自动客服演示
架构、部署与知识问答培训手册

az group create -n rg-lip777-foundry-iq -l eastus2 --tags issue=LIP-777 purpose=oppo-demo -o json

操作成功。资源配置详情见本节及附带审计记录。

az search service create -g rg-lip777-foundry-iq -n srch-lip777-f697e914 --sku basic -l eastus2 --partition-count 1 --replica-count 1 --identity-type SystemAssigned --disable-local-auth true --semantic-search free --tags issue=LIP-777 purpose=oppo-demo -o json

ERROR: (InsufficientResourcesAvailable) The region 'eastus2' is currently out of the resources required to provision new services. Try creating the service in another region. RequestId: f6cb8bd5-b34a-4b25-b038-c4a7aa2b5b0d Code: InsufficientResourcesAvailable Message: The region 'eastus2' is currently out of the resources required to provision new services. Try creating the service in another region. RequestId: f6cb8bd5-b34a-4b25-b038-c4a7aa2b5b0d

az role assignment create --assignee-object-id 9c537c3c-9595-496c-8a2f-a16f0cb9ce79 --assignee-principal-type ServicePrincipal --role 'Cognitive Services OpenAI User' --scope /subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/peng-foundry/providers/Microsoft.CognitiveServices/accounts/peng-azfoundry-eastus2 -o json

操作成功。资源配置详情见本节及附带审计记录。

az cognitiveservices account deployment create -g peng-foundry -n peng-azfoundry-eastus2 --deployment-name lip777-gpt-5-mini --model-name gpt-5-mini --model-version 2025-08-07 --model-format OpenAI --sku-name GlobalStandard --sku-capacity 1 -o json

操作成功。资源配置详情见本节及附带审计记录。

az cognitiveservices account deployment create -g peng-foundry -n peng-azfoundry-eastus2 --deployment-name lip777-gpt-5-mini --model-name gpt-5-mini --model-version 2025-08-07 --model-format OpenAI --sku-name GlobalStandard --sku-capacity 30 -o json

操作成功。资源配置详情见本节及附带审计记录。

az role assignment create --assignee-object-id fc4cdfba-6de8-4707-bd66-c505d13f4f07 --assignee-principal-type ServicePrincipal --role 'Cognitive Services User' --scope /subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/peng-foundry/providers/Microsoft.CognitiveServices/accounts/peng-azfoundry-eastus2 -o json

操作成功。资源配置详情见本节及附带审计记录。

az role assignment create --assignee-object-id 9c537c3c-9595-496c-8a2f-a16f0cb9ce79 --assignee-principal-type ServicePrincipal --role 'Search Index Data Reader' --scope /subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/rg-lip777-foundry-iq/providers/Microsoft.Search/searchServices/srch-lip777-f697e914 -o json

操作成功。资源配置详情见本节及附带审计记录。

az role assignment create --assignee-object-id 707b10f8-b414-4c3a-b90b-c2665198b716 --assignee-principal-type User --role 'Search Service Contributor' --scope /subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/rg-lip777-foundry-iq/providers/Microsoft.Search/searchServices/srch-lip777-f697e914 -o json

操作成功。资源配置详情见本节及附带审计记录。

az role assignment create --assignee-object-id 707b10f8-b414-4c3a-b90b-c2665198b716 --assignee-principal-type User --role 'Search Index Data Contributor' --scope /subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/rg-lip777-foundry-iq/providers/Microsoft.Search/searchServices/srch-lip777-f697e914 -o json

操作成功。资源配置详情见本节及附带审计记录。

az rest --method patch --url 'https://management.azure.com/subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/rg-lip777-foundry-iq/providers/Microsoft.Search/searchServices/srch-lip777-f697e914?api-version=2026-03-01-preview' --body '{"properties":{"knowledgeRetrieval":"standard"}}' -o json

操作成功。资源配置详情见本节及附带审计记录。

az search service create -g rg-lip777-foundry-iq -n srch-lip777-f697e914 --sku basic -l eastus --partition-count 1 --replica-count 1 --identity-type SystemAssigned --disable-local-auth true --semantic-search free --tags issue=LIP-777 purpose=oppo-demo -o json

操作成功。资源配置详情见本节及附带审计记录。

az search service show -g rg-lip777-foundry-iq -n srch-lip777-f697e914 -o json

操作成功。资源配置详情见本节及附带审计记录。

az rest --method get --url 'https://management.azure.com/subscriptions/f697e914-f091-4f00-a091-a77ae4c0a6c1/resourceGroups/rg-lip777-foundry-iq/providers/Microsoft.Search/searchServices/srch-lip777-f697e914?api-version=2026-03-01-preview' --query '{status:properties.status,knowledgeRetrieval:properties.knowledgeRetrieval,semanticSearch:properties.semanticSearch,disableLocalAuth:properties.disableLocalAuth,replicas:properties.replicaCount,partitions:properties.partitionCount,sku:sku.name}' -o json

操作成功。资源配置详情见本节及附带审计记录。

{
  "name": "microsoft-faq",
  "fields": [
    {
      "name": "id",
      "type": "Edm.String",
      "key": true,
      "filterable": true
    },
    {
      "name": "title",
      "type": "Edm.String",
      "searchable": true,
      "retrievable": true
    },
    {
      "name": "content",
      "type": "Edm.String",
      "searchable": true,
      "retrievable": true
    },
    {
      "name": "url",
      "type": "Edm.String",
      "retrievable": true
    },
    {
      "name": "product",
      "type": "Edm.String",
      "filterable": true,
      "retrievable": true
    }
  ],
  "semantic": {
    "defaultConfiguration": "faq-semantic",
    "configurations": [
      {
        "name": "faq-semantic",
        "prioritizedFields": {
          "titleField": {
            "fieldName": "title"
          },
          "prioritizedContentFields": [
            {
              "fieldName": "content"
            }
          ]
        }
      }
    ]
  }
}

{
  "name": "microsoft-faq-source",
  "kind": "searchIndex",
  "description": "Official Microsoft FAQs and Foundry IQ documentation. Public demo corpus only.",
  "searchIndexParameters": {
    "searchIndexName": "microsoft-faq",
    "semanticConfigurationName": "faq-semantic",
    "sourceDataFields": [
      {
        "name": "id"
      },
      {
        "name": "title"
      },
      {
        "name": "content"
      },
      {
        "name": "url"
      }
    ]
  }
}

{
  "name": "oppo-support-demo",
  "description": "OPPO India customer service demonstration using Microsoft official FAQs; contains no OPPO product policies.",
  "knowledgeSources": [
    {
      "name": "microsoft-faq-source"
    }
  ],
  "models": [
    {
      "kind": "azureOpenAI",
      "azureOpenAIParameters": {
        "resourceUri": "https://peng-azfoundry-eastus2.openai.azure.com",
        "deploymentId": "lip777-gpt-5-mini",
        "modelName": "gpt-5-mini"
      }
    }
  ],
  "retrievalInstructions": "Translate questions to English search queries as needed. Search only the Microsoft official documentation. For questions involving multiple topics, retrieve evidence for each topic. The corpus contains no OPPO product warranty or repair policies.",
  "outputMode": "extractiveData",
  "retrievalReasoningEffort": {
    "kind": "low"
  }
}

Microsoft does not use customers’ data to train its foundation models without permission. For Azure Foundry / Models sold by Azure (including Azure OpenAI), prompts, completions, embeddings, and training data: are not available to other customers or to OpenAI/third-party providers; are not used to improve those providers’ models or services; and are not used to train generative AI foundation models without your explicit permission or instruction. Azure OpenAI likewise doesn’t use customer data to retrain models. (Some service features do store customer data for the service to operate—stored data is kept in the customer’s tenant geography, encrypted at rest, and can be deleted by the customer; stored data can be used for customer-requested fine-tuning or for abuse monitoring as described.)

[0] Foundry Agent Service FAQ — Does Microsoft use my data for training models?

[1] Foundry Models data privacy — Foundry Models data privacy

[2] Microsoft Foundry classic FAQ — Do you use my company data to train any of the models?

[3] Foundry Models data privacy — Data storage for Models sold by Azure features

Basic setup stores agent state in Microsoft-managed resources; standard setup stores agent data (threads, files, and vector stores) in your Azure resources that you declare through capability settings. In basic setup data is kept in a secure, Microsoft-managed storage account that's logically separated. In standard setup you store data in your own Azure resources: Azure Storage for files and attachments, Azure Cosmos DB for threads and conversation history, and Azure AI Search for vector stores. Foundry Agent Service endpoints are regional and data is stored in the same region as the endpoint. [0][2][4]

[0] Foundry Agent Service FAQ — What's the difference between basic setup and standard setup?

[2] Foundry Agent Service FAQ — Where is this data stored?

[4] Foundry Agent Service FAQ — Where is data stored geographically?

根据微软文档：
- Foundry Agent Service 会保存两类数据：有状态实体（例如对话和响应）以及上传的文件和向量存储（用于设置或生成响应）[4]。
- 此类数据会一直保留，除非你显式删除。你可以通过 API 或 SDK 删除线程、文件或向量存储来删除代理数据[1]。
- 微软/Foundry 不会将你的数据用于训练通用基础模型。客户的提示、生成内容、嵌入和用于微调的训练数据：不会被其他客户访问，也不会提供给模型供应商用于改进其模型；在未经你明确允许或指示的情况下，不会用于训练任何生成式基础模型。用于微调的模型和数据仅供该客户使用，且可被客户删除[3][6][7]。

请参阅相关文档以了解加密、地理位置和可选的客户管理密钥等详细存储条件[6]。

[1] Foundry Agent Service FAQ — How long is this data stored?

[3] Foundry Models data privacy — Foundry Models data privacy

[6] Foundry Models data privacy — Data storage for Models sold by Azure features

[7] Foundry Models data privacy — What data does Foundry process to provide Models sold by Azure?

[4] Foundry Agent Service FAQ — Do you store any data used in the Foundry Agent Service API?

Microsoft के दस्तावेज़ों के अनुसार, Models sold by Azure (जैसे Foundry/Azure OpenAI सेवाएँ) ग्राहक द्वारा दी गई इनपुट (प्रॉम्प्ट), आउटपुट (कम्प्लीशन्स), एम्बेडिंग्स और ट्रेनिंग डेटा को अन्य ग्राहकों या मॉडल प्रदाताओं (जैसे OpenAI) के साथ साझा नहीं करता, और इन्हें बिना आपके स्पष्ट अनुमति या निर्देश के जेनरेटिव फाउंडेशन मॉडलों को प्रशिक्षित करने के लिए उपयोग नहीं किया जाता। इसके अलावा Microsoft कहता है कि वह ग्राहक डेटा का उपयोग मॉनिटरिंग और दुरुपयोग-निवारण के लिए कर सकता है; दुरुपयोग मॉनिटरिंग के दौरान नमूनों की स्वचालित और—आवश्यक होने पर—मानव समीक्षा की जा सकती है (यह समीक्षा सीमित परिस्थितियों में और नियंत्रित पहुँच के साथ होती है)। (यदि ग्राहक ने फ़ाइन-ट्यूनिंग के लिए डेटा अपलोड किया तो वही फाइन-ट्यून किए गए मॉडल केवल उस ग्राहक के लिए उपलब्ध रहते हैं और बिना अनुमति फाउंडेशन मॉडल के प्रशिक्षण में उपयोग नहीं होते.)

[0] Foundry Models data privacy — Foundry Models data privacy

[4] Foundry Agent Service FAQ — Does Microsoft use my data for training models?

[9] Foundry Models data privacy — Data storage for Models sold by Azure features

[1] Foundry Models data privacy — Preventing abuse

The Microsoft knowledge base does not provide sufficient evidence to answer this question. Please refer it to a human support agent; no support ticket has been created.

Foundry IQ uses an agentic retrieval pipeline to get permission-aware enterprise knowledge and return answers with citations. You create a knowledge base that references knowledge sources (indexed or remote) such as Azure Blob, SharePoint, OneLake, or the web. When an agent queries the knowledge base, Foundry IQ: (1) uses agentic retrieval to decompose complex queries into subqueries, run parallel searches (keyword, vector, or hybrid) across the configured sources, and semantically rerank and aggregate results; (2) can run extractive retrieval so agents can reason over raw content and trace answers to source documents; and (3) enforces end-to-end permissions by synchronizing and honoring source ACLs and running queries under the caller’s Microsoft Entra identity so only authorized content is returned. You can connect a Foundry IQ knowledge base to a Copilot Studio (GitHub Copilot harness) agent to get grounded answers with source citations.

[0] What is Foundry IQ? — What is Foundry IQ?

[1] What is Foundry IQ? — Connect your Foundry IQ (Azure AI Search) knowledge base to a Microsoft Copilot Studio agent

[2] What is Foundry IQ? — Capabilities

[3] What is Foundry IQ? — Workflow

[4] What is Foundry IQ? — Components

[7] What is Foundry IQ? — Capabilities and security