このページでは、ストリーミング回答メソッドについて説明します。
ストリーミング回答メソッドには、回答メソッドと同じ機能の多くに加えて、ストリーミングという機能が 1 つあります。回答をストリーミングすると、生成された回答は複数の部分に分割され、順番に送信されます。
回答のストリーミングは、生成された回答が長く、回答全体を一度に送信すると遅延が発生する場合に特に便利です。回答をストリーミングすると、レイテンシが軽減されます。
制限事項
ストリーミング回答メソッドには、回答メソッドと同じ機能がありますが、次の例外があります。
ストリーミング回答は英語でのみサポートされています。
言い換えステップの数は 1 です。言い換えを無効にしたり、ステップの最大数を変更したりすることはできません。
ストリーミング回答のグラウンディング スコアは取得できません。また、十分に根拠のある回答のみを返すこともできません。
ストリーミング回答メソッドで使用できるのは Gemini モデルのみです。モデルのリストについては、使用可能なモデルをご覧ください。
回答をストリーミングする
次のコマンドは、ストリーミング回答メソッドを呼び出して、生成された回答を一連の JSON レスポンスの形式で返す方法を示しています。通常、各レスポンスには回答の 1 つの文が含まれます。
この基本コマンドには、必要な入力のみが表示されます。オプションはデフォルトのままにします。
他のオプションの例については、回答とフォローアップを取得するをご覧ください。一部の回答オプションは回答ストリーミングでは使用できません。このページの制限事項をご覧ください。
REST
生成された回答をストリーミングして検索結果を取得する手順は次のとおりです。
次の curl コマンドを実行します。
curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" \ -H "Content-Type: application/json" \ "https://discoveryengine.googleapis.com/v1/projects/PROJECT_ID/locations/global/collections/default_collection/engines/APP_ID/servingConfigs/default_search:streamAnswer" \ -d '{ "query": { "text": "QUERY"} }'以下を置き換えます。
PROJECT_ID: Google Cloud プロジェクトの ID。APP_ID: クエリする Vertex AI Search アプリの ID。QUERY: 質問または検索クエリを含む自由形式の文字列。たとえば、「bigquery と spanner のどちらが速いですか？」などです。
コマンドの例と部分的な結果
curl -X POST -H "Authorization: Bearer $(gcloud auth print-access-token)" \ -H "Content-Type: application/json" \ "https://discoveryengine.googleapis.com/v1/projects/my-project-123/locations/global/collections/default_collection/engines/my-app/servingConfigs/default_search:streamAnswer" \ -d '{ "query":{"text":"Which database is faster, bigquery or spanner?"} }'
[{ "answer": { "state": "STREAMING", "steps": [ { "description": "Rephrase the query and search.", "actions": [ { "searchAction": { "query": " What is the performance of Spanner?" }, "observation": { "searchResults": [ { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "QUERY : a free-text string that contains the question or search query. For example, "Compare the BigQuery and \u003cb\u003eSpanner\u003c/b\u003e databases?". Example command and result.", "snippetStatus": "SUCCESS" } ] }, ... { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/b95bb201a0adb24f769627f56cf34405", "uri": "https://abc.xyz/assets/investor/static/pdf/2017_Q1_Earnings_Transcript.pdf", "title": "\u200b \u200b", "snippetInfo": [ { "snippet": "well as Hardware related costs, reflecting the continued strong \u003cb\u003eperformance\u003c/b\u003e of our new Made by ... We introduced dozens of new products, including \u003cb\u003eSpanner\u003c/b\u003e, a ...", "snippetStatus": "SUCCESS" } ] } ] } }, { "searchAction": { "query": " What is the performance of BigQuery?" }, "observation": { "searchResults": [ { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/18bcc727bfd6a3d1be0aa4bd49fe2c50", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/evaluate-search-quality", "title": "Evaluate search quality | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "You can evaluate the \u003cb\u003eperformance\u003c/b\u003e of generic search apps that contain structured, unstructured, and website data. ... Import from \u003cb\u003eBigQuery\u003c/b\u003e: import \u003cb\u003eBigQuery\u003c/b\u003e data ...", "snippetStatus": "SUCCESS" } ] }, ... { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "QUERY : a free-text string that contains the question or search query. For example, "Compare the \u003cb\u003eBigQuery\u003c/b\u003e and Spanner databases?". Example command and result.", "snippetStatus": "SUCCESS" } ] } ] } } ] } ] } } , { "answer": { "state": "STREAMING", "references": [ { "chunkInfo": { "content": "Example command and partial result curl -X POST -H \"Authorization: Bearer $(gcloud auth print-access-token)\" -H \"Content-Type: application/json\" \"https://discoveryengine.googleapis.com/v1/projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/servingConfigs/default_search:answer\" -d '{ \"query\": { \"text\": \"Compare bigquery with spanner database?\"} \"queryUnderstandingSpec\": { \"queryRephraserSpec\": { \"disable\": true } } }' { \"answer\": { \"state\": \"SUCCEEDED\", \"answerText\": \"You can compare BigQuery and Spanner databases using the following criteria:\\n\\n* **Pricing:** BigQuery is priced per GB of data processed, while Spanner is priced per hour of compute time.\\n* **Performance:** BigQuery is designed for fast analytics, while Spanner is designed for high availability and scalability.\\n* **Features:** BigQuery supports a wide range of features, including SQL, machine learning, and streaming. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud" } } }, ... { "chunkInfo": { "content": "Here is an example of a summary, with citations and citation metadata, returned at the end of a search response: See more code actions. Dismiss View Light code theme Dark code theme \"summary\": { \"summaryText\": \"BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse [1]. BigQuery supports all data types, works across clouds, and has built-in machine learning and business intelligence, all within a unified platform [2, 3].\", \"summaryWithMetadata\": { \"summary\": \"BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/f7ba2e8666f5514b5bc14f5e300d7678", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/get-search-summaries", "title": "Get search summaries | Vertex AI Agent Builder | Google Cloud" } } } ] } } , { "answer": { "state": "STREAMING", "answerText": "Span" } } , { "answer": { "state": "STREAMING", "answerText": "ner is Google's large-scale database that scales 20 times better than" } } , ... { "answer": { "state": "STREAMING", "answerText": " Web Services, and on-premises data sources. " } } , { "answer": { "state": "STREAMING", "answerText": "Spanner is a distributed, strongly consistent, SQL database designed to scale to 10 million servers. \n" } } , { "answer": { "state": "SUCCEEDED", "answerText": "Spanner is Google's large-scale database that scales 20 times better than any competitor. Spanner is designed for high availability and scalability, while BigQuery is designed for fast analytics. BigQuery is a serverless, highly scalable, and cost-effective cloud data warehouse that enables businesses to analyze all their data very quickly. BigQuery is a very powerful tool that can be used to analyze data from many different sources, including Google Cloud Platform, Amazon Web Services, and on-premises data sources. Spanner is a distributed, strongly consistent, SQL database designed to scale to 10 million servers. \n", "references": [ { "chunkInfo": { "content": "Example command and partial result curl -X POST -H \"Authorization: Bearer $(gcloud auth print-access-token)\" -H \"Content-Type: application/json\" \"https://discoveryengine.googleapis.com/v1/projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/servingConfigs/default_search:answer\" -d '{ \"query\": { \"text\": \"Compare bigquery with spanner database?\"} \"queryUnderstandingSpec\": { \"queryRephraserSpec\": { \"disable\": true } } }' { \"answer\": { \"state\": \"SUCCEEDED\", \"answerText\": \"You can compare BigQuery and Spanner databases using the following criteria:\\n\\n* **Pricing:** BigQuery is priced per GB of data processed, while Spanner is priced per hour of compute time.\\n* **Performance:** BigQuery is designed for fast analytics, while Spanner is designed for high availability and scalability.\\n* **Features:** BigQuery supports a wide range of features, including SQL, machine learning, and streaming. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud" } } }, { "chunkInfo": { "content": "Second, we also give them the ability to build applications using a set of technology that can run on any environment that they have. When we say on any environment - at their premise, on our cloud or on any other cloud. So, in other words, they can learn once, write once, deploy anywhere; and we make money no matter where they deploy. An example of that is a recent product we introduced called AlloyDB. It's the fastest-performing relational database in the market. We run it in all four environments: Our cloud, on-premise and on other clouds. And it's the only relational database that can run in any of those configurations. You see that in our adoption, both at the top end of the market where a system - for example, like Spanner, which is our large-scale database - scales 20 times better than the largest scalable system of any competitor. So for high-end, we work extremely well. And, also, because we made it so easy to use, startups and small businesses are growing very quickly in their adoption of our platform. When we introduced our AI systems, we introduced a platform called Vertex AI. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/8ad08d0844d601733e135381512e2a16", "uri": "http://abc.xyz/thomas-kurian-ceo-google-cloud-at-the-goldman-sachs-2023-communacopia-technology-conference-on-september-7th-2023", "title": "Thomas Kurian, CEO, Google Cloud at the Goldman Sachs 2023 Communacopia + Technology Conference on September 7th, 2023 - Alphabet Investor Relations" } } }, ... { "chunkInfo": { "content": "BigQuery is also integrated with other Google Cloud services, such as Google Kubernetes Engine, Cloud Data Fusion, and Cloud Dataproc, making it easy to build and deploy data pipelines. Here are some of the benefits of using BigQuery: * **Fast and scalable:** BigQuery can process petabytes of data very quickly, and it can scale to handle even the most demanding workloads. * **Cost-effective:** BigQuery is a very cost-effective way to store and analyze data. You only pay for the data that you use, and there are no upfront costs or commitments. * **Secure:** BigQuery is a secure platform that meets the needs of even the most security-conscious organizations. * **Easy to use:** BigQuery is easy to use, even for non-technical users. It has a simple and intuitive user interface, and it supports a variety of data sources. * **Integrated with other Google Cloud services:** BigQuery is integrated with other Google Cloud services, making it easy to build and deploy data pipelines. If you are looking for a fast, scalable, and cost-effective way to analyze your data, then BigQuery is a great option. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/f7ba2e8666f5514b5bc14f5e300d7678", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/get-search-summaries", "title": "Get search summaries | Vertex AI Agent Builder | Google Cloud" } } }, { "chunkInfo": { "content": "Here is an example of a summary, with citations and citation metadata, returned at the end of a search response: See more code actions. Dismiss View Light code theme Dark code theme \"summary\": { \"summaryText\": \"BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse [1]. BigQuery supports all data types, works across clouds, and has built-in machine learning and business intelligence, all within a unified platform [2, 3].\", \"summaryWithMetadata\": { \"summary\": \"BigQuery is Google Cloud's fully managed and completely serverless enterprise data warehouse. ", "documentMetadata": { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/f7ba2e8666f5514b5bc14f5e300d7678", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/get-search-summaries", "title": "Get search summaries | Vertex AI Agent Builder | Google Cloud" } } } ], "steps": [ { "description": "Rephrase the query and search.", "actions": [ { "searchAction": { "query": " What is the performance of Spanner?" }, "observation": { "searchResults": [ { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "QUERY : a free-text string that contains the question or search query. For example, "Compare the BigQuery and \u003cb\u003eSpanner\u003c/b\u003e databases?". Example command and result.", "snippetStatus": "SUCCESS" } ] }, { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/9d022f7bdf24bac6714a9cf61a5458c7", "uri": "https://abc.xyz/assets/87/4c/162ca71d4178a3f4d39002467439/thomas-kurian-goldman-sachs-090723.pdf", "title": "Thomas Kurian Goldman Sachs 090723", "snippetInfo": [ { "snippet": "2X better training \u003cb\u003eperformance\u003c/b\u003e per dollar1 compared to a leading cloud alternative. More than 70% of gen AI unicorns are Google Cloud customers. Best ...", "snippetStatus": "SUCCESS" } ] }, ... { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/20641e370fa86c78f1c81f3dab22efe1", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/release-notes", "title": "Vertex AI Agent Builder release notes | Google Cloud", "snippetInfo": [ { "snippet": "Generative answers have been updated with \u003cb\u003eperformance\u003c/b\u003e improvements. ... This lets you assess your search engine's \u003cb\u003eperformance\u003c/b\u003e ... Importing data from \u003cb\u003eSpanner\u003c/b\u003e, Cloud ...", "snippetStatus": "SUCCESS" } ] }, { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/b95bb201a0adb24f769627f56cf34405", "uri": "https://abc.xyz/assets/investor/static/pdf/2017_Q1_Earnings_Transcript.pdf", "title": "\u200b \u200b", "snippetInfo": [ { "snippet": "well as Hardware related costs, reflecting the continued strong \u003cb\u003eperformance\u003c/b\u003e of our new Made by ... We introduced dozens of new products, including \u003cb\u003eSpanner\u003c/b\u003e, a ...", "snippetStatus": "SUCCESS" } ] } ] } }, { "searchAction": { "query": " What is the performance of BigQuery?" }, "observation": { "searchResults": [ { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/18bcc727bfd6a3d1be0aa4bd49fe2c50", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/evaluate-search-quality", "title": "Evaluate search quality | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "You can evaluate the \u003cb\u003eperformance\u003c/b\u003e of generic search apps that contain structured, unstructured, and website data. ... Import from \u003cb\u003eBigQuery\u003c/b\u003e: import \u003cb\u003eBigQuery\u003c/b\u003e data ...", "snippetStatus": "SUCCESS" } ] }, { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/2a3221d40533a4bdaf35778962a2a079", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/check-media-data-quality", "title": "Check data quality for media recommendations | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "... model that will result in \u003cb\u003eperformance\u003c/b\u003e issue if not met for all media recommendations models and all business objectives.", "condition": { "expression ...", "snippetStatus": "SUCCESS" } ] }, ... { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/18c258b9c770f4d762e6233d1a1bc81c", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/user-events", "title": "About user events | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "This section provides the data formats for each event type supported by media recommendations. Examples for JavaScript Pixel are provided. For \u003cb\u003eBigQuery\u003c/b\u003e, the ...", "snippetStatus": "SUCCESS" } ] }, { "document": "projects/123456/locations/global/collections/default_collection/dataStores/my-data-store/branches/0/documents/1a9f55e00c42c06ca97bf5a5868dbcdc", "uri": "https://cloud.google.com/generative-ai-app-builder/docs/answer", "title": "Get answers and follow-ups | Vertex AI Agent Builder | Google Cloud", "snippetInfo": [ { "snippet": "QUERY : a free-text string that contains the question or search query. For example, "Compare the \u003cb\u003eBigQuery\u003c/b\u003e and Spanner databases?". Example command and result.", "snippetStatus": "SUCCESS" } ] } ] } } ] } ] } }この例では、「bigquery と spanner のどちらが速いですか？」というクエリに対する回答が、一連の JSON 出力に表示されます。最終的な出力にはステータス
SUCCEEDEDが付与され、完全な回答が含まれます。
その他の例
回答をストリーミングするで説明した基本コマンドは、オプションを指定しない場合の最もシンプルなコマンドです。ただし、このページに記載されている制限事項を除き、answer メソッドで使用できるオプションと同じオプションを適用できます。
回答のストリーミングは、フォローアップ セッションでも使用できます。