You have a chat app in a Microsoft Foundry project and an Azure AI Search vectorized index.You need to connect to the index to meet the following requirements:▪ Complex questions must retrieve information from multiple chunks.▪ Multi-turn conversations must influence retrieval planning.▪ Retrievals must run in parallel to reduce latency.Which retrieval approach should you use?
Answer(s): B
An agentic Retrieval-Augmented Generation (RAG) architecture is the best fit for your requirements.Traditional, linear RAG pipelines struggle with complex, multi-turn dependencies, whereas an agentic approach natively solves them through iterative reasoning and tool orchestration.Here is how an agentic RAG pattern specifically satisfies each the three technical requirements:1. Complex Questions & Multi-Chunk RetrievalThe Challenge: Complex questions often require aggregating distinct pieces of information scattered across different documents or sections.The Agentic Solution: An agent uses a Reasoning and Acting (ReAct) loop. It evaluates the user's query,breaks it down into sub-questions, and executes multiple distinct search queries. It then synthesizes the information from these diverse chunks before generating a final response.2. Multi-Turn Conversations & Retrieval PlanningThe Challenge: Standard RAG often just passes the latest user message or a simple chat history summary to the search index, which can lose context or misinterpret the user's true intent in long conversations.The Agentic Solution: The agent acts as a query planner. It maintains the state of the conversation and dynamically decides if a retrieval is needed, what specific keywords or vectors to target based on past turns,and how to reformulate the query to bridge context gaps.3. Parallel Retrievals for Low LatencyThe Challenge: Sequential lookups drastically increase time-to-first-token (TTFT), harming the chat user experience.The Agentic Solution: Modern AI agent frameworks allow the LLM to emit multiple tool calls simultaneously.The application layer can intercept these calls and execute parallel asynchronous queries against your AzureAI Search vectorized index, drastically minimizing latency.
https://arxiv.org/html/2501.09136v1
HOTSPOT (Drag and Drop is not supported)You need to recommend a plan to create a customer support agent by using the Microsoft Foundry AgentService. The agent must meet the following requirements:▪ Retain user preferences across multiple conversations.▪ Enable users to provide contextual grounding by directly uploading documents during a chat.Which Foundry capability should you recommend for each requirement? To answer, select the appropriate options in the answer area.Note: Each correct selection is worth one point.Hot Area:
Answer(s): A
Box 1: Agent memory that uses persistent storageRetain user preferences across multiple conversations.To enable a customer support agent built on the Microsoft Foundry Agent Service to retain user preferences across multiple distinct conversations, you must implement the platform's native managed long-term memory system.The solution architecture must include the following specific components and configurations:*-> 1. Dedicated Long-Term Memory StoreAgent-to-Store Mapping: Provision one dedicated memory store attached directly to your customer support agent to establish a clear knowledge boundary.Managed Lifecycle Operations: Rely on the service runtime to automatically manage the asynchronous extraction of preferences, consolidation (merging duplicates and resolving conflicting facts), and hybrid search retrieval.2. User-Scoped Partitioning (Memory Scope)Isolation Vaults: Define a mandatory Memory Scope for every interaction.Unique Identifiers: Bind the scope parameter to a persistent identifier like the customer’s Microsoft Entra ID or a custom system UUID. This ensures user preferences are securely isolated and do not leak into other customers' sessions.3. Dual Memory TypesBox 2: File Search toolTo enable users to directly upload documents during a chat for real-time contextual grounding, you should include the File Search tool.This built-in tool is specifically designed to augment agents with knowledge from uploaded files or proprietary documents. When users upload a document during a chat thread, the Foundry Agent Service File Search Tool automatically handles the extraction, chunking, embedding, and vector searching of that file to ground the agent's responses dynamically.Incorrect:* Azure AI SearchWhile it is a knowledge grounding tool, the Azure AI Search Tool is built to connect an agent to a pre-existing,centralized corporate search index. It does not handle ad-hoc user file uploads directly inside a chat session.* Code InterpreterThe Code Interpreter Tool is used for executing Python code in a sandboxed environment to perform advanced mathematical calculations, data analysis, and chart generation, rather than document vector retrieval.
https://learn.microsoft.com/en-us/azure/foundry-classic/agents/how-to/tools-classic/overview
You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.You need to improve response completeness. The solution must be implemented in the logic of the application code before responses are returned.What should you do?
Correct:* Add a reflection pass before the responses are returned.A reflection pass acts as an internal verification step within your application's logic. It directs the agent to evaluate its own draft response against the retrieved policy documents to ensure no mandatory regulatory clauses were left out. If it detects that critical text is missing, it regenerates or amends the answer before sending it to the user. This directly fulfills the core requirement of improving response completeness.* Add a retry evaluation before the responses are returned.You should implement an evaluation and retry loop in your application code.You must wrap the agent call inside a conditional code loop that evaluates the output against required criteria(such as a checklist of regulatory clauses), and programmatically forces a retry if information is omitted.Incorrect:* Decrease the value of the max_tokens parameter.* Decrease the value of the temperature paramerer.* Enable response streaming.* Increase the value of the presence_penalty parameter* Replace the model with a smaller deployment.* Switch to Retrieval Augmented Generation (RAG).
https://learn.microsoft.com/en-us/azure/foundry/agents/how-to/tools/ai-search
HOTSPOT (Drag and Drop is not supported)You have a Microsoft Foundry project that contains an agent.You use a GitHub Actions workflow for CI/CD.You need to configure the workflow to automatically evaluate the agent when a pull request (PR) is created and prevent branches from merging if the evaluation results do NOT meet the defined thresholds.How should you configure the workflow? To answer, select the appropriate options in the answer area.Note: Each correct selection is worth one point.Hot Area:
Box 1: An Azure Login action that uses OpenID Connect(OIDC).The most appropriate authentication method for this scenario is Microsoft Entra ID authentication using OpenIDConnect (OIDC).Using OIDC with the azure/login GitHub Action eliminates the need to store long-lived credentials (like passwords or client secrets) inside GitHub secrets, offering a highly secure and passwordless integration.Why OIDC Is the Recommended ChoiceShort-Lived Tokens: GitHub Actions requests a temporary, scoped token from Microsoft Entra ID dynamically for each workflow run.No Secret Rotations: Eliminates administrative overhead and safety risks associated with managing long-livedAzure Service Principal secrets.Granular RBAC: Permissions can be tightly scoped by assigning the Azure AI User role to the specific repository/workflow identity at the Microsoft Foundry resource level.Box 2: FailConfiguring the GitHub Actions workflow to exit with a non-zero status code (fail) when thresholds are not met is the standard mechanism to block pull requests.To enforce this restriction, you must pair the workflow configuration with GitHub repository settings.Example code:- name: Enforce Thresholds run: |# Use node, python, or bash to read JSON and validate python -c "import json, sys with open('results.json') as f:data = json.load(f)if data['groundedness'] < 0.85:print('Error: Groundedness score below 0.85 threshold.')sys.exit(1)
https://oneuptime.com/blog/post/2026-02-02-github-actions-pull-requests/view
Note: This section contains one or more sets of questions with the same scenario and problem. Each question presents a unique solution to the problem. You must determine whether the solution meets the stated goals. More than one solution in the set might solve the problem. It is also possible that none of the solutions in the set solve the problem.After you answer a question in this section, you will NOT be able to return. As a result, these questions do not appear on the Review Screen.You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.Users report that some responses omit required regulatory clauses, even when the clauses are present in the retrieved content.You need to improve response completeness.Solution: You increase the value of the max_tokens parameter.Does this meet the goal?
Correct:* You add a reflection pass that regenerates the response if the required clauses are missing.This is Self-Correction Strategy: A reflection pass allows an agent to evaluate its own initial output against specified constraints (e.g., checking for the presence of mandatory regulatory clauses). If the required text is missing, the agent triggers a programmatic self-correction or regeneration loop to include them before final delivery.Incorrect:* You increase the value of the max_tokens parameter.Increasing the max_tokens parameter prevents the response from being cut off mid-sentence due to length constraints. However, it does not force the model's logic to explicitly include missing information that it chose to leave out earlier in the text.* You increase the value of the temperature parameter.Raising the temperature parameter increases randomness and creativity. For rigid compliance tasks like summarizing regulatory documents, higher temperature actually increases the risk of hallucination and omission.* You run an evaluation flow that scores responses for completeness and blocks responses that fall below a defined threshold.Evaluation Flow Block: Running an evaluation flow to score completeness and blocking bad responses identifies and stops low-quality outputs, but it does not fix or actively improve the response completeness. It simply filters failures out of the system.
https://pub.towardsai.net/reflection-with-llm-how-to-make-ai-review-its-own-work-2db122fca1d8
Note: This section contains one or more sets of questions with the same scenario and problem. Each question presents a unique solution to the problem. You must determine whether the solution meets the stated goals. More than one solution in the set might solve the problem. It is also possible that none of the solutions in the set solve the problem.After you answer a question in this section, you will NOT be able to return. As a result, these questions do not appear on the Review Screen.You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.Users report that some responses omit required regulatory clauses, even when the clauses are present in the retrieved content.You need to improve response completeness.Solution: You add a reflection pass that regenerates the response if the required clauses are missing.Does this meet the goal?
Note: This section contains one or more sets of questions with the same scenario and problem. Each question presents a unique solution to the problem. You must determine whether the solution meets the stated goals. More than one solution in the set might solve the problem. It is also possible that none of the solutions in the set solve the problem.After you answer a question in this section, you will NOT be able to return. As a result, these questions do not appear on the Review Screen.You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.Users report that some responses omit required regulatory clauses, even when the clauses are present in the retrieved content.You need to improve response completeness.Solution: You increase the value of the temperature parameter.Does this meet the goal?
Note: This section contains one or more sets of questions with the same scenario and problem. Each question presents a unique solution to the problem. You must determine whether the solution meets the stated goals. More than one solution in the set might solve the problem. It is also possible that none of the solutions in the set solve the problem.After you answer a question in this section, you will NOT be able to return. As a result, these questions do not appear on the Review Screen.You have a Microsoft Foundry project that contains an agent. The agent generates summaries from retrieved policy documents.Users report that some responses omit required regulatory clauses, even when the clauses are present in the retrieved content.You need to improve response completeness.Solution: You run an evaluation flow that scores responses for completeness and blocks responses that fall below a defined threshold.Does this meet the goal?
Share your comments for Microsoft AI-103 exam with other users:
Question 23:Question 23 describes a multimodal model where users can upload unsafe images that could contain hidden instructions. The goal is to implement controls to mitigate this risk. Key points to understand
Keeping this site free takes real effort. We constantly battle automated scraping and unauthorized content copying. A quick account helps us protect the community and keep the site free.
To continue studying for your AI-103, please sign in or create a free account.