へのリンクも新たに追加され、複雑なスキルセットの定義やインデックスの作成をサポートします。
さらに、ベクトル検索の量子化での圧縮技術についての情報拡充は、特にリソースの最適化を重視する開発者にとって有益であり、Azure AI Searchの利用コストを大幅に削減する可能性を秘めています。
まとめると、この一連の更新は、ユーザーにとってAzure AIの機能を最大限に活用するための手助けとなり、実装から運用までのプロセスをよりスムーズに行えるよう整備されたものです。
| Filename | Type | Title | Status | A | D | M |
|---|---|---|---|---|---|---|
| cognitive-search-aml-skill.md | minor update | AMLスキルの更新: モデルカタログの明確化 | modified | 2 | 2 | 4 |
| cognitive-search-defining-skillset.md | minor update | スキルセット作成に関するドキュメントの更新 | modified | 2 | 2 | 4 |
| cognitive-search-skill-vision-vectorize.md | minor update | Azure AI Visionマルチモーダル埋め込みスキルのドキュメント更新 | modified | 38 | 24 | 62 |
| search-get-started-portal-import-vectors.md | minor update | Azureポータルでのベクトルのインポートに関するクイックスタートガイドの更新 | modified | 3 | 3 | 6 |
| search-region-support.md | minor update | Azure AIサービス統合に関する地域サポートの記事の更新 | modified | 9 | 9 | 18 |
| vector-search-how-to-create-index.md | minor update | ベクトル検索インデックス作成に関するドキュメントの更新 | modified | 71 | 55 | 126 |
| vector-search-how-to-quantization.md | minor update | ベクトル検索における量子化に関するドキュメントの更新 | modified | 3 | 0 | 3 |
@@ -10,15 +10,15 @@ ms.custom:
- ignite-2023
- build-2024
ms.topic: reference
-ms.date: 08/05/2024
+ms.date: 04/18/2025
---
# AML skill in an Azure AI Search enrichment pipeline
> [!IMPORTANT]
> Support for indexer connections to the Azure AI Foundry model catalog is in public preview under [supplemental terms of use](https://azure.microsoft.com/support/legal/preview-supplemental-terms/). Preview REST APIs support this skill.
-The **AML** skill allows you to extend AI enrichment with a custom [Azure Machine Learning (AML)](../machine-learning/overview-what-is-azure-machine-learning.md) model or deployed base embedding model on Azure AI Foundry. Once an AML model is [trained and deployed](../machine-learning/concept-azure-machine-learning-architecture.md#workspace), an **AML** skill integrates it into a skillset.
+The **AML** skill allows you to extend AI enrichment with a custom [Azure Machine Learning (AML)](../machine-learning/overview-what-is-azure-machine-learning.md) model or deployed base embedding model in the Azure AI Foundry model catalog. Once an AML model is [trained and deployed](../machine-learning/concept-azure-machine-learning-architecture.md#workspace), an **AML** skill integrates it into a skillset.
## AML skill usage
{
"modification_type": "minor update",
"modification_title": "AMLスキルの更新: モデルカタログの明確化"
}この修正では、Azure AI SearchのAMLスキルに関する文書が更新されました。主な変更点は、文中の「Azure AI Foundryモデルカタログ」という表現が追加されることで、ユーザーがAMLスキルの機能をより明確に理解できるようになっています。具体的には、「deployed base embedding model on Azure AI Foundry」との記述が「deployed base embedding model in the Azure AI Foundry model catalog」に変更され、参照先が明確になりました。また、日付が2024年8月5日から2025年4月18日に更新され、内容の適時性が保たれるようにされています。この変更は、文書の一貫性とユーザーに対する情報提供の質を向上させることを目的としています。
@@ -8,7 +8,7 @@ ms.service: azure-ai-search
ms.custom:
- ignite-2023
ms.topic: conceptual
-ms.date: 12/06/2024
+ms.date: 04/18/2025
---
# Create a skillset in Azure AI Search
@@ -192,7 +192,7 @@ Skills read from and write to an enriched document. Skill inputs specify the ori
| `source`: `/document/some-named-field` | For text-based skills, such as entity recognition or key phrase extraction, the origin should be a field that contains sufficient text to be analyzed, such as a *description* or *summary*. |
| `source`: `/document/normalized_images/*` | For image content, the source is image that's been normalized during document cracking. |
-If the skill iterates over an array, both context and input source should include `/*` in the correct positions.
+If the skill iterates over an array, both context and input source should include `/*` in the correct positions. For more information about the complete syntax, see [Skill context and input annotation language](cognitive-search-skill-annotation-language.md).
## Define outputs
{
"modification_type": "minor update",
"modification_title": "スキルセット作成に関するドキュメントの更新"
}この修正は、Azure AI Searchにおけるスキルセットの作成に関するドキュメントに対するマイナーな更新です。主な変更点として、ドキュメントの日付が2024年12月6日から2025年4月18日に更新され、情報の適時性が確保されています。また、スキルの説明に新たに、「スキルのコンテキストと入力アノテーション言語に関する完全な構文については、Skill context and input annotation languageを参照してください」という文が追加され、ユーザーがより詳細な情報にアクセスできるようになっています。この変更は、ユーザーの理解を深め、ドキュメントの有用性を向上させることを目的としています。
@@ -9,44 +9,55 @@ ms.custom:
- build-2024
- references_regions
ms.topic: reference
-ms.date: 08/05/2024
+ms.date: 04/18/2025
---
# Azure AI Vision multimodal embeddings skill
> [!IMPORTANT]
-> This skill is in public preview under [Supplemental Terms of Use](https://azure.microsoft.com/support/legal/preview-supplemental-terms/). The [2024-05-01-Preview REST API](/rest/api/searchservice/skillsets/create-or-update?view=rest-searchservice-2024-05-01-Preview&preserve-view=true) supports this feature.
+> This skill is in public preview under [Supplemental Terms of Use](https://azure.microsoft.com/support/legal/preview-supplemental-terms/). The [2024-05-01-Preview REST API](/rest/api/searchservice/skillsets/create-or-update?view=rest-searchservice-2024-05-01-Preview&preserve-view=true) and newer preview APIs support this feature.
The **Azure AI Vision multimodal embeddings** skill uses Azure AI Vision's [multimodal embeddings API](/azure/ai-services/computer-vision/concept-image-retrieval) to generate embeddings for image or text input.
-The skill is only supported in search services located in a region that supports the [Azure AI Vision Multimodal embeddings API](/azure/ai-services/computer-vision/how-to/image-retrieval). Review [region availability for multimodal embeddings](/azure/ai-services/computer-vision/overview-image-analysis?tabs=4-0#region-availability). Your data is processed in the [Geo](https://azure.microsoft.com/explore/global-infrastructure/data-residency/) where your model is deployed.
+This skill must be [attached to a billable Azure AI multi-service resource](cognitive-search-attach-cognitive-services.md) for transactions that exceed 20 documents per indexer per day. Execution of built-in skills is charged at the existing [Azure AI services pay-as-you go price](https://azure.microsoft.com/pricing/details/cognitive-services/).
-> [!NOTE]
-> This skill is bound to Azure AI services and requires [a billable resource](cognitive-search-attach-cognitive-services.md) for transactions that exceed 20 documents per indexer per day. Execution of built-in skills is charged at the existing [Azure AI services pay-as-you go price](https://azure.microsoft.com/pricing/details/cognitive-services/).
->
-> In addition, image extraction is [billable by Azure AI Search](https://azure.microsoft.com/pricing/details/search/).
->
+In addition, image extraction is [billable by Azure AI Search](https://azure.microsoft.com/pricing/details/search/).
+
+Location of resources is an important consideration. Because you're using a preview API version to create a skillset that contains preview skills, you have the option of a [keyless connection](cognitive-search-attach-cognitive-services.md#bill-through-a-keyless-connection), which relaxes the region requirement. However, if you're connecting with an API key, then Azure AI Search and Azure AI multi-service must be in the same region.
+
++ First, find a [supported region for multimodal embeddings](/azure/ai-services/computer-vision/overview-image-analysis?tabs=4-0#region-availability).
+
++ Second, verify the [region provides AI enrichment](search-region-support.md).
+
+The Azure AI multi-service resource is used for billing purposes only. Content processing occurs on separate resources managed and maintained by Azure AI Search within the same geo. Your data is processed in the [Geo](https://azure.microsoft.com/explore/global-infrastructure/data-residency/) where your resource is deployed.
## @odata.type
Microsoft.Skills.Vision.VectorizeSkill
## Data limits
-The input limits for the skill can be found in [the Azure AI Vision documentation](/azure/ai-services/computer-vision/concept-image-retrieval#input-requirements) for images and text respectively. Consider using the [Text Split skill](cognitive-search-skill-textsplit.md) if you need data chunking for text inputs.
+The input limits for the skill can be found in the [Azure AI Vision documentation](/azure/ai-services/computer-vision/concept-image-retrieval#input-requirements) for images and text respectively. Consider using the [Text Split skill](cognitive-search-skill-textsplit.md) if you need data chunking for text inputs.
+
+Applicable inputs include:
+
++ Image input file size must be less than 20 megabytes (MB). Image size must be greater than 10 x 10 pixels and less than 16,000 x 16,000 pixels.
++ Text input string must be between (inclusive) one word and 70 words.
## Skill parameters
Parameters are case-sensitive.
| Inputs | Description |
|---------------------|-------------|
-| `modelVersion` | (Required) The model version to be passed to the Azure AI Vision multimodal embeddings API for generating embeddings. It's important that all embeddings stored in a given index field are generated using the same `modelVersion`. For information about version support for this model, refer to [multimodal embeddings](/azure/ai-services/computer-vision/concept-image-retrieval#what-are-vector-embeddings).|
+| `modelVersion` | (Required) The model version (`2023-04-15`) to be passed to the Azure AI Vision multimodal embeddings API for generating embeddings. Vector embeddings can only be compared and matched if they're from the same model type. Images vectorized by one model won't be searchable through a different model. The latest Image Analysis API offers two models, version `2023-04-15` which supports text search in many languages, and the legacy `2022-04-11` model which supports only English. Azure AI Search uses the newer version. |
## Skill inputs
+Skill definition inputs include name, source, and inputs. The following table provides valid values for name of the input. You can also specify recursive inputs. For more information, see the [REST API reference](/rest/api/searchservice/skillsets/create?view=rest-searchservice-2025-03-01-preview#inputfieldmappingentry&preserve-view=true) and [Create a skillset](cognitive-search-defining-skillset.md).
+
| Input | Description |
-|--------------------|-------------|
+|--------|-------------|
| `text` | The input text to be vectorized. If you're using data chunking, the source might be `/document/pages/*`. |
| `image` | Complex Type. Currently only works with "/document/normalized_images" field, produced by the Azure blob indexer when ```imageAction``` is set to a value other than ```none```. |
| `url` | The URL to download the image to be vectorized. |
@@ -62,15 +73,15 @@ Only one of `text`, `image` or `url`/`queryString` can be configured for a singl
## Sample definition
-For text input, consider a record that has the following fields:
+For text input, consider a blob that has the following content:
```json
{
- "content": "Microsoft released Windows 10."
+ "content": "Forests, grasslands, deserts, and mountains are all part of the Patagonian landscape that spans more than a million square kilometers of South America."
}
```
-Then your skill definition might look like this:
+For text inputs, your skill definition might look like this:
```json
{
@@ -85,14 +96,15 @@ Then your skill definition might look like this:
],
"outputs": [
{
- "name": "vector"
+ "name": "vector",
+ "targetName": "text_vector"
}
]
}
```
-For image input, your skill definition might look like this:
+For image input, a second skill definition in the same skillset might look like this:
```json
{
@@ -107,13 +119,14 @@ For image input, your skill definition might look like this:
],
"outputs": [
{
- "name": "vector"
+ "name": "vector",
+ "targetName": "image_vector"
}
]
}
```
-If you want to vectorize images directly from your blob storage datasource, your skill definition might look like this:
+If you want to vectorize images directly from your blob storage data source rather than extract images during indexing, your skill definition should specify a URL, and perhaps a SAS token depending on storage security. For this scenario, your skill definition might look like this:
```json
{
@@ -132,19 +145,20 @@ If you want to vectorize images directly from your blob storage datasource, your
],
"outputs": [
{
- "name": "vector"
+ "name": "vector",
+ "targetName": "image_vector"
}
]
}
```
## Sample output
-For the given input text, a vectorized embedding output is produced.
+For the given input, a vectorized embedding output is produced. Output is 1,024 dimensions, which is the number of dimensions supported by the Azure AI Vision multimodal API.
```json
{
- "vector": [
+ "text_vector": [
0.018990106880664825,
-0.0073809814639389515,
....
@@ -153,7 +167,7 @@ For the given input text, a vectorized embedding output is produced.
}
```
-The output resides in memory. To send this output to a field in the search index, you must define an [outputFieldMapping](cognitive-search-output-field-mapping.md) that maps the vectorized embedding output (which is an array) to a [vector field](vector-search-how-to-create-index.md). Assuming the skill output resides in the document's **vector** node, and **content_vector** is the field in the search index, the outputFieldMapping in indexer should look like:
+The output resides in memory. To send this output to a field in the search index, you must define an [outputFieldMapping](cognitive-search-output-field-mapping.md) that maps the vectorized embedding output (which is an array) to a [vector field](vector-search-how-to-create-index.md). Assuming the skill output resides in the document's **vector** node, and **content_vector** is the field in the search index, the outputFieldMapping in the indexer should look like:
```json
"outputFieldMappings": [
@@ -164,7 +178,7 @@ The output resides in memory. To send this output to a field in the search index
]
```
-For mapping image embeddings to the index, you'll need to use the [Index Projections](index-projections-concept-intro.md) feature. The payload for `indexProjections` might look something like this:
+For mapping image embeddings to the index, you use [index projections](index-projections-concept-intro.md). The payload for `indexProjections` might look something like the following example. image_content_vector is a field in the index, and it's populated with the content found in the **vector** of the **normalized_images** array.
```json
"indexProjections": {
@@ -175,7 +189,7 @@ For mapping image embeddings to the index, you'll need to use the [Index Project
"sourceContext": "/document/normalized_images/*",
"mappings": [
{
- "name": "content_vector",
+ "name": "image_content_vector",
"source": "/document/normalized_images/*/vector"
}
]{
"modification_type": "minor update",
"modification_title": "Azure AI Visionマルチモーダル埋め込みスキルのドキュメント更新"
}この修正は、Azure AI Visionのマルチモーダル埋め込みスキルに関するドキュメントに大幅な改訂が行われたものです。主要な変更点は、日付の更新（2024年8月5日から2025年4月18日）および、スキルの使用に関する重要な情報の追加です。このスキルは、Azure AIマルチサービスリソースに結び付けられる必要があり、毎日20ドキュメントを超えるトランザクションには料金が発生することを明確に述べています。また、スキルの定義セクションにおいて、スキルの入力要件や出力マッピングに関する具体的な記述が追加され、ユーザーが正しい構文でスキルを定義できるようになっています。さらに、処理されるデータの地域やプレビューAPIバージョンに関する情報もアップデートされ、ユーザーが適切な設定を行えるよう配慮されています。全体として、この変更はスキルの透明性とユーザーの利便性を向上させることを目的としています。
@@ -9,7 +9,7 @@ ms.custom:
- build-2024
- ignite-2024
ms.topic: quickstart
-ms.date: 03/04/2025
+ms.date: 04/18/2025
---
# Quickstart: Vectorize text and images in the Azure portal
@@ -189,9 +189,9 @@ The wizard supports text-embedding-ada-002, text-embedding-3-large, and text-emb
The wizard supports Azure AI Vision image retrieval through multimodal embeddings (version 4.0). Internally, the wizard calls the [multimodal embeddings skill](cognitive-search-skill-vision-vectorize.md) to connect to Azure AI Vision.
-1. [Create an Azure AI Vision service in a supported region](/azure/ai-services/computer-vision/how-to/image-retrieval?tabs=csharp#prerequisites).
+1. [Create an Azure AI multi-service resource](/azure/ai-services/multi-service-resource?pivots=azportal#azure-ai-services-resource-for-azure-ai-search-skills). [Choose a region](/azure/ai-services/computer-vision/overview-image-analysis#region-availability) that provides the multimodal embeddings model.
-1. Make sure your Azure AI Search service is in the same region.
+1. Make sure your Azure AI Search service is in the same region, and the [region supports AI enrichment](search-region-support.md).
1. After the service is deployed, go to the resource and select **Access control** to assign the **Cognitive Services User** role to your search service's managed identity. Optionally, you can use key-based authentication for the connection.
{
"modification_type": "minor update",
"modification_title": "Azureポータルでのベクトルのインポートに関するクイックスタートガイドの更新"
}この修正は、Azureポータルにおけるテキストと画像のベクトル化に関するクイックスタートガイドに対するマイナーな更新です。主な変更点は、ドキュメントの日付が2025年3月4日から2025年4月18日に更新されたことです。また、Azure AI Visionサービスの作成に関する手順が明確化され、ユーザーが「Azure AIマルチサービスリソースを作成し、マルチモーダル埋め込みモデルを提供する地域を選択する」ことが強調されています。さらに、Azure AI Searchサービスが同じ地域に設置されていることに加え、その地域がAIエンリッチメントをサポートしていることも確認するように指示が追加されています。この変更は、ユーザーがサービスを適切にセットアップするための理解を助けることを目的としています。
@@ -27,7 +27,7 @@ Some features take a dependency on other Azure services or infrastructure that a
| [Availability zones](search-reliability.md#availability-zone-support) | Divides a region's data centers into distinct physical location groups, providing high availability within the same geo. | Regional support is noted in this article. |
| [Semantic ranker](semantic-search-overview.md) | Takes a dependency on Microsoft-hosted models in specific regions. | Regional support is noted in this article. |
| [Query rewrite](semantic-how-to-query-rewrite.md) | Takes a dependency on Microsoft-hosted models in specific regions. | Regional support is noted in this article. |
-| [AI service integration](cognitive-search-concept-intro.md) | Refers to [built-in skills](cognitive-search-predefined-skills.md) that make internal calls to Azure AI for enrichment and transformation during indexing. Integration requires that Azure AI Search coexists with an [Azure AI services multi-service account](/azure/ai-services/multi-service-resource#azure-ai-services-resource-for-azure-ai-search-skills) in the same physical region. You can bypass region requirements if you use [identity-based connections](cognitive-search-attach-cognitive-services.md#bill-through-a-keyless-connection), currently in public preview. | Regional support is noted in this article. |
+| [AI enrichment](cognitive-search-concept-intro.md) | Refers to [built-in skills](cognitive-search-predefined-skills.md) that make internal calls to Azure AI for enrichment and transformation during indexing. Integration requires that Azure AI Search coexists with an [Azure AI services multi-service account](/azure/ai-services/multi-service-resource#azure-ai-services-resource-for-azure-ai-search-skills) in the same physical region. You can bypass region requirements if you use [identity-based connections](cognitive-search-attach-cognitive-services.md#bill-through-a-keyless-connection), currently in public preview. | Regional support is noted in this article. |
| [Azure OpenAI integration](vector-search-integrated-vectorization.md) | Refers to the AzureOpenAIEmbedding skill and vectorizer that make internal calls to deployed embedding models on Azure OpenAI. | Check [Azure OpenAI model region availability](/azure/ai-services/openai/concepts/models#model-summary-table-and-region-availability) for the most current list of regions for each embedding and chat model. Specific Azure OpenAI models are in fewer regions, so check for model availability first, and then verify Azure AI Search is available in the same region.|
| [Azure AI Foundry integration](vector-search-integrated-vectorization-ai-studio.md) | Refers to skills and vectorizers that make internal calls to the models hosted in the model catalog. | Check [Azure AI Foundry region availability](/azure/ai-foundry/reference/region-support) for the most current list of regions. |
| [Azure AI Vision 4.0 multimodal APIs](search-get-started-portal-image-search.md) | Refers to the Azure AI Vision multimodal embeddings skill and vectorizer that call the multimodal embedding API. | Check the [Azure AI Vision region list](/azure/ai-services/computer-vision/overview-image-analysis#region-availability) first, and then verify Azure AI Search is available in the same region.|
@@ -36,11 +36,11 @@ Some features take a dependency on other Azure services or infrastructure that a
You can create an Azure AI Search resource in any of the following Azure public regions. Almost all of these regions support [higher capacity tiers](search-limits-quotas-capacity.md#service-limits). Exceptions are noted where they apply.
-AI service integration refers to internal connections to an Azure AI services multi-service account and doesn't include Azure OpenAI integration.
+AI enrichment refers to internal connections to an Azure AI services multi-service account and doesn't include Azure OpenAI integration.
### Americas
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| Brazil South | ✅ | | ✅ | |
| Canada Central | ✅ | ✅ | ✅ | |
@@ -58,7 +58,7 @@ AI service integration refers to internal connections to an Azure AI services mu
### Europe
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| North Europe <sup>1</sup> | ✅ | ✅ | ✅ | ✅ |
| West Europe | ✅ | ✅ | ✅ | |
@@ -80,7 +80,7 @@ AI service integration refers to internal connections to an Azure AI services mu
### Middle East
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| Israel Central <sup>1</sup> | | ✅ | | |
| Qatar Central <sup>1</sup> | | ✅ | | |
@@ -90,13 +90,13 @@ AI service integration refers to internal connections to an Azure AI services mu
### Africa
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| South Africa North | ✅ | ✅ | | |
### Asia Pacific
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| Australia East | ✅ | ✅ | ✅ | |
| Australia Southeast | | | ✅ | |
@@ -115,15 +115,15 @@ AI service integration refers to internal connections to an Azure AI services mu
## Azure Government regions
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| Arizona | ✅ | | ✅ | |
| Texas | | | | |
| Virginia | ✅ | ✅ | ✅ | |
## Azure operated by 21Vianet
-| Region | AI service integration | Availability zones | Semantic ranker | Query rewrite |
+| Region | AI enrichment | Availability zones | Semantic ranker | Query rewrite |
|--|--|--|--|--|
| China East | | | | |
| China East 2 <sup>1</sup> | ✅ | | | |{
"modification_type": "minor update",
"modification_title": "Azure AIサービス統合に関する地域サポートの記事の更新"
}この修正は、Azure AIサービス統合に関連する地域サポートに関する記事の内容を更新したものです。主な変更点は、「AIサービス統合」という表現を「AIエンリッチメント」に置き換えることで、文書の内容がより正確に反映されるよう改訂されました。また、表のヘッダーにも同様の変更が施され、各地域におけるAIエンリッチメントのサポート状況がすっきりと整理されています。この変更により、ユーザーがAzure AI関連のリソースを適切に利用できる場所を把握しやすくなり、サービスの使用にあたっての理解が深まることを目的としています。全体として、地域サポートに関する情報が明確化され、ユーザーの利便性が向上しています。
@@ -9,20 +9,20 @@ ms.service: azure-ai-search
ms.custom:
- ignite-2024
ms.topic: how-to
-ms.date: 02/14/2025
+ms.date: 04/17/2025
---
# Create a vector index
-In Azure AI Search, you can store vectors in a search index and send vector queries to match on semantic similarity. A vector store in Azure AI Search is defined by an index schema that has both vector and nonvector fields. The schema also has a vector configuration for specifying the algorithms used to create and compress the embedding space.
+In Azure AI Search, you can store vectors in a search index and send vector queries for matching based on semantic similarity. A vector index is defined by an index schema that has: vector fields, nonvector fields, and a vector configuration section.
[Create or Update Index API](/rest/api/searchservice/indexes/create-or-update) creates the vector store. Follow these steps to index vectors in Azure AI Search:
> [!div class="checklist"]
> + Start with a basic schema definition
> + Add vector algorithms and optional compression
> + Add vector field definitions
-> + Load prevectorized data [as a separate step](#load-vector-data-for-indexing), or use [integrated vectorization](vector-search-integrated-vectorization.md) for data chunking and encoding during indexing
+> + Load prevectorized data [as a separate step](#load-vector-data-for-indexing), or use [integrated vectorization](vector-search-integrated-vectorization.md) for data chunking and embedding during indexing
This article explains the workflow using the REST API for illustration. Once you understand the basic workflow, continue with the Azure SDK code samples in the [azure-search-vector-samples](https://github.com/Azure/azure-search-vector-samples) repository for guidance on using vectors in test and production code.
@@ -37,35 +37,37 @@ This article explains the workflow using the REST API for illustration. Once you
+ You should know the dimensions limit of the model used to create the embeddings so that you can assign that limit to the vector field. For **text-embedding-ada-002**, dimensions are fixed at 1536. For **text-embedding-3-small** or **text-embedding-3-large**, dimensions range from 1 to 1536 and 3072, respectively.
-+ You should also know what similarity metric to use. For embedding models on Azure OpenAI, similarity is [computed using `cosine`](/azure/ai-services/openai/concepts/understand-embeddings#cosine-similarity).
++ You should also know what similarity metric to use. For embedding models on Azure OpenAI, similarity is computed using [`cosine`](/azure/ai-services/openai/concepts/understand-embeddings#cosine-similarity).
-+ You should be familiar with [creating an index](search-how-to-create-search-index.md). The schema must include a field for the document key, other fields you want to search or filter, and other configurations for behaviors needed during indexing and queries.
++ You should be familiar with [creating an index](search-how-to-create-search-index.md). A schema always includes a field for the document key, fields used for search or filters, and other configurations for behaviors needed during indexing and queries.
## Limitations
For search services created before January 2019, there's a small subset that can't create a vector index. If this applies to you, create a new service to use vectors.
## Prepare documents for indexing
-Before indexing, assemble a document payload that includes fields of vector and nonvector data. The document structure must conform to the index schema.
+Before indexing, assemble a document payload that includes fields of vector and nonvector data. The document structure must conform to the fields collection of index schema.
Make sure your source documents provide the following content:
| Content | Description |
|---------|-------------|
| Unique identifier | A field or a metadata property that uniquely identifies each document. All search indexes require a document key. To satisfy document key requirements, a source document must have one field or property uniquely identifies it in the index. If you're indexing blobs, it might be the metadata_storage_path that uniquely identifies each blob. If you're indexing from a database, it might be primary key. This source field must be mapped to an index field of type `Edm.String` and `key=true` in the search index.|
-| Non-vector content | Provide other fields with human-readable content. Human readable content is useful for the query response, and for hybrid query scenarios that include full text search or semantic ranking in the same request. If you're using a chat completion model, most models like ChatGPT don't accept raw vectors as input. |
-| Vector content| A vectorized version of your non-vector content. A vector is an array of single-precision floating point numbers generated by an embedding model. Each vector field contains an array generated by an embedding model, one embedding per field, where the field is a top-level field (not part of a nested or complex type). For the simplest integration, we recommend the embedding models in [Azure OpenAI](https://aka.ms/oai/access), such as a **text-embedding-3** model for text documents or the [Image Retrieval REST API](/rest/api/computervision/image-retrieval) for images and multimodal embeddings. <br><br>If you can take a dependency on indexers and skillsets, consider using [integrated vectorization](vector-search-integrated-vectorization.md) that encodes images and textual content during indexing. Your field definitions are for vector fields, but incoming source data can be text or images, which are converted to vector arrays during indexing. |
+| Non-vector content | Provide other fields with human-readable content. Human readable content is useful for the query response, and for hybrid query scenarios that include full text search or semantic ranking in the same request. If you're using a chat completion model, most models like ChatGPT expect human-readable text and don't accept raw vectors as input. |
+| Vector content| A vectorized version of your non-vector content. Vectors are used at query time. A vector is an array of single-precision floating point numbers generated by an embedding model. Each vector field contains an array generated by an embedding model, one embedding per field, where the field is a top-level field (not part of a nested or complex type). For the simplest integration, we recommend the embedding models in [Azure OpenAI](https://aka.ms/oai/access), such as a **text-embedding-3** model for text documents or the [Image Retrieval REST API](/rest/api/computervision/image-retrieval) for images and multimodal embeddings. <br><br>If you can take a dependency on indexers and skillsets, consider using [integrated vectorization](vector-search-integrated-vectorization.md) that encodes images and textual content during indexing. Your field definitions are for vector fields, but incoming source data can be text or images, which are converted to vector arrays during indexing. |
-Your search index should include fields and content for all of the query scenarios you want to support. Suppose you want to search or filter over product names, versions, metadata, or addresses. In this case, vector similarity search isn't especially helpful. Keyword search, geo-search, or filters would be a better choice. A search index that includes a comprehensive collection of both vector and nonvector fields provides maximum flexibility for query construction and response composition.
+Your search index should include fields and content for all of the query scenarios you want to support. Suppose you want to search or filter over product names, versions, metadata, or addresses. In this case, vector similarity search isn't especially helpful. Keyword search, geo-search, or filters that iterate over verbatim content would be a better choice. A search index that's inclusive of both vector and nonvector fields provides maximum flexibility for query construction and response composition.
A short example of a documents payload that includes vector and nonvector fields is in the [load vector data](#load-vector-data-for-indexing) section of this article.
## Start with a basic index
Start with a minimum schema so that you have a definition to work with before adding a vector configuration and vector fields. A simple index might look the following example. You can learn more about an index schema in [Create a search index](search-how-to-create-search-index.md).
-Notice that it has a required name, a required document key (`"key": true`), and fields for human readable content in plain text. It's common to have a human-readable version of whatever content you intend to vectorize. For example, if you have a chunk of text from a PDF file, your index schema should have the plain text equivalent of the vectorized text.
+Notice that it has a required name, a required document key (`"key": true`), and fields for human readable content in plain text. It's common to have a human-readable version of whatever content you intend to vectorize. For example, if you have a chunk of text from a PDF file, your index schema should have a field for plain text chunks and a second field for vectorized chunks.
+
+Here's a basic index with a "name", a "fields" collection, and a few other constructs for extra configuration.
```http
POST https://[servicename].search.windows.net/indexes?api-version=[api-version]
@@ -76,35 +78,35 @@ POST https://[servicename].search.windows.net/indexes?api-version=[api-version]
{ "name": "myHumanReadableNameField", "type": "Edm.String", "retrievable": true, "searchable": true, "filterable": false, "sortable": true, "facetable": false },
{ "name": "myHumanReadableContentField", "type": "Edm.String", "retrievable": true, "searchable": true, "filterable": false, "sortable": false, "facetable": false, "analyzer": "en.microsoft" },
],
- "suggesters": [ ],
+ "analyzers": [ ],
"scoringProfiles": [ ],
- "analyzers":(optional)[ ... ]
+ "suggesters": [ ],
+ "vectorSearch": [ ]
}
```
## Add a vector search configuration
-Next, add a vector search configuration (profile) to your schema. Configuration occurs before field definitions because you specify a profile on each field as part of its definition. In the schema, vector configuration is typically added after the fields collection, perhaps after `"suggesters"`, `"scoringProfiles"`, or `"analyzers"`.
-
-A vector configuration specifies the parameters used during indexing to create "nearest neighbor" information among the vector nodes. Algorithms include:
-
-+ Hierarchical Navigable Small World (HNSW)
-+ Exhaustive k-Nearest Neighbor (KNN)
-
-If you choose HNSW on a field, you can opt in for exhaustive KNN at query time. But the other direction doesn’t work: if you choose exhaustive for indexing, you can’t later request HNSW search because the extra data structures that enable approximate search don’t exist.
+Next, add a "vectorSearch" configuration to your schema. It's useful to specify a configuration first, before field definitions, because the profiles you define here become part of the vector field's definition. In the schema, vector configuration is typically inserted after the fields collection, perhaps after `"analyzers"`, `"scoringProfiles"`, and `"suggesters"` (although order doesn't matter).
-Optionally, a vector configuration also specifies quantization methods for reducing vector size:
+A vector configuration includes:
-+ Scalar
-+ Binary (available in 2024-07-01 only and in newer Azure SDK packages)
++ `vectorSearch.algorithms` used during indexing to create "nearest neighbor" information among the vector nodes.
++ `vectorSearch.compressions` for scalar or binary quantization, oversampling, and for reranking with original vectors.
++ `vectorSearch.profiles` for specifying multiple combinations of algorithm and compression configurations.
### [**2024-07-01**](#tab/config-2024-07-01)
[**2024-07-01**](/rest/api/searchservice/search-service-api-versions#2024-07-01) is generally available. It supports a vector configuration having:
-+ `vectorSearch.algorithms` support HNSW and exhaustive KNN.
-+ `vectorSearch.compressions` support scalar and binary quantization, oversampling, and reranking with original vectors.
-+ `vectorSearch.profiles` for specifying multiple combinations of algorithm and compression configurations.
++ Hierarchical Navigable Small World (HNSW) algorithm
++ Exhaustive k-Nearest Neighbor (KNN) algorithm
++ Scalar compression
++ Binary compression (available in 2024-07-01 only and in newer Azure SDK packages)
++ Oversampling
++ Reranking with original vectors
+
+If you choose HNSW on a field, you can opt in for exhaustive KNN at query time. But the other direction doesn’t work: if you choose exhaustive for indexing, you can’t later request HNSW search because the extra data structures that enable approximate search don’t exist.
Be sure to have a strategy for [vectorizing your content](vector-search-how-to-generate-embeddings.md). We recommend [integrated vectorization](vector-search-integrated-vectorization.md) and [query-time vectorizers](vector-search-how-to-configure-vectorizer.md) for built-in encoding.
@@ -175,7 +177,7 @@ Be sure to have a strategy for [vectorizing your content](vector-search-how-to-g
+ Names for each configuration of compression, algorithm, and profile must be unique for its type within the index.
- + `vectorSearch.compressions` can be `scalarQuantization` or `binaryQuantization`. Scalar quantization compresses float values into narrower data types. Binary quantization converts floats into binary 1 bit values.
+ + `vectorSearch.compressions` can be `scalarQuantization` or `binaryQuantization`. Scalar quantization compresses float values into narrower data types. Binary quantization converts floats into binary 1-bit values.
+ `vectorSearch.compressions.rerankWithOriginalVectors` uses the original, uncompressed vectors to recalculate similarity and rerank the top results returned by the initial search query. The uncompressed vectors exist in the search index even if `stored` is false. This property is optional. Default is true.
@@ -197,14 +199,9 @@ Be sure to have a strategy for [vectorizing your content](vector-search-how-to-g
### [**2024-05-01-preview**](#tab/config-2024-05-01-Preview)
-[**2024-05-01-preview**](/rest/api/searchservice/search-service-api-versions#2024-05-01-preview) is the most recent preview version.
+[**2024-05-01-preview**](/rest/api/searchservice/search-service-api-versions#2024-05-01-preview) is the most recent preview version. It's inclusive of previous preview versions.
-+ `vectorSearch.algorithms` with support for HNSW and exhaustive KNN.
-+ `vectorSearch.compressions` with properties for scalar (but not binary) quantization, oversampling, and reranking with original vectors.
-+ `vectorSearch.profiles` for multiple combinations of algorithm and compression configurations.
-+ Inclusive of 2024-03-01-preview.
-+ Inclusive of 2023-10-01-preview.
-+ Inclusive of 2023-11-01 `vectorSearch.algorithms` and `vectorSearch.profiles`.
+Preview and stable API versions support the same `"vectorSearch"` configurations. You would choose the preview over the stable version for other reasons, such as [more compression options](vector-search-how-to-quantization.md) or [newer vectorizers](vector-search-how-to-configure-vectorizer.md) invoked at query time.
1. Use the [Create or Update Index Preview REST API](/rest/api/searchservice/indexes/create-or-update?view=rest-searchservice-2024-05-01-preview&preserve-view=true) to create the index.
@@ -278,7 +275,7 @@ Be sure to have a strategy for [vectorizing your content](vector-search-how-to-g
+ `vectorSearch.algorithms.efConstruction` is the number of nearest neighbors used during indexing. Default is 400. The range is 100 to 1,000.
- + `"vectorSearch.algorithms.fSearch` is the number of nearest neighbors used during search. Default is 500. The range is 100 to 1,000.
+ + `"vectorSearch.algorithms.efSearch` is the number of nearest neighbors used during search. Default is 500. The range is 100 to 1,000.
+ `vectorSearch.algorithms.metric` should be "cosine" if you're using Azure OpenAI, otherwise use the similarity metric associated with the embedding model you're using. Supported values are `cosine`, `dotProduct`, `euclidean`, `hamming` (used for [indexing binary data](vector-search-how-to-index-binary-data.md)).
@@ -294,25 +291,30 @@ Once you have a vector configuration, you can add a vector field to the fields c
Vector fields are characterized by [their data type](/rest/api/searchservice/supported-data-types#edm-data-types-for-vector-fields), a `dimensions` property based on the embedding model used to output the vectors, and a vector profile that you created in a previous step.
-```json
-{
- "name": "contentVector",
- "type": "Collection(Edm.Single)",
- "searchable": true,
- "retrievable": false,
- "stored": false,
- "dimensions": 1536,
- "vectorSearchProfile": "vector-profile-1"
-}
-```
-
### [**2024-07-01**](#tab/rest-2024-07-01)
[**2024-07-01**](/rest/api/searchservice/search-service-api-versions#2024-07-01) is generally available.
-1. Use the [Create or Update Index](/rest/api/searchservice/indexes/create-or-update) to create the index.
+1. Use the [Create or Update Index](/rest/api/searchservice/indexes/create-or-update) to create the index and add a vector field to the fields collection.
-1. Define a vector field with the following attributes. You can store one generated embedding per field. For each vector field:
+ ```json
+ {
+ "name": "example-index",
+ "fields": [
+ {
+ "name": "contentVector",
+ "type": "Collection(Edm.Single)",
+ "searchable": true,
+ "retrievable": false,
+ "stored": false,
+ "dimensions": 1536,
+ "vectorSearchProfile": "vector-profile-1"
+ }
+ ]
+ }
+ ```
+
+1. Specify a vector field with the following attributes. You can store one generated embedding per field. For each vector field:
+ `type` must be a [vector data types](/rest/api/searchservice/supported-data-types#edm-data-types-for-vector-fields). `Collection(Edm.Single)` is the most common for embedding models.
+ `dimensions` is the number of dimensions generated by the embedding model. For text-embedding-ada-002, it's fixed at 1536. For the text-embedding-3 model series, there's a range of values. If you're using integrated vectorization and an embedding skill to generate vectors, make sure this property is set to the [same dimensions value](cognitive-search-skill-azure-openai-embedding.md#supported-dimensions-by-modelname) used by the embedding skill.
@@ -401,14 +403,28 @@ Vector fields are characterized by [their data type](/rest/api/searchservice/sup
### [**2024-05-01-preview**](#tab/rest-2024-05-01-Preview)
-[**2024-05-01-preview**](/rest/api/searchservice/search-service-api-versions#2024-05-01-preview) is the most recent preview version.
+[**2024-05-01-preview**](/rest/api/searchservice/search-service-api-versions#2024-05-01-preview) is the most recent preview version. It supports the same vector field definitions as the stable version, including support for all [vector data types](/rest/api/searchservice/supported-data-types#edm-data-types-for-vector-fields).
-+ Supports all [vector data types](/rest/api/searchservice/supported-data-types#edm-data-types-for-vector-fields).
-+ Inclusive of `2024-03-01-preview`.
+1. Use the [Create or Update Index Preview REST API](/rest/api/searchservice/indexes/create-or-update?view=rest-searchservice-2024-05-01-preview&preserve-view=true) to create the index and add a vector field to the fields collection.
-1. Use the [Create or Update Index Preview REST API](/rest/api/searchservice/indexes/create-or-update?view=rest-searchservice-2024-05-01-preview&preserve-view=true) to define the fields collection of an index.
+ ```json
+ {
+ "name": "example-index",
+ "fields": [
+ {
+ "name": "contentVector",
+ "type": "Collection(Edm.Single)",
+ "searchable": true,
+ "retrievable": false,
+ "stored": false,
+ "dimensions": 1536,
+ "vectorSearchProfile": "vector-profile-1"
+ }
+ ]
+ }
+ ```
-1. Add vector fields to the fields collection. You can store one generated embedding per document field. For each vector field:
+1. Specify a vector field with the following attributes. You can store one generated embedding per document field. For each vector field:
+ `type` can be `Collection(Edm.Single)`, `Collection(Edm.Half)`, `Collection(Edm.Int16)`, `Collection(Edm.SByte)`
+ `dimensions` is the number of dimensions generated by the embedding model. For text-embedding-ada-002, it's 1536.
@@ -649,6 +665,6 @@ Key points include:
As a next step, we recommend [Query vector data in a search index](vector-search-how-to-query.md).
-Code samples in the [azure-search-vector](https://github.com/Azure/azure-search-vector-samples) repository demonstrate end-to-end workflows that include schema definition, vectorization, indexing, and queries.
+Code samples in the [azure-search-vector-samples](https://github.com/Azure/azure-search-vector-samples) repository demonstrate end-to-end workflows that include schema definition, vectorization, indexing, and queries.
There's demo code for [Python](https://github.com/Azure/azure-search-vector-samples/tree/main/demo-python), [C#](https://github.com/Azure/azure-search-vector-samples/tree/main/demo-dotnet), and [JavaScript](https://github.com/Azure/azure-search-vector-samples/tree/main/demo-javascript).{
"modification_type": "minor update",
"modification_title": "ベクトル検索インデックス作成に関するドキュメントの更新"
}この修正は、Azure AI Searchにおけるベクトル検索インデックスの作成に関するドキュメントの更新です。主な変更点は、インデックスの構成やマッピングに関する説明を明確化し、新たに追加された情報や修正された表現を含んでいます。
具体的には、インデックススキーマの定義についてより詳細な説明が加えられ、ベクトルフィールドと非ベクトルフィールドの区分が明示化されました。さらに、ベクトルアルゴリズムや圧縮オプションに関する情報も拡充され、特にHNSWアルゴリズムやk-Nearest Neighbor（KNN）アルゴリズムに関する使用方法について詳述されています。
また、ドキュメントの日付が2025年2月14日から2025年4月17日に更新されたことや、文中の項目における表現の改善も行われています。全体として、ユーザーがベクトルインデックスを作成し、データをインデックス化するための手順が、より理解しやすくなっています。この変更は、Azure AI Searchを利用する開発者やユーザーにとって、実装時の利便性向上を目的としています。
@@ -28,6 +28,9 @@ To use built-in quantization, follow these steps:
> - Load the index with float32 or float16 data that's quantized during indexing with the configuration you defined
> - Optionally, [query quantized data](#query-a-quantized-vector-field-using-oversampling) using the oversampling parameter. If the vector field doesn't specify oversampling in its definition, you can add it at query time.
+> [!TIP]
+> [Azure AI Search: Cut Vector Costs Up To 92.5% with New Compression Techniques](https://aka.ms/AISearch-cut-cost) compares compression strategies and explains savings in storage and costs. It also includes metrics for measuring relevance based on Normalized discounted cumulative gain (NDCG), demonstrating that you can compress your data without sacrificing search quality.
+
## Prerequisites
- [Vector fields in a search index](vector-search-how-to-create-index.md), with a `vectorSearch` configuration specifying either the Hierarchical Navigable Small Worlds (HNSW) or exhaustive K-nearest neighbor (eKNN) algorithm, and a new vector profile.{
"modification_type": "minor update",
"modification_title": "ベクトル検索における量子化に関するドキュメントの更新"
}この修正は、Azure AI Searchにおけるベクトル検索の量子化に関するドキュメントの更新です。新たに追加された情報として、コスト削減に関する重要なタイップが含まれています。「Azure AI Search: Cut Vector Costs Up To 92.5% with New Compression Techniques」というリンクが追加され、圧縮戦略の比較や、保存コストの節約に関する説明が含まれています。この情報は、検索の品質を損なうことなくデータを圧縮できることを示す、Normalized Discounted Cumulative Gain (NDCG) の基準も含んでいます。
さらに、ドキュメントは量子化を利用するための手順や、必要な前提条件に関するセクションを維持したまま、詳細な手続きが明記されています。これにより、ユーザーは量子化のプロセスを理解しやすくなり、関連するリソースへのアクセスがさらに容易になります。全体として、ユーザーにとって有用な情報が充実し、Azure AI Searchを利用する際の効率が向上することを目的とした変更です。