[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"article-awesome-open-source-ai-projects-list-zh":3,"article-related-awesome-open-source-ai-projects-list-zh":30,"series-tools-feb9176d-89c6-4bd0-a82a-8440625d8c94":86},{"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":11,"views":27,"created_at":28,"published_at":29,"topic_cluster_id":11},"feb9176d-89c6-4bd0-a82a-8440625d8c94","awesome-open-source-ai-projects-list-zh","開源 AI 專案清單怎麼挑","\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Falvinreal\u002Fawesome-opensource-ai\" target=\"_blank\" rel=\"noopener\">awesome-opensource-ai\u003C\u002Fa> 這個 GitHub repo，很像一份開源 AI 採購單。它有 2,486 顆星、219 次 fork，還是 Python 專案。重點不是收很多連結，而是挑能上線的工具。\u003C\u002Fp>\u003Cp>說白了，現在 AI 工具太多了。真正難的不是找得到，而是選得對。這份清單把模型、推理、RAG、代理、評測、訓練、ML\u003Ca href=\"\u002Fnews\u002Fopenai-revenue-valuation-funding-2026-zh\">Op\u003C\u002Fa>s 分開整理，對台灣工程師很實用。\u003C\u002Fp>\u003Ch2>這份清單到底在挑什麼\u003C\u002Fh2>\u003Cp>README 寫得很直白。只有「battle-tested」和「production-prov\u003Ca href=\"\u002Fnews\u002Fagents-radar-ai-digest-10-sources-zh\">en\u003C\u002Fa>」的專案會進榜。這句話聽起來很像行銷，但看分類就知道它不是亂吹。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775999036470-b4zr.png\" alt=\"開源 AI 專案清單怎麼挑\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>它把 AI 系統拆成幾塊。像是核心框架、基礎模型、推理引擎、代理系統、檢索、媒體生成、訓練、MLOps、benchmark 和安全工具。這種切法很務實。\u003C\u002Fp>\u003Cp>很多目錄站只是連結堆疊。這份清單比較像給工程團隊看的地圖。你要做產品、部署服務、還要顧成本，它都幫你先分好類。\u003C\u002Fp>\u003Cul>\u003Cli>GitHub 目前有 2,486 顆星\u003C\u002Fli>\u003Cli>有 219 次 fork\u003C\u002Fli>\u003Cli>主專案是 Python\u003C\u002Fli>\u003Cli>內容分成 14 個區塊\u003C\u002Fli>\u003C\u002Ful>\u003Cp>第一個區塊就能看到熟面孔。像是 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fpytorch\u002Fpytorch\" target=\"_blank\" rel=\"noopener\">PyTorch\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ftensorflow\u002Ftensorflow\" target=\"_blank\" rel=\"noopener\">TensorFlow\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fjax-ml\u002Fjax\" target=\"_blank\" rel=\"noopener\">JAX\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fgoogle\u002Fflax\" target=\"_blank\" rel=\"noopener\">Flax\u003C\u002Fa>。但它也放了 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ftinygrad\u002Ftinygrad\" target=\"_blank\" rel=\"noopener\">tinygrad\u003C\u002Fa> 這種小而怪的專案。\u003C\u002Fp>\u003Cp>我覺得這種搭配很有意思。大框架負責生產力，小專案負責理解底層。兩種都需要，少一個都會踩雷。\u003C\u002Fp>\u003Ch2>它反映的是整個 AI stack\u003C\u002Fh2>\u003Cp>開源 AI 早就不是單一類別了。現在它是一整層 stack。這份 repo 把訓練、推理、檢索、代理和上線工具拆開，才方便比較。\u003C\u002Fp>\u003Cp>像推理區塊會看到 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm\" target=\"_blank\" rel=\"noopener\">vLLM\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftext-generation-inference\" target=\"_blank\" rel=\"noopener\">Text Generation Inference\u003C\u002Fa>。這兩個名字，很多團隊都拿來比延遲、吞吐量和記憶體用量。\u003C\u002Fp>\u003Cp>代理區塊則有 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flanggraph\" target=\"_blank\" rel=\"noopener\">LangGraph\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen\" target=\"_blank\" rel=\"noopener\">AutoGen\u003C\u002Fa>、\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FcrewAIInc\u002FcrewAI\" target=\"_blank\" rel=\"noopener\">CrewAI\u003C\u002Fa>。這些工具都在解同一題：怎麼把多次 LLM 呼叫串起來，還不把流程寫成一團亂。\u003C\u002Fp>\u003Cblockquote>“The future of AI is not about one model. It’s about systems.” — Andrew Ng\u003C\u002Fblockquote>\u003Cp>這句話很適合拿來看這份清單。因為它關心的不是單一模型多強，而是整個系統怎麼組。\u003C\u002Fp>\u003Cp>講白了就是，模型只是零件。真正決定產品能不能活下來的，是外面那一圈工具。\u003C\u002Fp>\u003Ch2>數字會告訴你誰比較成熟\u003C\u002Fh2>\u003Cp>這份清單厲害的地方，在於它不只列名字。很多條目還會放下載量、使用量或生態規模。這對選型很重要。\u003C\u002Fp>\n\u003Cfigure class=\"my-6\">\u003Cimg src=\"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775999037516-l64y.png\" alt=\"開源 AI 專案清單怎麼挑\" class=\"rounded-xl w-full\" loading=\"lazy\" \u002F>\u003C\u002Ffigure>\n\u003Cp>AI 世界很容易被 README 騙。頁面寫得漂亮，不代表能扛正式流量。數字至少能先幫你過濾一輪。\u003C\u002Fp>\u003Cp>幾個例子很有代表性。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers\" target=\"_blank\" rel=\"noopener\">Transformers\u003C\u002Fa> 有超過 100 萬個模型，還有每天 25 萬次以上下載。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FPaddlePaddle\u002FPaddle\" target=\"_blank\" rel=\"noopener\">PaddlePaddle\u003C\u002Fa> 宣稱支援 2,300 萬以上開發者與 76 萬家企業。\u003C\u002Fp>\u003Cul>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftransformers\" target=\"_blank\" rel=\"noopener\">Transformers\u003C\u002Fa>：模型生態最大宗之一\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FUKPLab\u002Fsentence-transformers\" target=\"_blank\" rel=\"noopener\">sentence-transformers\u003C\u002Fa>：做 embedding 很常見\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fpola-rs\u002Fpolars\" target=\"_blank\" rel=\"noopener\">Polars\u003C\u002Fa>：資料量一大就很有感\u003C\u002Fli>\u003Cli>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Ftracel-ai\u002Fburn\" target=\"_blank\" rel=\"noopener\">Burn\u003C\u002Fa> 和 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Fcandle\" target=\"_blank\" rel=\"noopener\">Candle\u003C\u002Fa>：Rust 派的代表\u003C\u002Fli>\u003C\u002Ful>\u003Cp>這些數字不是拿來炫耀而已。它們代表文件、社群、 bug 回報、範例數量，通常都比較完整。你\u003Ca href=\"\u002Fnews\u002Flinux-finally-dropping-i486-support-zh\">真的要\u003C\u002Fa>上線，這些都很重要。\u003C\u002Fp>\u003Cp>而且這份清單也提醒一件事。Python 不是唯一答案。Rust、Julia、甚至更偏底層的實作，都在開源 AI 裡佔一席之地。\u003C\u002Fp>\u003Ch2>競品比一比，差很多\u003C\u002Fh2>\u003Cp>如果你只看名稱，會以為這些工具差不多。其實差很大。推理引擎、代理框架、資料處理工具，各自解的問題都不同。\u003C\u002Fp>\u003Cp>拿推理來說，\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fvllm-project\u002Fvllm\" target=\"_blank\" rel=\"noopener\">vLLM\u003C\u002Fa> 主打高吞吐。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fhuggingface\u002Ftext-generation-inference\" target=\"_blank\" rel=\"noopener\">TGI\u003C\u002Fa> 則和 Hugging Face 生態綁得更緊。你如果已經大量用 Transformers，TGI 會比較順手。\u003C\u002Fp>\u003Cp>代理框架也一樣。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Flangchain-ai\u002Flanggraph\" target=\"_blank\" rel=\"noopener\">LangGraph\u003C\u002Fa> 偏流程編排。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fmicrosoft\u002Fautogen\" target=\"_blank\" rel=\"noopener\">AutoGen\u003C\u002Fa> 偏多代理對話。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002FcrewAIInc\u002FcrewAI\" target=\"_blank\" rel=\"noopener\">CrewAI\u003C\u002Fa> 則主打角色分工。你要的是哪一種工作流，答案會完全不同。\u003C\u002Fp>\u003Cul>\u003Cli>vLLM：適合看吞吐與延遲\u003C\u002Fli>\u003Cli>TGI：適合 Hugging Face 使用者\u003C\u002Fli>\u003Cli>LangGraph：適合複雜流程\u003C\u002Fli>\u003Cli>AutoGen：適合多代理互動\u003C\u002Fli>\u003Cli>CrewAI：適合任務分工型設計\u003C\u002Fli>\u003C\u002Ful>\u003Cp>資料處理也有差。\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fpola-rs\u002Fpolars\" target=\"_blank\" rel=\"noopener\">Polars\u003C\u002Fa> 在大資料集上常比傳統 pandas 更俐落。這不是情懷問題，是效能問題。\u003C\u002Fp>\u003Cp>我會建議團隊先問三件事。你的瓶頸是訓練、推理，還是資料管線。你要的是單機工具，還是要扛伺服器流量。你能接受多少維護成本。\u003C\u002Fp>\u003Ch2>這類清單為什麼現在更重要\u003C\u002Fh2>\u003Cp>以前大家找 AI 工具，常常先看 demo。現在不行了。很多團隊已經進到實作階段，開始在意延遲、成本、可觀測性和部署方式。\u003C\u002Fp>\u003Cp>這也是為什麼 curated list 很有價值。有人先幫你把噪音過濾掉，你就不用每個 repo 都 clone 一次。這對時間很省。\u003C\u002Fp>\u003Cp>放到台灣的情境也很現實。很多新創和企業都在做內部知識庫、客服助理、文件摘要、資料搜尋。這些場景最怕選錯工具，然後後面整套重寫。\u003C\u002Fp>\u003Cp>開源 AI 的生態也越來越像分工市場。模型、推理、檢索、評測、監控，各有專門工具。你很少再看到一個 repo 想包山包海，還能每個都做好。\u003C\u002Fp>\u003Cp>所以這份清單不是拿來收藏而已。它更像一份選型起點。你可以先看分類，再去比 benchmark、文件、社群活躍度和 issue 處理速度。\u003C\u002Fp>\u003Ch2>我會怎麼用這份清單\u003C\u002Fh2>\u003Cp>如果你是工程師，我會先從三個區塊下手。第一個看模型與訓練。第二個看推理。第三個看資料和 RAG。\u003C\u002Fp>\u003Cp>如果你是產品或創業團隊，就先看能不能快速上線。能不能接 API。能不能把成本壓住。這些問題比「哪個最潮」重要太多。\u003C\u002Fp>\u003Cp>我的預測很直接。接下來 12 個月，大家會更少問「哪個模型最強」。大家會更常問「哪個 stack 最穩」。如果你現在就要開始選，先從 \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Falvinreal\u002Fawesome-opensource-ai\" target=\"_blank\" rel=\"noopener\">awesome-opensource-ai\u003C\u002Fa> 這種清單下手，會比亂搜 GitHub 省很多時間。\u003C\u002Fp>\u003Cp>你也可以先挑一個場景。是要做 RAG，還是做模型服務。先把問題定義清楚，再去看工具，會準很多。\u003C\u002Fp>","這份 GitHub 清單收錄可直接上線的開源 AI 專案，從 PyTorch 到 vLLM 都有，2,486 顆星，適合想找模型、推理、RAG 和代理工具的工程師。","github.com","https:\u002F\u002Fgithub.com\u002Falvinreal\u002Fawesome-opensource-ai",null,"https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1775999036470-b4zr.png","tools","zh","00a0853d-92b0-45e5-bfcd-97d7f77ec8a0",[17,18,19,20,21,22,23,24,25,26],"開源 AI","GitHub","PyTorch","vLLM","Transformers","RAG","代理框架","MLOps","推理引擎","台灣開發者",8,"2026-04-12T13:03:35.795784+00:00","2026-04-12T13:03:35.58+00:00",{"tags":31,"relatedLang":45,"relatedPosts":49},[32,33,35,37,39,41,42,44],{"name":23,"slug":23},{"name":22,"slug":34},"rag",{"name":18,"slug":36},"github",{"name":20,"slug":38},"vllm",{"name":24,"slug":40},"mlops",{"name":25,"slug":25},{"name":19,"slug":43},"pytorch",{"name":26,"slug":26},{"id":15,"slug":46,"title":47,"language":48},"awesome-open-source-ai-projects-list-en","Awesome Open Source AI: the best projects list","en",[50,56,62,68,74,80],{"id":51,"slug":52,"title":53,"cover_image":54,"image_url":54,"created_at":55,"category":13},"8520cd4f-2531-4808-a95d-26f590239d7a","500-ai-agent-projects-show-where-agents-work-now-zh","500 個 AI agent 專案，現在能做什麼","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781033591132-c0nh.png","2026-06-09T19:32:37.03924+00:00",{"id":57,"slug":58,"title":59,"cover_image":60,"image_url":60,"created_at":61,"category":13},"c557ef1c-7fde-4c86-918e-4fb9680ee9df","chocolatey-go-package-policy-installs-zh","Chocolatey 的 Go 安裝變成政策","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781029110289-xkbh.png","2026-06-09T18:18:05.078435+00:00",{"id":63,"slug":64,"title":65,"cover_image":66,"image_url":66,"created_at":67,"category":13},"90b2df54-df6e-417d-9e16-91e9ad2f53d7","go-support-policy-turns-releases-into-a-checklist-zh","Go 支援政策把發版變清單","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781028200122-3m4u.png","2026-06-09T18:02:49.50176+00:00",{"id":69,"slug":70,"title":71,"cover_image":72,"image_url":72,"created_at":73,"category":13},"119c23c6-8ae7-4c4e-820e-1eba0730d702","rustdesk-self-hosting-secure-remote-access-zh","RustDesk 自架遠端存取部署指南","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781017373324-g7et.png","2026-06-09T15:02:24.118819+00:00",{"id":75,"slug":76,"title":77,"cover_image":78,"image_url":78,"created_at":79,"category":13},"b84491ba-e4af-4581-8c04-1890df39a1ad","aider-open-source-coding-agent-repo-edits-zh","Aider 讓開源編碼變成 repo 編輯","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781013817997-l4il.png","2026-06-09T14:02:56.179093+00:00",{"id":81,"slug":82,"title":83,"cover_image":84,"image_url":84,"created_at":85,"category":13},"b6bc009f-238c-4466-b7ec-c7085c7fdbe8","wwdc-2026-rumors-siri-assistant-ios-27-zh","WWDC 2026 讓 Siri 變助手","https:\u002F\u002Fxxdpdyhzhpamafnrdkyq.supabase.co\u002Fstorage\u002Fv1\u002Fobject\u002Fpublic\u002Fcovers\u002Finline-1781007517876-bmuu.png","2026-06-09T12:18:03.608802+00:00",[87,92,97,102,107,112,117,122,127,132],{"id":88,"slug":89,"title":90,"created_at":91},"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":93,"slug":94,"title":95,"created_at":96},"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":98,"slug":99,"title":100,"created_at":101},"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":103,"slug":104,"title":105,"created_at":106},"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":108,"slug":109,"title":110,"created_at":111},"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":113,"slug":114,"title":115,"created_at":116},"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":118,"slug":119,"title":120,"created_at":121},"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":123,"slug":124,"title":125,"created_at":126},"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":128,"slug":129,"title":130,"created_at":131},"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":133,"slug":134,"title":135,"created_at":136},"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"]