AWS Certified Generative AI Developer - Professional (AIP-C01) (AIP-C01) (AIP-C01) — Questions 151–169

ConverseStream is the Bedrock Converse API operation that returns incremental response chunks over an event stream. Selecting it satisfies the stem's streaming requirement, whereas the non-streaming Converse call returns the complete response only after generation finishes.

Amazon Bedrock Guardrails apply sensitive-information filters that detect and mask PII in prompts before inference, satisfying the requirement that data be redacted prior to reaching the foundation model. Configured via `sensitiveInformationPolicyConfig`, they intercept the request at the service boundary, so the model never receives raw personal data.

Amazon Bedrock Guardrails applies configurable policies that evaluate prompts and responses against denied topics and harmful content categories, blocking or filtering them. This directly satisfies the requirement to restrict user queries by topic or content category, which base model invocation alone cannot enforce.

A system prompt supplies persistent instructions that shape a model's tone, role, and boundaries across every turn, unlike per-request user messages. This directly satisfies the stem's constraint of guiding behaviour, establishing consistent operating rules before conversation begins. It does not configure infrastructure, guardrails, or knowledge retrieval.

PrivateLink interface endpoints for Bedrock and a VPC endpoint for OpenSearch Serverless keep traffic on the AWS private network, so requests never traverse the public internet. This satisfies the constraint that all data remains within the VPC.

InvocationLatency is emitted natively by Amazon Bedrock to CloudWatch, measuring end-to-end time from request receipt to full response completion. It directly satisfies the stem's requirement for invocation performance monitoring without custom instrumentation, unlike token-level or throttling metrics that address different concerns.

Amazon Bedrock encrypts all data in transit between the client and the service using TLS 1.2 or higher, protecting prompts containing sensitive data during transmission. This satisfies the compliance requirement for encryption of prompts while in flight.

Batching multiple products into one invocation reduces the number of API calls and makes better use of the model's token capacity per request. This directly lowers the chance of hitting per-account request quotas and improves throughput. The model and output quality remain unchanged because the same model processes a larger prompt. This is the most operationally efficient approach for the described scenario.

Keeping data in a Region-bound bucket satisfies residency, scoping the customization action to the data scientist role enforces authorization, and a cross-account IAM role or resource policy limits model invocation to the production account. This combination meets all three legal requirements.

Amazon Bedrock Model Evaluation provides managed jobs that score foundation model outputs against built-in or custom metrics for quality, safety and robustness. It directly satisfies the requirement for a formal testing framework, unlike ad-hoc prompt testing or general monitoring tools.

Amazon Bedrock's knowledge bases require embeddings produced by a supported embedding model, and vector spaces are model-specific. Embeddings from the on-premises model are therefore incompatible and cannot be reused. Re-embedding the source documents with a Bedrock-supported model ensures retrieval works correctly.

A Sync operation on the Knowledge Base re-ingests and re-indexes the underlying data sources, so newly added or changed documents become retrievable. This satisfies the freshness constraint: without syncing, the RAG system keeps serving stale embeddings from the previous ingestion.

Bedrock Guardrails content filters evaluate prompts and responses against configurable hate, violence, sexual and insult categories, blocking matching content. This satisfies the requirement centrally at the model invocation layer, without custom prompt engineering or separate moderation infrastructure per application.

Removing personally identifiable information satisfies data privacy obligations and limits exposure if the knowledge base is breached or over-shared. Retrieval-augmented generation can surface indexed content to any permitted user, so minimising stored PII directly reduces regulatory and reputational risk under frameworks such as UK GDPR.

IAM least privilege restricts each agent's permissions to only the actions its task requires, limiting blast radius if the agent is compromised or misbehaves. This satisfies the security requirement by constraining agentic autonomy through tightly scoped, auditable role assignments.

Amazon Bedrock Prompt Management stores prompt templates as versioned artefacts, letting you create, compare and roll back variants natively. It satisfies the version-control requirement without building custom tooling, unlike generic stores such as S3 or Parameter Store.

Proper chunking directly impacts the relevance and efficiency of retrieval.

The Converse API accepts documents as content blocks inside the messages array, letting each document be attached to a user turn for the model to consider. Placing them in the top-level request body or system prompt does not satisfy the stem's requirement for document content blocks.

AWS Service Quotas provides a central console and API for viewing and requesting increases to Bedrock quotas, including per-region token and request limits. It satisfies the stem's regional constraint because quota values are scoped per AWS Region, letting you raise limits in the specific region your application calls.