AI-103 Practice Questions

The correct built-in skills are Azure OpenAI Embedding and Text Split. The case study requires an indexing pipeline that enables semantic and vector search over the product sheets stored in Azure Blob Storage, so Agent1 can retrieve relevant product information for natural language customer questions. For a RAG pipeline, long PDF content must first be broken into retrievable chunks, and each chunk must then be vectorized for semantic similarity retrieval.

Microsoft’s Azure AI Search integrated vectorization guidance states that you create a skillset that calls the Text Split skill for chunking and the Azure OpenAI Embedding skill to vectorize the chunks. The Text Split skill breaks text into chunks and provides positional metadata, making it suitable when downstream embedding skills have input-length limits. The Azure OpenAI Embedding skill connects to an embedding model deployed in Azure OpenAI or a Microsoft Foundry project and generates embeddings during indexing.

Language Detection, Entity Recognition, and key phrase extraction can enrich text, but they do not create vector embeddings. Merge is useful for combining OCR text with document text, but it does not satisfy the core vector-search requirement. Reference topics: Azure AI Search skillsets, Text Split skill, Azure OpenAI Embedding skill, integrated vectorization, and RAG indexing.

Because the existing classifier is an Azure AI Custom Vision project, the lowest-effort workflow is performed directly in the Custom Vision portal , not Azure Machine Learning studio or Vision Studio. Open the existing project so that the new product classes can be incorporated into its current training dataset.

Next, upload representative sample images of each new product. The images should include sufficient variation in viewing angle, lighting, background, size, and presentation to support reliable classification. Label each image with the appropriate new product tag. Custom Vision uses these tags as the target classes during supervised training. Microsoft’s portal workflow explicitly supports uploading and manually tagging images without requiring application code.

Retrain the classifier after the labeled images have been added. Training creates a new iteration based on all current training images and tags, including the newly introduced product categories.

Finally, publish the trained iteration to the prediction resource. Publishing makes that iteration available through the Custom Vision Prediction API used by the application.

Study Guide alignment: create, train, evaluate, retrain, publish, and consume custom image-classification models .

The schema requires that a receipt has a discount_code value — No

The schema can be used as is for Azure OpenAI structured outputs — No

The generated JSON object will preserve the property order merchant, order_number, discount_code, line_items, and total — Yes

The discount_code definition permits a nullable result. Consequently, the output can contain " discount_code " : null when the source receipt contains no discount code. The property can be structurally present without containing a substantive discount-code value; therefore, the first statement is No .

The schema cannot be submitted unchanged. Azure OpenAI structured outputs support only a defined subset of JSON Schema. Every property declared in an object must be included in that object’s required array. Optional information must instead be represented as a required property whose type includes null. Each object must also specify " additionalProperties " : false. Furthermore, although nested anyOf constructs are supported, an anyOf construct cannot be used as the root schema. A schema that violates any of these restrictions must be revised before use, making the second statement No .

The third statement is Yes . Azure OpenAI structured outputs preserve keys in the order in which the properties are defined in the supplied schema. Therefore, the receipt object will emit merchant, order_number, discount_code, line_items, and total in that sequence. This behavior supports predictable parsing and downstream integration. The relevant curriculum area is Use Azure OpenAI in Foundry Models to generate content , including application integration and structured model responses.

Use Microsoft Entra ID authentication and pass an OAuth 2.0 bearer token in the REST request’s Authorization header. Assign the required inference role, such as Cognitive Services User , only to each application’s managed identity or service principal. This authenticates the applications as distinct security principals and supports resource-scoped RBAC, auditing, token expiration, and immediate access revocation. Microsoft recommends Entra ID for production workloads because API keys are static, resource-wide secrets that cannot identify individual callers or provide granular authorization.

Each application must specify its appropriate deployment name when making an inference request. Foundry deployments assign a unique name and configuration to each deployed model. In the deployment-level REST API, that name appears in the request route:

/openai/deployments/{deployment-name}/chat/completions

The resource endpoint and authentication mechanism can remain common across all three applications, while the deployment name directs each request to the GPT-3.5 deployment optimized for that application’s workload. A deployment type describes capacity or provisioning characteristics and does not select a particular deployed model.

Study Guide alignment: configure authentication, managed identities, RBAC, model deployments, and REST-based model consumption.

AnalyzeTextOptions

request = AnalyzeTextOptions(

text=input_text,

blocklist_names=[blocklist_name],

halt_on_blocklist_hit=True

)

response = client.analyze_text(request)

AnalyzeTextOptions is the request model used by the Azure AI Content Safety Python SDK to submit text for moderation. Its text property contains the user message, while blocklist_names identifies one or more existing custom blocklists that must be applied during analysis. The value must be supplied as a list, so the correct assignment is [blocklist_name]. The request is then passed to ContentSafetyClient.analyze_text(), which returns an AnalyzeTextResult containing any blocklist matches and applicable harm-category analysis.

Setting halt_on_blocklist_hit=True stops additional harm-category processing when a blocklisted term is detected. This is appropriate when a blocklist match alone is sufficient to prevent the message from reaching the language model.

AddOrUpdateTextBlocklistItemsOptions manages terms within a blocklist; it does not analyze input. AnalyzeImageOptions is intended for image moderation, and TextBlocklist represents blocklist metadata rather than a text-analysis request.

Study Guide alignment: configure safety filters, guardrails, blocklists, risk detection, and content moderation for Azure AI applications.