1Z0-1127-25 Building LLM Applications with RAG and Vector Search Practice Test 2 — 15 Questions

Question 1

A company is building a RAG application using OCI Generative AI and OCI Search with OpenSearch. Users report that the responses from the LLM are not relevant to the queries, even though the document chunks seem appropriate. What is the most likely cause?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Reranking is not enabled in the OpenSearch query.. In a RAG pipeline, retrieving the top-K chunks by vector similarity often returns chunks that are semantically close but not the most relevant to the specific query. Reranking applies a cross-encoder or relevance model to rescore the retrieved candidates, reordering them so the most relevant chunks are passed to the LLM. If reranking is not enabled, the LLM receives poorly ordered context, producing irrelevant answers even when the chunks themselves are appropriate.

Answer

The embedding model is not suited for the domain.

Answer

The top K value is set too high.

Answer

The chunk size is too small, causing loss of context.

Question 2

An application using OCI Generative AI returns a 403 Forbidden error when attempting to invoke a model. The user's API key is valid and the endpoint is correct. What is the most likely cause?

Accepted Answer

The IAM policy does not grant the necessary permission to use the model.. A 403 Forbidden error indicates that the server understood the request but refused to authorize it. Since the API key is valid and the endpoint is correct, the most likely cause is that the IAM policy attached to the user or resource does not include a statement granting the `inference` permission for the specific model family or compartment. OCI Generative AI requires explicit IAM policies to allow model invocation, and a missing or misconfigured policy will result in a 403 even with a valid key.

Answer

The model is out of capacity.

Answer

The region is not supported for the selected model.

Answer

The API request body is malformed.

Question 3

Which OCI service provides a managed vector database capability that can be used as a knowledge base in a RAG architecture?

Accepted Answer

OCI Search with OpenSearch. OCI Search with OpenSearch provides a managed vector database capability through its k-nearest neighbor (k-NN) plugin, which supports storing and querying vector embeddings. This makes it suitable as a knowledge base in a Retrieval-Augmented Generation (RAG) architecture, where vector similarity search retrieves relevant context for LLM prompts.

Answer

OCI MySQL HeatWave

Answer

OCI Database (Autonomous Database)

Answer

OCI Object Storage

Question 4

Which TWO are required components to implement a basic RAG system using OCI services? (Choose two.)

Accepted Answer

OCI Search with OpenSearch. OCI Search with OpenSearch (D) is correct because a RAG system needs a vector store to index and retrieve document embeddings; OCI Search with OpenSearch supports vector search and serves as the retrieval layer that supplies relevant context to the LLM. OCI Document Understanding (E) is correct because it extracts text and structured content from source documents (PDFs, images, forms), which is a prerequisite for chunking and embedding in a basic RAG pipeline. OCI Object Storage (A) is not required for a basic RAG implementation, since documents can be ingested directly and the vector index lives in OpenSearch. OCI Functions (B) is not a required component; it is an optional serverless compute service for orchestration or custom logic. OCI Data Flow (C) is not required either, as it is a managed Spark service for large-scale batch processing, which is unnecessary for a basic RAG setup.

Answer

OCI Object Storage

Answer

OCI Functions

Answer

OCI Data Flow

Question 5

A company has deployed a RAG application using OCI Generative AI service with a vector store in OCI OpenSearch. Users report that answers are often incomplete or irrelevant. The application uses a single prompt template with a fixed chunk size of 1000 tokens. Which action is most likely to improve answer quality?

Accepted Answer

Implement a re-ranking step after vector search. Implementing a re-ranking step after vector search improves answer quality by reordering the top-K retrieved chunks based on their relevance to the query, using a cross-encoder or similar model. This addresses the issue of incomplete or irrelevant answers caused by the fixed chunk size and single prompt template, as re-ranking ensures that only the most contextually appropriate chunks are passed to the LLM for generation.

Answer

Disable vector search and rely solely on the LLM's pre-trained knowledge

Answer

Use a smaller embedding model to reduce noise

Answer

Increase the chunk size to 2000 tokens

Question 6

A team is designing a RAG system for a multilingual knowledge base. Which TWO strategies are appropriate? (Choose two.)

Accepted Answer

Store separate vector indices per language. Option A is correct because maintaining separate vector indices per language keeps each language's embeddings in a homogeneous semantic space, avoiding cross-lingual interference and allowing language-specific retrieval and ranking without diluting similarity scores across languages. Option E is correct because a single embedding model trained for multilingual text (e.g., multilingual E5, LaBSE, or multilingual Sentence-BERT) maps semantically equivalent phrases from different languages into a shared vector space, enabling cross-lingual retrieval with one index and consistent query handling. Option B is wrong because disabling vector search for non-English queries removes the core retrieval mechanism of a RAG system and would cripple multilingual coverage rather than improve it. Option C is not appropriate as a blanket strategy because machine-translating all documents to English before indexing introduces translation latency, cost, and semantic drift, and it discards the original language content needed for multilingual answers. Option D is wrong because using a different embedding model per language produces incompatible vector spaces, so a single query cannot be compared across indices and cross-lingual retrieval becomes impossible without additional alignment.

Answer

Disable vector search for non-English queries

Answer

Translate all documents to English before indexing

Answer

Use a different embedding model per language

Question 7

A developer is troubleshooting low recall in a vector search. Which THREE factors should be checked? (Choose three.)

Accepted Answer

Embedding model quality and relevance to domain. Option A is correct because the embedding model determines how well semantic meaning is captured; a model poorly matched to the domain (e.g., a general-purpose model on legal or medical text) will place relevant items far apart in vector space, directly lowering recall. Option B is correct because chunk size and overlap control how content is segmented; chunks that are too large dilute the embedding's semantic focus, while too little overlap splits related context across chunks, causing relevant passages to be missed. Option C is correct because the query embedding must be generated with the same model and preprocessing as the indexed documents; a mismatched or poorly formed query vector will not align with relevant document vectors, reducing recall. Option D is not among the marked answers: increasing k returns more neighbors and can improve observed recall, but it is a retrieval parameter rather than a root-cause factor in embedding or indexing quality. Option E is not among the marked answers: LLM temperature affects the randomness of generated text, not the vector similarity search, so it has no bearing on recall.

Answer

The number of results returned (k) in the search

Answer

The LLM's temperature setting

Question 8

Refer to the exhibit. A developer has set this policy to allow an OCI Data Science session to generate embeddings. However, the API call returns a 403 Forbidden. Which of the following is likely missing?

Accepted Answer

The dynamic group does not include the Data Science session. The 403 Forbidden error indicates that the Data Science session does not have the necessary permissions to call the API. In OCI, policies grant permissions to dynamic groups, not directly to resources. The dynamic group must include the Data Science session (e.g., by matching the session's OCID or resource type) for the policy to apply. Without this, the session is not recognized as a principal, and the API call is denied.

Answer

The policy needs a 'where request.region != ...' condition

Answer

The policy should include 'in tenancy' instead of compartment

Answer

The service requires 'manage' permission instead of 'use'

Question 9

Refer to the exhibit. A RAG application logs this error when trying to search. What is the most likely cause?

Accepted Answer

The index name is misspelled in the application configuration. The error log indicates that the RAG application cannot find the specified index when performing a vector search. This is most commonly caused by a mismatch between the index name configured in the application and the actual index name in OpenSearch. Option C is correct because a misspelled index name will cause the search request to fail with a '404 index_not_found_exception' or similar error, even if the cluster is healthy and the embedding model is valid.

Answer

The embedding model is incompatible

Answer

The OpenSearch cluster is not accessible

Answer

The query syntax is incorrect

Question 10

A development team notices that their RAG application returns responses slowly when processing large PDF documents (100+ pages). They need to improve response time without significantly reducing retrieval quality. Which action is most effective?

Accepted Answer

Use a smaller chunk size during document ingestion. Using a smaller chunk size during document ingestion reduces the amount of text per chunk, which decreases the latency of embedding generation and vector search. For large PDFs, smaller chunks also minimize the risk of retrieving overly large context windows that slow down the LLM's response generation, while still preserving retrieval quality if overlap is properly configured.

Answer

Add a reranking step after initial retrieval

Answer

Increase the topK parameter to retrieve more context

Answer

Switch to a larger embedding model for better accuracy

Question 11

An AI engineer observes that the RAG application fails to retrieve relevant documents for certain user queries, despite having a comprehensive knowledge base. The issue appears to be a semantic gap between query phrasing and document content. Which technique should the engineer implement first to address this?

Accepted Answer

Apply query expansion techniques before embedding the user query. Query expansion techniques (e.g., synonym injection, back-translation, or LLM-based paraphrasing) directly address the semantic gap by enriching the user query with alternative phrasings before embedding. This increases the likelihood of matching relevant document chunks in the vector space, without altering the retrieval architecture or requiring additional inference steps.

Answer

Switch from dense to sparse vector embeddings

Answer

Implement a re-ranking model to reorder retrieved results

Answer

Increase the chunk overlap to ensure more context

Question 12

An organization wants to combine keyword search and vector search to improve retrieval accuracy in their RAG pipeline. Which OCI service provides built-in hybrid search capabilities?

Accepted Answer

OCI OpenSearch. OCI OpenSearch is the correct answer because it natively supports hybrid search, which combines keyword-based (BM25) and vector-based (k-NN) queries in a single search request. This allows the RAG pipeline to retrieve documents that match both exact terms and semantic meaning, improving overall accuracy without requiring separate search systems.

Answer

OCI Search with AI

Answer

Autonomous Database with AI Vector Search

Answer

OCI Logging

Question 13

Which TWO best practices should be followed when designing a RAG application using OCI GenAI? (Select two.)

Accepted Answer

Use OCI Dedicated AI Cluster for inference to ensure data privacy. Option C is correct because an OCI Dedicated AI Cluster provides isolated, single-tenant compute for model inference, so prompts and completions never share infrastructure with other customers, which is essential for data privacy in a RAG application handling sensitive content. Option E is correct because dedicated AI endpoints expose the model through a private, workload-specific endpoint that can be scoped to a VCN and controlled with IAM policies, keeping retrieval-augmented generation traffic off shared public endpoints for sensitive workloads. Option A is wrong because batching user queries degrades response latency and relevance in an interactive RAG flow and does not address retrieval quality or privacy. Option B is wrong because a vector database should store embeddings plus metadata and document references, not raw source documents, which belong in Object Storage or another document store for update and lifecycle management. Option D is wrong because, while OCI Vault is the right service for secrets, embedding API keys are not the primary design concern here and frequent rotation alone does not satisfy the privacy requirements that dedicated infrastructure provides.

Answer

Batch all user queries to minimize costs

Answer

Store raw documents in the vector database for easy updates

Answer

Store embedding API keys in OCI Vault and rotate frequently

Question 14

A developer calls the OCI GenAI embedding API as shown in the exhibit. What is the most likely cause of the error?

Accepted Answer

The input string is too long for the embedding model. The error is most likely caused by the input string exceeding the maximum token limit for the embedding model. OCI GenAI embedding models have a fixed context window (e.g., 512 or 1024 tokens), and if the input text is longer than that, the API returns an error. The developer must truncate or chunk the input before calling the API.

Answer

The API key does not have permission to call the endpoint

Answer

The endpoint ID is incorrect

Answer

The model ID is not supported for embedding

Question 15

An IAM policy is shown in the exhibit. A user reports that they cannot call the OCI GenAI embedding API, but they can use OCI AI Language. Which policy statement is missing to allow embedding API access?

Accepted Answer

The policy needs 'use' action on 'oci-generative-ai-family'. The user can already use OCI AI Language, indicating that the existing policy grants access to the 'oci-ai-language-family' resource. To also allow access to the OCI GenAI embedding API, the policy must include a 'use' verb on the 'oci-generative-ai-family' resource type. The 'use' verb is the minimum required permission for invoking inference operations like embedding, while 'manage' would be excessive and 'inspect' is insufficient for API calls.

Answer

The policy needs 'inspect' on 'oci-generative-ai-endpoint'

Answer

The policy needs 'manage' action on 'oci-generative-ai-family'

Answer

The policy needs 'inspect' on 'oci-ai-language-family'