[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-databricks-ai-gateway-inference-tables-served-models-zh":3,"article-related-databricks-ai-gateway-inference-tables-served-models-zh":31,"series-tools-2ffaf5e5-e155-47dd-80cd-af62c3705516":74},{"id":4,"slug":5,"title":6,"content":7,"summary":8,"source":9,"source_url":10,"author":11,"image_url":12,"cover_image":12,"category":13,"language":14,"translated_content":11,"related_article_id":15,"keywords":16,"key_takeaways":23,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":30},"2ffaf5e5-e155-47dd-80cd-af62c3705516","databricks-ai-gateway-inference-tables-served-models-zh","Databricks 為模型服務加上 AI Gateway 推論表","\u003Cp data-speakable=\"summary\">Databricks 在 6 月 30 日更新文件，為模型\u003Ca href=\"\u002Fnews\u002Fbootdev-go-course-turns-syntax-into-services-zh\">服務\u003C\u002Fa>端點加入 AI Gateway 推論表，會把請求與回應資料寫進 Unity Catalog。\u003C\u002Fp>\u003Cp>Databricks 於 2026 年 6 月 30 日更新 \u003Ca href=\"\u002Ftag\u002Faws\">AWS\u003C\u002Fa> 文件，說明 \u003Ca href=\"https:\u002F\u002Fwww.databricks.com\u002Fproduct\u002Fai-gateway\" target=\"_blank\" rel=\"noopener\">Databricks\u003C\u002Fa> 的 AI Gateway 已可搭配 \u003Ca href=\"\u002Ftag\u002Finference\">inference\u003C\u002Fa> tables 使用，支援的對象包含 Model Serving 與已部署的 \u003Ca href=\"\u002Ftag\u002Fai-agent\">AI agent\u003C\u002Fa>。這項功能會自動蒐集請求、回應、HTTP 狀態碼、執行時間、request ID 與 trace，集中存到 \u003Ca href=\"https:\u002F\u002Fdocs.databricks.com\u002Fen\u002Fcatalog\u002Findex.html\" target=\"_blank\" rel=\"noopener\">Unity Catalog\u003C\u002Fa> 的 Delta table。\u003Cp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>項目\u003C\u002Fth>\u003Cth>數值\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>文件更新日期\u003C\u002Ftd>\u003Ctd>2026 年 6 月 30 日\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>支援的端點類型\u003C\u002Ftd>\u003Ctd>5 種\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>每個部署自動建立的 AI agent 表\u003C\u002Ftd>\u003Ctd>3 類資料\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>payload 資料可用時間\u003C\u002Ftd>\u003Ctd>1 小時內\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>發生了什麼\u003C\u002Fh2>\u003Cp>這次更新把 inference tables 從單純的記錄機制，提升成 Model Serving 的內建可觀測層。只要啟用，系統就會把進站請求、出站回應、狀態碼、執行時間、request ID 與 traces 寫入 Unity Catalog，讓團隊可以直接查詢，不必再翻分散的日誌。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782939770301-2a9e.png\" alt=\"Databricks 為模型服務加上 AI Gateway 推論表\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>文件也把範圍擴到 AI \u003Ca href=\"\u002Ftag\u002Fagent\">agent\u003C\u002Fa>。對 agent 來說，Databricks 會保存 payload、request 細節與 MLflow Trace logs；如果是透過 \u003Ca href=\"https:\u002F\u002Fmlflow.org\u002Fdocs\u002Flatest\u002Fpython_api\u002Fmlflow.deploy.html\" target=\"_blank\" rel=\"noopener\">mlflow.deploy()\u003C\u002Fa> 部署，系統還會自動建立 inference tables。支援的端點類型包含：\u003C\u002Fp>\u003Cul>\u003Cli>Provisioned throughput\u003C\u002Fli>\u003Cli>Pay-per-token\u003C\u002Fli>\u003Cli>External models\u003C\u002Fli>\u003Cli>Deployed AI agents\u003C\u002Fli>\u003Cli>Custom models\u003C\u002Fli>\u003C\u002Ful>\u003Cp>文件同時列出限制：工作區必須有 Unity Catalog、serverless compute，且所在區域要支援 model serving。Databricks 也提醒，端點建立者與修改者都需要 Can Manage 權限，外加 USE CATALOG、USE SCHEMA、CREATE TABLE 等權限；而且系統只會為新表建立 inference table，既有表不支援直接套用。\u003C\u002Fp>\u003Ch2>為什麼重要\u003C\u002Fh2>\u003Cp>對\u003Ca href=\"\u002Fnews\u002Fzhihe-a210-risc-v-soc-dev-kit-breakdown-zh\">開發\u003C\u002Fa>者來說，這代表模型流量不再只是黑盒子裡的雜訊，而是可以直接查詢、聯集、比對的資料。團隊能把 inference tables 跟 ground truth 標籤接起來，做評估集、找出錯誤樣本、追 drift，甚至直接在 Databricks SQL 或 notebook 裡檢查失敗案例。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782939768160-qsou.png\" alt=\"Databricks 為模型服務加上 AI Gateway 推論表\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>對維運與平台團隊，這也讓 agent debugging 更好做。因為表格會帶著 MLflow traces 和 request metadata，團隊可以追慢請求、比對歷史輸入，快速看出延遲或回應品質是從哪一版開始變化。Databricks 另外提到新的 \u003Ca href=\"https:\u002F\u002Fdocs.databricks.com\u002Faws\u002Fen\u002Fai-gateway\u002F\" target=\"_blank\" rel=\"noopener\">Unity AI Gateway\u003C\u002Fa> beta，定位是管理 \u003Ca href=\"\u002Ftag\u002Fllm\">LLM\u003C\u002Fa> 端點與 coding agents 的企業控制平面，顯示推論、治理與觀測正在往同一套工作流集中。\u003C\u002Fp>\u003Cp>Databricks 也明講，改動 schema、重新命名或刪除表格，都可能讓 logging 中斷或損壞表格；對 AI agent 而言，舊的 request logs 與 assessment logs 也正被新 payload tables 取代。對團隊來說，\u003Ca href=\"\u002Fnews\u002Fbineval-binary-questions-llm-evals-zh\">問題\u003C\u002Fa>已不是要不要記錄模型流量，而是要不要從第一天就把這些資料納入 Unity Catalog 管理。\u003C\u002Fp>","Databricks 於 6 月 30 日更新 AWS 文件，為 Model Serving 與 AI agent 加入 AI Gateway 推論表，將請求、回應與追蹤資料寫入 Unity Catalog。","docs.databricks.com","https:\u002F\u002Fdocs.databricks.com\u002Faws\u002Fen\u002Fai-gateway\u002Finference-tables",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1782939770301-2a9e.png","tools","zh","610d3dfe-c451-42a0-a51a-adbee93932f5",[17,18,19,20,21,22],"Databricks","AI Gateway","Inference Tables","Unity Catalog","Model Serving","AI agent",[24,25,26],"推論表把模型服務請求、回應與 trace 直接寫進 Unity Catalog。","AI agent 也納入支援，mlflow.deploy() 部署可自動建立表格。","對開發與維運團隊，這讓除錯、評估與 drift 監控更容易在同一平台完成。",3,"2026-07-01T21:02:20.660434+00:00","2026-07-01T21:02:20.652+00:00","ddbe17bf-4560-43f7-af76-3e7d6e08e601",{"tags":32,"relatedLang":33,"relatedPosts":37},[],{"id":15,"slug":34,"title":35,"language":36},"databricks-ai-gateway-inference-tables-served-models-en","Databricks adds AI Gateway inference tables for served models","en",[38,44,50,56,62,68],{"id":39,"slug":40,"title":41,"cover_image":42,"image_url":42,"created_at":43,"category":13},"419deb6d-ea65-4fa7-9dd1-73763f373b53","bailian-token-plan-agent-credits-guide-zh","百炼Token Plan把Agent接入变简单","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783929835818-cyp6.png","2026-07-13T08:03:24.024727+00:00",{"id":45,"slug":46,"title":47,"cover_image":48,"image_url":48,"created_at":49,"category":13},"5c5c6733-6f41-49ca-b61e-c0a53399c327","one-api-gateway-turns-six-ai-apis-into-one-zh","一個閘道把六個 AI API 收成一套","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783883027640-gz6i.png","2026-07-12T19:03:22.171684+00:00",{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"953466bc-6d94-4ffe-944b-ac7728d09184","fde-playbook-fixes-broken-agent-projects-zh","FDE 讓烂尾 Agent 回正軌","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783881221882-yw1n.png","2026-07-12T18:33:16.193979+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"bd3b5d5c-02d9-49c4-8ca4-dcb6df373374","ai-zaobao-template-daily-to-reusable-report-zh","AI 早报拆成可复用周报","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783879420413-rlwb.png","2026-07-12T18:03:11.617929+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"33302498-49ec-4a00-8e2c-61b6f9fc3ece","claude-reflect-turns-usage-into-retention-zh","Claude Reflect 把使用變留存","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783793005956-n1ea.png","2026-07-11T18:03:01.486301+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"5f08bf54-0ddb-4884-b541-75e33ad21b30","midjourney-turns-prompt-ideas-into-art-zh","Midjourney 把提示詞變成可重用流程","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1783694003831-eilt.png","2026-07-10T14:32:55.765266+00:00",[75,80,85,90,95,100,105,110,115,120],{"id":76,"slug":77,"title":78,"created_at":79},"855cd52f-6fab-46cc-a7c1-42195e8a0de4","surepath-real-time-mcp-policy-controls-zh","SurePath 推出即時 MCP 政策控管","2026-03-26T07:57:40.77233+00:00",{"id":81,"slug":82,"title":83,"created_at":84},"9b19ab54-edef-4dbd-9ce4-a51e4bae4ebb","mcp-in-2026-the-ai-tool-layer-teams-use-zh","2026 年 MCP：團隊真的在用的 AI 工具層","2026-03-26T08:01:46.589694+00:00",{"id":86,"slug":87,"title":88,"created_at":89},"af9c46c3-7a28-410b-9f04-32b3de30a68c","prompting-in-2026-what-actually-works-zh","2026 提示工程，真正有用的是什麼","2026-03-26T08:08:12.453028+00:00",{"id":91,"slug":92,"title":93,"created_at":94},"05553086-6ed0-4758-81fd-6cab24b575e0","garry-tan-open-sources-claude-code-toolkit-zh","Garry Tan 開源 Claude Code 工具包","2026-03-26T08:26:20.068737+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"042a73a2-18a2-433d-9e8f-9802b9559aac","github-ai-projects-to-watch-in-2026-zh","2026 必看 20 個 GitHub AI 專案","2026-03-26T08:28:09.619964+00:00",{"id":101,"slug":102,"title":103,"created_at":104},"a5f94120-ac0d-4483-9a8b-63590071ac6a","claude-code-vs-cursor-2026-zh","Claude Code 與 Cursor 深度對比：202…","2026-03-26T13:27:14.279193+00:00",{"id":106,"slug":107,"title":108,"created_at":109},"0975afa1-e0c7-4130-a20d-d890eaed995e","practical-github-guide-learning-ml-2026-zh","2026 機器學習入門 GitHub 實用指南","2026-03-27T01:16:49.712576+00:00",{"id":111,"slug":112,"title":113,"created_at":114},"bfdb467a-290f-4a80-b3a9-6f081afb6dff","aiml-2026-student-ai-ml-lab-repo-review-zh","AIML-2026：像課綱的學生實驗 Repo","2026-03-27T01:21:51.467798+00:00",{"id":116,"slug":117,"title":118,"created_at":119},"80cabc3e-09fc-4ff5-8f07-b8d68f5ae545","ai-trending-github-repos-and-research-feeds-zh","AI Trending：把 AI 資源收成一張表","2026-03-27T01:31:35.262183+00:00",{"id":121,"slug":122,"title":123,"created_at":124},"3ce6e6e2-bac5-463e-9f8d-45caabcc61f7","awesome-ai-for-science-research-tools-map-zh","AI 科研工具清單，開始像地圖了","2026-03-27T01:46:50.521945+00:00"]