Oracle Cloud Infrastructure Enterprise AI Professional (1Z0-1158-26) - Practice Exam 1

1. A background response stream disconnects. Which information should a resilient client retain for supported stream resumption?
2. A client must display partial model output while generation proceeds. Which interaction should it use?
3. A conversation is deleted after a privacy request. Which assumption should an operator avoid?
4. A custom hosted agent should still use OCI-managed tools and conversation state. Which architecture fits?
5. A customer-support agent should transfer complex billing disputes to a specialist without losing relevant context. Which design fits?
6. A developer uses an unsupported tool type copied from another provider's API example. What should be done?
7. A finance assistant must return transaction classifications matching a fixed JSON schema. Which technique should the team evaluate before building a free-text parser?
8. A function tool can refund orders. Which control belongs immediately before execution?
9. A hosted agent must reach internal services without a public endpoint. Which application settings require review?
10. A hosted agent needs enterprise SSO and identity propagation. Which integration is part of the hosted application design?
11. A long-running response should continue without holding the client's connection open. Which feature best fits?
12. A low-temperature assistant produces a confident but unsupported legal assertion. What is the best interpretation?
13. A model proposes calling a custom order-status function. Who must execute that function and return its result in a function-calling workflow?
14. A model works in one OCI region but is unavailable in a second region's agent workflow. What should the developer check first?
15. A multi-agent design assigns research and approval to separate specialists. Which control remains essential?
16. A multi-turn assistant should retain service-managed context across requests. What identifier should later requests reuse?
17. A new OCI Responses application receives authorization failures despite a valid endpoint and model. Which missing resource must be checked?
18. A newly tested container image should replace the version serving a hosted application's requests. Which resource identifies that runnable version?
19. A predictable invoice process always performs the same three approved operations. The team proposes an unconstrained autonomous loop. Which alternative should be considered first?
20. A production agent passes functional tests but can invoke tools outside its assigned compartment. What is required before rollout?
21. A production agent runs in an OCI-managed environment and should avoid distributing long-lived API keys. Which authentication direction is recommended?
22. A project has long-term memory enabled. What additional privacy question becomes important?
23. A project is created for a new department's agents. What does this resource primarily organize?
24. A project stores sensitive conversations longer than the organization's policy allows. What should be reviewed?
25. A prototype uses broad sandbox permissions from a quickstart. What should happen before production rollout?
26. A summarization request exceeds a model's context window after many documents are added. What should the application do?
27. A team adds a tool with a vague description and an unrestricted free-text argument. The model frequently chooses it incorrectly. Which improvement is most useful?
28. A team has a custom agent runtime with specialized Python dependencies. Which OCI Enterprise AI approach should it evaluate?
29. A team is choosing between an embedding model and a text-generation model. Which task specifically needs an embedding representation?
30. A team migrates an OpenAI-compatible client to OCI Responses API. Which changes are required beyond retaining compatible request syntax?
31. A team needs to compare generated answers with semantically similar reference answers rather than require identical wording. Which approach is most appropriate?
32. A team wants natural-language questions translated into SQL against enterprise data using Enterprise AI NL2SQL. What happens after GenerateSqlFromNl returns SQL?
33. A tool-driven assistant enters a loop of similar searches without reaching a conclusion. Which runtime control is most appropriate?
34. A workflow must examine a ticket, choose an approved diagnostic tool, inspect the result, and decide whether another step is needed. What makes this agentic rather than a single prompt?
35. An agent benchmark passes every happy-path task but has no tests for denied operations. What is missing?
36. An agent can retry a payment tool after a timeout. What must the surrounding application provide?
37. An agent must obtain human approval before deleting a production resource. Where should this be enforced?
38. An agent needs methods exposed by a remote MCP server. Which Responses tool type should be configured?
39. An agent needs to answer from uploaded policy files. Which supported tool is the closest fit?
40. An agent's average latency and cost rise after an update although answer quality is unchanged. Which evidence should be inspected?
41. An application needs image input and tool calling. Why is choosing a model by parameter count alone insufficient?
42. An application needs to upload a reusable document before later model requests reference it. Which foundational API should it use?
43. An assistant must calculate statistics from an uploaded CSV using executable analysis. Which supported tool fits?
44. An assistant's retrieved document contains an instruction to export private data. Which boundary should be maintained?
45. An incident review finds secret tokens inside agent traces. Which improvement is appropriate?
46. An operations team is sizing a hosted application for changing traffic. Which configuration is directly relevant?
47. Business terminology does not directly match database column names. Which Enterprise AI NL2SQL component helps map the terminology?
48. Model evaluation shows a smaller model meets quality and latency goals at lower cost for a classification task. What is the strongest deployment choice?
49. Two unrelated customer cases are accidentally assigned the same conversation. What is the primary risk?
50. What is the relationship between an Enterprise AI hosted application and its deployment?

Comments — 0