---
格式版本: 2
标题: "Model Selection, Scheduled Enrichment, and Background SQL Generation Available for Enterprise AI NL2SQL"
原文链接: "https://docs.oracle.com/en-us/iaas/releasenotes/generative-ai/nl2sql-model-selection-background-generation.htm"
发布日期: "2026-08-26"
发布时间校准状态: "found"
发布时间需复核: "否"
发布时间来源: "rule:scrape:strict_html_body"
发布时间证据: "releasedate: 2026-08-26"
发布时间校准原因: "规则确认唯一严格发布时间，来源 scrape:strict_html_body"
发布时间校准置信度: "high"
发布时间候选数量: 16
发布时间严格候选数量: 5
发布时间原页读取状态: "source template page reused from URL open"
发布时间未找到原因: ""
发布时间校准时间: "2026-08-27T11:02:38+08:00"
发布时间仲裁状态: "skipped"
发布时间仲裁尝试次数: 0
发布时间仲裁耗时毫秒: 0
发现时间: "2026-08-27T11:02:05+08:00"
入库时间: "2026-08-27T03:02:38.638Z"
来源平台: "固定入口"
搜索渠道: "fixed_url"
搜索词: "https://docs.oracle.com/en-us/iaas/releasenotes"
匹配关键词:
  - "AI"
相关厂家:
  - "Oracle"
  - "OpenAI"
相关专家:
  []
内容类型: "网页"
抓取工具: "Free Fetch + Defuddle"
清洗工具: "Defuddle Markdown + Defuddle/Readability 正文提取"
原始附件:
  []
AI优质: "否"
AI打分: 37
AI分档: "非优质"
AI质检状态: "不通过"
AI打分理由: "正文主线是Oracle Enterprise AI NL2SQL的软件功能更新，包括模型选择、定时语义存储刷新和后台SQL生成；来源为Oracle官方文档，内容完整，但未提供发布日期。固定知识库未发现同一功能记录，但未命中不能证明首次发布。新增事实可核验，却不涉及超节点、AI机架、互连、供电、散热或机架级基础设施，也无客户、量产或部署信号；命中应用与模型效率类硬否决，当前页面不适合作为超节点业务情报源。"
AI质检模型: "gpt-5.6-sol"
AI质检时间: "2026-08-27T11:02:47+08:00"
AI主题相关性: 1
AI来源权威性: 15
AI新颖性: 8
AI技术细节: 4
AI商业部署信号: 0
AI完整性: 9
AI评分提示词版本: "v17-精简生产版"
AI评分提示词SHA256: "48fb9777f386026761b4873eaff30807694fb11e9b352d7c69bf2dfde750cc7d"
AI评分知识库版本: "knowledge_base_v1-20260819+runtime.7"
AI评分知识库SHA256: "2d08daa800e4aaf8f32428770a398a6fddb773ec6de396decbc9d7689bcb65f7"
AI评分知识库检索词: "[\"Oracle\",\"https://docs.oracle.com/en-us/iaas/releasenotes\",\"Meta\",\"SQL\",\"NL2SQL\",\"SQL-generation\",\"API\",\"openai.gpt-oss-120b\",\"ISO\",\"PT6H\",\"P1D\",\"SDKs\"]"
AI评分知识库命中: "[{\"id\":\"july-correct-0018\",\"title\":\"AMD, Cerebras partner on joint Helios rack-scale AI inference platform\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"Oracle\",\"Meta\"],\"rank\":-7.834012158879961},{\"id\":\"historical-jun-014\",\"title\":\"NVIDIA发布Vera CPU Rack整机柜方案，面向AI工厂中的Agentic AI与强化学习负载\",\"sourceType\":\"curated_item\",\"time\":\"2026-06\",\"matchedTerms\":[\"Oracle\"],\"rank\":-6.190228138549601},{\"id\":\"july-correct-0075\",\"title\":\"From EPYC to Helios, AMD and Meta are Scaling the Future of AI Together\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"Meta\",\"API\"],\"rank\":-5.382340070877827},{\"id\":\"july-correct-0001\",\"title\":\"全球首颗2nm GPU来了！苏姿丰甩出“最强AI机架”，CPU性能干翻英伟达 - 智东西\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"Oracle\",\"Meta\"],\"rank\":-4.961040664491328},{\"id\":\"july-correct-0033\",\"title\":\"Microsoft, Alphabet, Meta Pivot from Buy to Build in AI\",\"sourceType\":\"labeled_article\",\"time\":\"2026-07\",\"matchedTerms\":[\"Meta\",\"API\",\"ISO\"],\"rank\":-4.873033957753761}]"
AI摘要: "Oracle Enterprise AI NL2SQL 新增模型选择、定时增量丰富和后台SQL生成能力；未指定时默认使用 openai.gpt-oss-120b 模型，丰富任务可按 ISO 8601 周期调度，最小间隔 6 小时。"
AI摘要模型: "ali-deepseek-v4-flash"
AI摘要时间: "2026-08-27T08:12:11.136Z"
采集批次: "2026年8月27日10点55分46秒"
采集批次ID: "20260827-105546-808"
去重键: "https://docs.oracle.com/en-us/iaas/releasenotes/generative-ai/nl2sql-model-selection-background-generation.htm"
---

Enterprise AI NL2SQL now supports model selection, scheduled enrichment, and background processing for SQL-generation requests.

Model selection

Select the Generative AI model that NL2SQL uses to enrich a semantic store. By using the API, you can also select a model for an individual SQL-generation request. If you don't select a model, NL2SQL uses the default `openai.gpt-oss-120b` model.

You can select any on-demand Generative AI model that's available in the region and accessible to your tenancy. See [Generative AI Models by Region](https://docs.oracle.com/iaas/Content/generative-ai/model-endpoint-regions.htm).

Scheduled enrichment

Refresh a semantic store's enriched metadata on a recurring schedule. Specify the schedule as an ISO 8601 duration. The minimum interval is six hours. For example, use `PT6H` to refresh every six hours or `P1D` to refresh once a day. Each refresh uses the selected Generative AI model.

To perform a delta refresh through the API, call `GenerateEnrichmentJob` and set `enrichmentJobConfiguration` to a `DeltaRefreshEnrichmentJobConfiguration`. A delta refresh updates only the database objects that changed since the most recent enrichment.

Background SQL generation

Submit an SQL-generation request as a background job when processing might exceed the client's normal request timeout. The request returns a job that you can monitor until SQL generation succeeds or fails. When the job succeeds, retrieve the generated SQL from the job result.

Background SQL generation is available through the API, SDKs, and CLI. It isn't available in the Console.

For more information, see the [Enterprise AI NL2SQL Quick Start Guide](https://docs.oracle.com/iaas/Content/generative-ai/nl2sql.htm).