[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-vector-databases-aws-explained-zh":3,"article-related-vector-databases-aws-explained-zh":38,"series-tools-6ca36c73-d147-4134-913d-7e1df080899f":89},{"id":4,"title":5,"content":6,"summary":7,"source":8,"source_url":9,"author":10,"image_url":11,"keywords":12,"language":21,"translated_content":10,"views":22,"is_premium":23,"created_at":24,"updated_at":24,"cover_image":11,"published_at":25,"rewrite_status":26,"rewrite_error":10,"rewritten_from_id":27,"slug":28,"category":29,"related_article_id":30,"status":31,"google_indexed_at":10,"x_posted_at":10,"tweet_text":10,"title_rewritten_at":10,"title_original":10,"key_takeaways":32,"topic_cluster_id":36,"embedding":37,"is_canonical_seed":23},"6ca36c73-d147-4134-913d-7e1df080899f","AWS 怎麼看向量資料庫","\u003Cp data-speakable=\"summary\">AWS 說向量\u003Ca href=\"\u002Fnews\u002Foracle-ai-doesnt-need-another-database-zh\">資料庫\u003C\u002Fa>會存 embeddings，讓應用程式快速找出語意相近的資料。\u003C\u002Fp>\u003Cp>說真的，這東西現在很實用。向量搜尋已經不是實驗室玩具，而是 \u003Ca href=\"\u002Fnews\u002Fhow-to-build-an-agentic-ai-crypto-stack-zh\">AI\u003C\u002Fa> 應用的基本零件。\u003C\u002Fp>\u003Cp>AWS 把重點講得很直白。它把向量當成高維資料點，再用最近鄰搜尋找出意思相近的內容，不是只看字面是否一樣。\u003C\u002Fp>\u003Cp>這件事會影響搜尋、推薦、聊天機器人。只要你的資料有文字、圖片、音訊，\u003Ca href=\"\u002Fnews\u002Fhow-to-choose-a-vector-database-in-2026-zh\">向量資料\u003C\u002Fa>庫就很可能派得上用場。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>主題\u003C\u002Fth>\u003Cth>AWS 說法\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>資料模型\u003C\u002Ftd>\u003Ctd>向量存成高維資料點\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>索引方法\u003C\u002Ftd>\u003Ctd>k-NN、HNSW、IVF\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>Bedrock 建議\u003C\u002Ftd>\u003Ctd>\u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fopensearch-service\u002F\" target=\"_blank\" rel=\"noopener\">Amazon OpenSearch Service\u003C\u002Fa>\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>MemoryDB 數字\u003C\u002Ftd>\u003Ctd>數百萬向量、毫秒級回應、每秒數萬次查詢、回收率超過 99%\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Ch2>向量資料庫到底在做什麼\u003C\u002Fh2>\u003Cp>向量資料庫存的是 embeddings。這些數字向量來自模型輸出，代表內容的語意和上下文。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778973842531-shap.png\" alt=\"AWS 怎麼看向量資料庫\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>講白了，它不是拿字串比對字串。它是拿向量空間裡的距離，去判斷兩筆資料像不像。\u003C\u002Fp>\u003Cp>AWS 文章把核心工作講得很簡單。存向量、找向量、快速找最近鄰。\u003C\u002Fp>\u003Cp>但真正上線時，事情沒那麼單純。資料庫還要處理索引、查詢規劃、權限、容錯和資料管理。\u003C\u002Fp>\u003Cp>這也是為什麼大家不直接拿原始 k-NN 硬幹。演算法只是起點，真正能跑 production 的，是整套資料庫能力。\u003C\u002Fp>\u003Cul>\u003Cli>向量可以代表文字、圖片、音訊、掃描文件。\u003C\u002Fli>\u003Cli>相似度搜尋可用在圖片查找和語意搜尋。\u003C\u002Fli>\u003Cli>資料庫會補上索引、權限、容錯。\u003C\u002Fli>\u003Cli>這些功能比單純的數學搜尋更像正式基礎設施。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>AWS 為什麼一直提向量資料庫\u003C\u002Fh2>\u003Cp>AWS 的立場很務實。embedding 有用，前提是開發者能把它用進系統裡。\u003C\u002Fp>\u003Cp>所以向量資料庫的價值，不是「能不能算距離」。而是能不能把 embeddings 建索引、查詢，還能跟 metadata 一起做 hybrid search。\u003C\u002Fp>\u003Cp>這裡才是重點。你可以同時看語意和關鍵字，再用向量相似度和 BM25 一起排序。\u003C\u002Fp>\u003Cp>對很多產品來說，這種做法比純語意搜尋更穩。結果比較像人類真的會接受的答案。\u003C\u002Fp>\u003Cp>向量資料庫也跟\u003Ca href=\"\u002Ftag\u002F生成式-ai\">生成式 AI\u003C\u002Fa> 的幻覺問題有關。外部知識庫可以幫聊天機器人找資料，降低亂答機率。\u003C\u002Fp>\u003Cblockquote>\"OpenSearch Service offers a scalable and high-performance vector database enabling vector-driven search capabilities.\" — Amazon Web Services\u003C\u002Fblockquote>\u003Cp>這句話很直接。AWS 沒把向量搜尋當附加功能，而是把 \u003Ca href=\"https:\u002F\u002Fopensearch.org\u002F\" target=\"_blank\" rel=\"noopener\">OpenSearch\u003C\u002Fa> 放到 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fbedrock\u002F\" target=\"_blank\" rel=\"noopener\">Amazon Bedrock\u003C\u002Fa> 的檢索路線上。\u003C\u002Fp>\u003Cp>對開發者來說，意思很清楚。你要做 \u003Ca href=\"\u002Ftag\u002Frag\">RAG\u003C\u002Fa>，AWS 想先讓你想到 OpenSearch Service。\u003C\u002Fp>\u003Cul>\u003Cli>向量搜尋可搭配 metadata 篩選。\u003C\u002Fli>\u003Cli>Hybrid search 比單一語意搜尋更穩。\u003C\u002Fli>\u003Cli>RAG 系統常需要外部知識庫。\u003C\u002Fli>\u003Cli>AWS 把 OpenSearch 放在 Bedrock 路線上。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>不同 AWS 服務怎麼卡位\u003C\u002Fh2>\u003Cp>AWS 給了好幾條路。這很像它一貫的作風，工具很多，選項也很多。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778973838316-30zk.png\" alt=\"AWS 怎麼看向量資料庫\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>\u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fopensearch-service\u002F\" target=\"_blank\" rel=\"noopener\">Amazon OpenSearch Service\u003C\u002Fa> 是 Bedrock 的推薦選項。這代表它最像通用型搜尋底座。\u003C\u002Fp>\u003Cp>如果你已經在用 PostgreSQL，那 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Frds\u002Fpostgresql\u002F\" target=\"_blank\" rel=\"noopener\">Amazon RDS for PostgreSQL\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Faurora\u002F\" target=\"_blank\" rel=\"noopener\">Amazon Aurora PostgreSQL-Compatible Edition\u003C\u002Fa> 的 pgvector 會很順手。\u003C\u002Fp>\u003Cp>如果你要的是低延遲，\u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fmemorydb\u002F\" target=\"_blank\" rel=\"noopener\">Amazon MemoryDB\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fdocumentdb\u002F\" target=\"_blank\" rel=\"noopener\">Amazon DocumentDB\u003C\u002Fa> 也都能做向量搜尋。\u003C\u002Fp>\u003Cp>這種切法很實際。AWS 不是只推一種資料庫，而是看你的資料本來放哪裡。\u003C\u002Fp>\u003Ctable>\u003Cthead>\u003Ctr>\u003Cth>服務\u003C\u002Fth>\u003Cth>適合情境\u003C\u002Fth>\u003Cth>AWS 提到的特徵\u003C\u002Fth>\u003C\u002Ftr>\u003C\u002Fthead>\u003Ctbody>\u003Ctr>\u003Ctd>OpenSearch Service\u003C\u002Ftd>\u003Ctd>Bedrock 檢索、通用搜尋\u003C\u002Ftd>\u003Ctd>可做向量驅動搜尋\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>PostgreSQL \u002F Aurora PostgreSQL\u003C\u002Ftd>\u003Ctd>SQL 資料旁邊放 embeddings\u003C\u002Ftd>\u003Ctd>支援 pgvector\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>MemoryDB\u003C\u002Ftd>\u003Ctd>低延遲、高 QPS\u003C\u002Ftd>\u003Ctd>數百萬向量、毫秒級回應、每秒數萬次查詢\u003C\u002Ftd>\u003C\u002Ftr>\u003Ctr>\u003Ctd>DocumentDB\u003C\u002Ftd>\u003Ctd>文件型工作負載\u003C\u002Ftd>\u003Ctd>毫秒級延遲的向量搜尋\u003C\u002Ftd>\u003C\u002Ftr>\u003C\u002Ftbody>\u003C\u002Ftable>\u003Cul>\u003Cli>OpenSearch Service：適合 Bedrock 檢索。\u003C\u002Fli>\u003Cli>PostgreSQL \u002F Aurora：適合 SQL 與 embeddings 並存。\u003C\u002Fli>\u003Cli>MemoryDB：適合低延遲、高 QPS。\u003C\u002Fli>\u003Cli>DocumentDB：適合文件型工作負載。\u003C\u002Fli>\u003C\u002Ful>\u003Ch2>數字怎麼看，差異在哪裡\u003C\u002Fh2>\u003Cp>這篇文章最有用的地方，是它把不同路線的數字放在一起。\u003C\u002Fp>\u003Cp>OpenSearch Service 偏通用。MemoryDB 偏快。DocumentDB 偏文件型。pgvector 則偏工程實用。\u003C\u002Fp>\u003Cp>你可以把它想成四種取捨。不是誰比較厲害，而是誰比較貼近你的系統。\u003C\u002Fp>\u003Cp>如果資料已經在 PostgreSQL，硬搬到別的引擎，常常只是多一層麻煩。\u003C\u002Fp>\u003Cp>如果你要的是即時查詢和更新，MemoryDB 的方向就更像你要的東西。\u003C\u002Fp>\u003Cul>\u003Cli>OpenSearch Service：AWS 對 Bedrock 的預設建議。\u003C\u002Fli>\u003Cli>MemoryDB：數百萬向量、單位是毫秒、QPS 可到數萬。\u003C\u002Fli>\u003Cli>DocumentDB：向量搜尋加上文件資料模型。\u003C\u002Fli>\u003Cli>pgvector：最像「把 embeddings 放進既有 SQL 系統」。\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這裡的關鍵不是功能名詞，而是維運成本。你多開一套系統，就多一份同步、備份、監控和權限管理。\u003C\u002Fp>\u003Cp>很多團隊最後會選最接近原本資料流的方案。這很無聊，但通常最少踩雷。\u003C\u002Fp>\u003Ch2>產業脈絡：為什麼大家都在做\u003C\u002Fh2>\u003Cp>向量資料庫的熱度，跟生成式 AI 一起起來。原因很簡單，\u003Ca href=\"\u002Ftag\u002Fllm\">LLM\u003C\u002Fa> 需要外部資料，不能只靠模型記憶。\u003C\u002Fp>\u003Cp>你要做企業知識庫、客服助理、產品搜尋，最後都會碰到同一題：怎麼把資料找回來。\u003C\u002Fp>\u003Cp>這就是向量資料庫的角色。它把非結構化資料變成可查詢的資料層。\u003C\u002Fp>\u003Cp>在這之前，很多團隊只能靠關鍵字搜尋。那種做法對精準詞很有效，但對語意理解很弱。\u003C\u002Fp>\u003Cp>現在的做法比較像混搭。先用 embedding 找候選，再用關鍵字、權限、時間戳去縮小範圍。\u003C\u002Fp>\u003Cp>這也解釋了為什麼 AWS、\u003Ca href=\"\u002Ftag\u002Fgoogle\">Google\u003C\u002Fa>、\u003Ca href=\"\u002Ftag\u002Fmicrosoft\">Microsoft\u003C\u002Fa> 都在推相關功能。大家都知道，AI 應用最後還是要回到資料層。\u003C\u002Fp>\u003Cp>你可以把 \u003Ca href=\"https:\u002F\u002Faws.amazon.com\u002Fs3\u002F\" target=\"_blank\" rel=\"noopener\">Amazon S3\u003C\u002Fa>、Aurora、MemoryDB 想成不同的資料落點。向量資料庫只是把檢索能力接上去。\u003C\u002Fp>\u003Ch2>結論：開發者該怎麼選\u003C\u002Fh2>\u003Cp>如果你已經在 AWS 上，先看資料在哪裡，再決定要不要上 OpenSearch Service。這通常比先追新名詞更實際。\u003C\u002Fp>\u003Cp>我覺得最務實的做法是這樣。SQL 資料留在 PostgreSQL，低延遲查詢放 MemoryDB，Bedrock 檢索先看 OpenSearch Service。\u003C\u002Fp>\u003Cp>接下來一兩年，向量搜尋大概會變成 AI 應用的標配。問題只剩下，你要把它放在哪一層。\u003C\u002Fp>\u003Cp>如果你現在正在做 RAG，我會先問一件事：你的 embeddings 是不是已經離資料太遠了？\u003C\u002Fp>","AWS 這篇在講向量資料庫怎麼存 embeddings、怎麼做相似度搜尋，以及為什麼 Bedrock 常搭配 OpenSearch Service。","aws.amazon.com","https:\u002F\u002Faws.amazon.com\u002Fwhat-is\u002Fvector-databases\u002F",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778973842531-shap.png",[13,14,15,16,17,18,19,20],"向量資料庫","embeddings","AWS","OpenSearch Service","Bedrock","RAG","pgvector","MemoryDB","zh",1,false,"2026-05-16T23:23:31.940718+00:00","2026-05-16T23:23:31.853+00:00","done","e37a33e3-06c1-491c-99d0-989f882043e1","vector-databases-aws-explained-zh","tools","92dfddab-f461-42ad-a8e2-ec8016195a70","published",[33,34,35],"向量資料庫存的是 embeddings，核心是用相似度找語意相近的資料。","AWS 把 OpenSearch Service 當 Bedrock 的主要向量搜尋選項。","PostgreSQL、MemoryDB、DocumentDB 都能做向量搜尋，但適合的場景不同。","c3c88dd2-a940-438a-b359-0e5a24562273","[-0.044712044,-0.018886922,0.0051631234,-0.10672095,-0.021448674,0.004949049,-0.003812951,0.009931865,0.012532753,0.026492732,-0.017305776,6.949976e-05,0.008458841,-0.013910614,0.118752785,0.010136494,0.019135522,-0.008252844,-0.0059411526,-0.032086004,-0.01361605,0.010298021,-0.010992237,-0.0002455634,0.014795263,0.0047179395,0.021263791,-0.0068859807,0.034590565,0.017794887,0.021867806,0.017528217,0.028840762,0.047475334,-0.007943158,0.005479513,-0.004893632,-0.0039102044,0.009008493,0.0311418,0.019517288,-0.017086234,-0.000891368,-0.0028487938,0.0024362132,-0.0014490675,0.0035025447,-0.017886652,0.008092694,-0.0065778354,0.020862574,0.01806586,-0.0060079326,-0.1657356,0.014101263,0.009160827,-0.022011172,-0.021915134,0.0030609204,0.0095242765,-0.014140265,0.029596062,-0.011706619,0.009242529,-0.008363496,0.0009266177,0.040096927,-0.002935069,-0.00036803202,-0.0037152574,0.010029836,-0.012332465,0.0042117205,-0.010296754,-0.0044925874,-0.038132783,0.0289582,0.0011182587,0.012254786,0.019148368,0.015855404,-0.003937335,-0.019714106,-0.008543442,-0.009090759,0.010312812,0.012232625,0.00945447,0.0109431995,0.012959404,-0.0063224025,-0.013847772,0.008315972,-0.02061608,0.014654114,-0.012171778,-0.039237842,-0.028491829,0.003280277,-0.0013348839,-0.0020307899,-0.008403746,-0.014296189,0.008473226,0.017680956,0.021168182,0.02454227,0.0022222397,-0.008499492,0.0001974823,0.013011697,-0.032700624,0.0002687867,-0.01029008,-0.003470167,-0.13600183,-0.009032822,0.0032759227,0.02439178,0.019820208,-0.030582462,-0.0056264154,0.01650463,0.003291545,0.0056924173,-0.020166755,-0.014149231,-0.014403168,-0.022037085,0.0047139633,-0.001000106,0.023418382,-0.0030888866,-0.0018195729,0.010387673,0.024719065,-0.017465971,-0.018432096,-0.032916192,-0.013736986,0.0094729895,0.02462212,-0.035136815,0.0063668517,-0.019506682,0.01070857,-0.03134078,0.026530895,-0.00014751618,0.0041290717,0.035594713,-0.014647614,-0.00550153,0.014556461,0.018357871,-0.005399305,0.015696933,0.007800208,0.008928033,0.012417622,0.023206575,0.0033430315,-0.013837261,0.02235183,0.021973595,0.014205611,-0.0096016135,-0.027937999,0.0077750417,0.017640905,-0.028896658,0.015546804,-0.002500437,0.007544978,0.006561019,-0.0277585,-0.0055582584,0.015072297,-0.0006385303,-0.0041568847,0.030587297,0.007715098,-0.013001702,0.01668677,-0.0062492127,0.0044559604,-0.0022404313,0.02226959,0.01710808,0.0055587925,-0.016524283,-0.010746,0.025784835,0.0015055831,-0.01616587,-0.03690826,-0.014638268,0.017944947,-0.0188432,0.02534018,-0.022675725,-0.014063702,0.025082262,-0.02465602,0.02248329,-0.0071046203,-0.022466443,-0.04443988,0.0084982775,-0.0074937413,0.003518157,0.020945448,0.010553135,-0.013961117,0.01072275,-0.015664943,0.0016875638,-0.016703017,-0.013489373,-0.01948732,0.0058369446,-0.006287201,0.038989622,0.01945637,0.015932249,-0.019118795,-0.030813808,-0.028165892,-0.011341293,-0.014650058,-0.013197549,0.023502152,0.014329897,0.014360219,0.011326519,-0.04045739,0.0034365682,-0.0049630445,-0.030318469,0.017039858,-0.005104505,0.021217743,-0.006804113,0.021596966,0.012917793,-0.014035311,-0.0047774673,-0.006096442,-0.0025773984,0.02619051,0.010296915,0.005412925,-0.0069111832,-0.013899584,0.011366527,0.009995513,-0.04428039,0.006043936,0.009550878,0.032717727,-0.007406419,-0.032311812,-0.010812271,-0.023551278,-0.021147128,0.039184485,0.0019174294,0.011077498,-0.00076039997,0.01838169,0.012649713,0.0001472644,0.022634631,-0.013754319,-0.008305444,0.033515617,-0.028374486,0.024647798,0.00056812743,-0.002769751,0.023440963,0.013625826,0.018353123,0.01220008,-0.015819255,0.023252463,-0.015101281,-0.0139103625,0.006162018,0.002956677,0.0083794,0.01010923,-0.031667333,0.00031802364,0.031184096,-0.043086525,0.0015756815,0.01698084,0.017018644,-0.007704734,-0.018181674,0.012110257,0.012282282,0.021154288,0.00087480823,0.016657438,-0.007271177,-0.017157897,0.01014808,0.024993492,-0.00573128,0.0003825023,-0.009266415,0.0054463455,-0.012531512,-0.0064033135,0.019560004,-0.012130252,-0.03911594,0.024056641,-0.00039890961,0.0122838365,-0.016435087,-8.3734405e-05,-0.039415672,-0.0003017551,0.014552813,-0.006451585,0.034198675,-0.018397527,-0.009004109,0.0068392814,-0.0005584592,-0.015756037,-0.01197896,-0.0048241396,-0.0037332305,-0.007922234,-0.013026492,0.02622651,0.009875566,0.0043178606,-0.02476999,0.012169234,-0.042002957,0.034091983,0.015893545,-0.016457867,-0.03584011,-0.020699276,0.008594602,0.026879784,0.017765662,-0.020648006,-0.025432039,-0.00763249,-0.031106781,-0.006673676,0.014061232,-0.00652084,0.0023421792,-0.011640669,0.005971379,0.0032833223,0.00053730735,-0.020623112,0.004921307,0.030213032,-0.014450865,0.0020334856,-0.020839209,0.017667077,0.028748818,0.0045235665,-0.0006346796,-0.00929329,0.0041009793,-0.01798078,-0.026843589,-0.00992827,0.008033356,0.04229076,-0.007177046,-0.017744523,-0.0043719993,-0.017600892,0.018828394,0.01331381,0.0066913334,0.014595923,0.004286484,-0.0037250838,0.0038020283,0.011391107,0.0006705233,-0.027131407,-0.030620767,0.00850789,0.012948033,-0.006324761,-0.034489013,0.0019341463,-0.014777407,-0.0063745296,0.007386735,-0.014353479,0.0025374258,0.0007844773,-0.00411333,-0.010212411,-0.0031806715,-0.028063385,-0.007836891,-0.019985747,0.017012116,0.016877428,-0.0015669105,0.0048763356,0.029377224,-0.015404056,0.0004302087,0.018709641,-0.014146096,-0.01289025,0.008869198,-0.013895475,-0.03568632,-0.011108735,-0.014909447,-0.031638812,0.023473777,-0.016932534,-0.021358965,-0.016938565,-0.031732425,-0.01733949,-0.022512013,0.008880584,-0.01898189,-0.022047095,0.013686613,-0.009394114,-0.0011228204,0.007931122,-0.005927679,0.003821698,0.00019771449,-0.005166465,-0.0144325355,0.006014505,0.036972493,0.0066251857,0.014210662,-0.01616177,0.021346902,-0.026103863,0.009264508,-0.0069021564,-0.024094967,-0.030131565,-0.0101511115,-0.008424955,0.0063469973,0.019495273,-0.00086926244,-0.030600064,-0.0051936773,-0.010799128,-0.034902174,0.03930418,-0.0064316704,0.015747244,-0.007070943,-0.024107348,-0.02102109,-0.006445156,-0.0038808158,0.028752076,0.013095187,-0.010682904,-0.009337848,-0.022392103,-0.0012931497,-0.006409035,0.012156315,0.034902833,0.0073996363,0.02742862,-0.0023752626,-0.013716213,0.0004931211,0.0037427156,0.019774318,0.0006669031,-0.006266956,-0.024386719,0.023045931,-0.03974336,0.021857457,0.021451296,-0.030808201,0.009315088,0.008886651,-0.026748475,0.0012398829,-0.017391674,0.0079927575,-0.016942738,-0.007923185,-0.010223384,-0.010963783,-0.0056362725,0.008850883,0.0057034115,0.007384954,-0.0017280506,-0.0013204422,0.02820078,0.0110893035,0.016855838,-0.0229743,-0.03371564,-0.020490257,0.030325968,-0.010171737,-0.0063957926,0.008098962,-0.005181566,0.025085436,-0.017426847,-0.009277661,0.0330256,-0.009207732,0.0037260312,0.004125533,0.02124999,0.027054967,0.0018911563,0.0045092506,0.010857662,-0.023541495,-0.009099483,-0.0105559835,0.002224333,0.025785316,-0.11146498,0.013284585,0.025159536,-0.00026009817,-0.02868114,-0.026940444,0.0028481386,-0.015940513,0.005337953,0.005054588,-0.0077457693,0.024786424,0.02049657,0.010390447,0.014161377,-0.009892817,-0.04313122,-0.016365228,0.020617865,-0.012306774,0.010781522,-0.01695984,0.006038971,0.015011932,-0.005791304,0.0021210173,0.015554187,0.0071885544,0.015187772,0.010376422,-0.0040799435,-0.031314123,-0.004324366,0.02642828,-0.00038510142,0.0007753209,0.013325765,-0.012069407,0.026347319,0.015375607,0.0028383352,-0.018165551,-0.016651396,-0.021849342,0.022157302,0.0002650105,-0.020197326,0.002580075,-0.01638978,0.008322332,-0.046618454,0.003809864,-0.023626296,-0.0041447976,0.012087432,0.01492328,-0.019912,0.0050265347,-0.009817331,0.01426953,-0.012289422,0.009692843,-0.011261969,0.01311938,-0.0049456437,-0.022771616,0.009350154,0.014341975,0.016203016,0.006762,-0.021991886,-0.0052886107,0.0082730055,0.0017680047,-0.0292331,-0.002409035,0.002462388,0.023697633,-0.0028824606,-0.0117507605,-0.033146624,-0.03638002,-0.104275994,-0.03230105,-0.0062236623,-0.0023273616,0.040185276,0.0027175413,-0.0063711726,-0.012691758,0.008725112,-0.025742708,-0.008459983,-0.01452406,-0.014053777,-0.02586978,-0.019402828,-0.0070020976,0.023538383,-0.01022535,-0.014453191,-0.033303063,-0.0037581273,-0.013142187,0.021211114,0.0073867994,-0.015280432,0.031946715,0.031128118,-0.006045519,-0.012233984,0.006535818,0.00649327,-0.111438386,-0.01195769,0.024472475,0.022659225,-0.003474036,0.0039298213,-0.0054199304,-0.0020394807,-0.006338009,0.015382904,-0.015311049,-0.029088965,-0.039051317,-0.002187996,-0.002232373,0.10037975,-0.012340748,-0.0003721362,0.011932927,0.0012482926,-0.010875914,-0.0052362,0.0039806985,0.011446057,0.0135003785,0.02825366,0.011185631,-0.0041537513,-0.006004382,0.035409383,0.017963525,0.029947009,-0.033393845,0.015299609,-0.005904739,-0.029824361,0.014827316,-0.009504706,0.02434085,0.007931005,0.021346478,0.016020365,0.023331122,-0.008701776,0.011005879,0.016113501,-0.021191072,0.0045245853,-0.018597303,-0.008773471,0.0050513153,-0.077148005,0.010198692,-0.011530974,0.013016689,-0.022875447,-0.0154296495,0.017503403,0.0022925937,-0.0060644755,0.0016249784,-0.0012478448,-0.010647931,0.024384463,-0.018775417,0.0039963517,0.007310098,0.0069505847,-0.0019298015,0.0066007953,-0.025049457,0.012937787,-0.017235348,0.00060640275,-0.030512813,-0.004781839,0.0014144277,0.004044159,0.038671084,-0.019679988,-0.0021733404,0.0019979954,0.0022157894,-0.0043057664,0.025055373,0.0007363758,0.01497741,-0.014769094,-0.005243272,-0.033809453,0.02222549,0.030144721,-0.00025685065,0.018471738,-0.011816307,0.017480532,0.01904409,0.027813222,0.020012032,0.002493989,-0.032203466,-0.022047425,-0.00032462034,-0.034606867,0.015182834,0.025229845,0.007771374,0.021832418,0.02456137,0.0039722165]",{"tags":39,"relatedLang":48,"relatedPosts":52},[40,42,44,45,46],{"name":15,"slug":41},"aws",{"name":17,"slug":43},"bedrock",{"name":14,"slug":14},{"name":13,"slug":13},{"name":16,"slug":47},"opensearch-service",{"id":30,"slug":49,"title":50,"language":51},"vector-databases-aws-explained-en","Vector Databases: How AWS Explains Them","en",[53,59,65,71,77,83],{"id":54,"slug":55,"title":56,"cover_image":57,"image_url":57,"created_at":58,"category":29},"dc666c3f-3dc1-43e3-908b-14d7f4cfdac0","ru-he-yong-cuda-oxide-jian-li-rust-gpu-he-xin-zh","如何用 cuda-oxide 建立 Rust GPU 核心","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778976835472-mbgt.png","2026-05-17T00:13:33.173664+00:00",{"id":60,"slug":61,"title":62,"cover_image":63,"image_url":63,"created_at":64,"category":29},"4a78a285-c9ab-400d-a4ea-ada7303fd327","how-to-choose-a-vector-database-in-2026-zh","2026 向量資料庫怎麼選","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778972632684-qdj2.png","2026-05-16T23:03:29.482997+00:00",{"id":66,"slug":67,"title":68,"cover_image":69,"image_url":69,"created_at":70,"category":29},"a109dac1-43f3-4a6b-982c-13b59e8f61e9","vibe-research-ai-tools-workflows-zh","Vibe Research：用 AI 加速研究流程","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778904653705-zekc.png","2026-05-16T04:10:33.15767+00:00",{"id":72,"slug":73,"title":74,"cover_image":75,"image_url":75,"created_at":76,"category":29},"cb68bb90-3638-4334-87c7-02580f59877a","aws-repository-wide-security-scanner-matters-zh","為什麼 AWS 的全倉庫安全掃描比更快的 SAST 更重要","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778901047875-n4l9.png","2026-05-16T03:10:24.757504+00:00",{"id":78,"slug":79,"title":80,"cover_image":81,"image_url":81,"created_at":82,"category":29},"7c966206-36f7-4d6b-b2e5-088a4732ede4","why-docker-microvm-sandboxes-ai-agents-zh","為什麼 Docker 的 microVM 沙盒才是 AI 代理的正解","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778893837158-gpxf.png","2026-05-16T01:10:20.668094+00:00",{"id":84,"slug":85,"title":86,"cover_image":87,"image_url":87,"created_at":88,"category":29},"d058a76f-6548-4135-8970-f3a97f255446","why-gemini-api-pricing-is-cheaper-than-it-looks-zh","為什麼 Gemini API 定價其實比看起來更便宜","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1778869845081-j4m7.png","2026-05-15T18:30:25.797639+00:00",[90,95,100,105,110,115,120,125,130,135],{"id":91,"slug":92,"title":93,"created_at":94},"de769291-4574-4c46-a76d-772bd99e6ec9","googles-biggest-gemini-launches-in-2026-zh","Google 2026 最大 Gemini 盤點","2026-03-26T07:26:39.21072+00:00",{"id":96,"slug":97,"title":98,"created_at":99},"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":101,"slug":102,"title":103,"created_at":104},"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":106,"slug":107,"title":108,"created_at":109},"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":111,"slug":112,"title":113,"created_at":114},"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":116,"slug":117,"title":118,"created_at":119},"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":121,"slug":122,"title":123,"created_at":124},"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":126,"slug":127,"title":128,"created_at":129},"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":131,"slug":132,"title":133,"created_at":134},"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":136,"slug":137,"title":138,"created_at":139},"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"]