https://whynoideas.dev/blog/ 香港本地 AI 部署實戰：律師行、診所、會計、中小企，粵語內容。 zh-hk Sun, 01 Nov 2026 01:33:17 +0000 https://whynoideas.dev/blog/moneylab-vs-whynoideas-truth.html https://whynoideas.dev/blog/moneylab-vs-whynoideas-truth.html Sat, 31 Oct 2026 16:00:00 +0000 Moneylab 公開佢用 AI agent 做 7 種 side-income 方法嘅 ledger，179 日只得 US$10.50 收入 — 但佢公開更正自己假數據嘅做法，係最好嘅信任建立策略。whynoideas 對照：0 sales、0 email subs、AdSense pending，誠實係唯一出路。 https://whynoideas.dev/blog/2026-11-01-video.html https://whynoideas.dev/blog/2026-11-01-video.html Sat, 31 Oct 2026 16:00:00 +0000 微調同推理硬件需求大不同。本文拆解顯存要求、QLORA實測、環境設定及算力瓶頸，幫你避開買錯機嘅坑。 https://whynoideas.dev/blog/2026-11-01-sdk.html https://whynoideas.dev/blog/2026-11-01-sdk.html Sat, 31 Oct 2026 16:00:00 +0000 唔止係陪聊天，教你用Anthropic官方SDK建立能執行動作嘅AI Agent，處理複雜流程。 https://whynoideas.dev/blog/2026-11-01-rag-flow-ai.html https://whynoideas.dev/blog/2026-11-01-rag-flow-ai.html Sat, 31 Oct 2026 16:00:00 +0000 唔想文件上傳雲端？RAG Flow幫你喺本地建立安全問答系統。詳解硬體要求、Docker部署步驟及模型選擇，兼顧私隱與效率。 https://whynoideas.dev/blog/2026-11-01-ollama-ai-windows.html https://whynoideas.dev/blog/2026-11-01-ollama-ai-windows.html Sat, 31 Oct 2026 16:00:00 +0000 唔使月費、唔使上网！只要你有NVIDIA或AMD顯示卡，即可在Windows本地運行AI。教你由零安裝Ollama，選擇合適模型，避開VRAM陷阱，享受私密智能體驗。 https://whynoideas.dev/blog/2026-11-01-nvidia-dgx-spark.html https://whynoideas.dev/blog/2026-11-01-nvidia-dgx-spark.html Sat, 31 Oct 2026 16:00:00 +0000 唔使機房養伺服器，DGX Spark 將數據中心算力縮細放書枱。解析 GB10 晶片、統一記憶體架構及實際跑大模型表現。 https://whynoideas.dev/blog/2026-11-01-n8n-line.html https://whynoideas.dev/blog/2026-11-01-n8n-line.html Sat, 31 Oct 2026 16:00:00 +0000 教你用n8n內建LINE節點，無需寫code，設定Webhook自動分派查詢，省錢又穩定，適合預算有限嘅小本經營。 https://whynoideas.dev/blog/2026-11-01-microsoft-365-copilot-email.html https://whynoideas.dev/blog/2026-11-01-microsoft-365-copilot-email.html Sat, 31 Oct 2026 16:00:00 +0000 每日開工 Inbox 塞爆？Copilot 自動整理郵件線程、建議回覆，幫你找回時間做重要事，務實提升效率。 https://whynoideas.dev/blog/2026-11-01-mfit.html https://whynoideas.dev/blog/2026-11-01-mfit.html Sat, 31 Oct 2026 16:00:00 +0000 記憶體不足唔代表玩唔到 AI。透過 Mfit 掃描硬件同簡單公式，幫你精準篩選適合本地運行嘅模型，唔使盲目追求最大，兼顧速度與穩定性。 https://whynoideas.dev/blog/2026-11-01-meta-whatsapp-cloud-api.html https://whynoideas.dev/blog/2026-11-01-meta-whatsapp-cloud-api.html Sat, 31 Oct 2026 16:00:00 +0000 告别灰色API封號風險，改用Meta官方Cloud API。免費開戶、穩定自動回覆，手把手教你由零建立智能客服系統。 https://whynoideas.dev/blog/2026-11-01-m5-pro-m6-mac-mini-ai.html https://whynoideas.dev/blog/2026-11-01-m5-pro-m6-mac-mini-ai.html Sat, 31 Oct 2026 16:00:00 +0000 2026款Mac mini M5 Pro與M6邊個先至跑得動大模型？拆解頻寬、記憶體同MLX、Ollama設定，幫你揀啱用嘅AI神器。 https://whynoideas.dev/blog/2026-11-01-gbnf-ai-json.html https://whynoideas.dev/blog/2026-11-01-gbnf-ai-json.html Sat, 31 Oct 2026 16:00:00 +0000 解決本地模型輸出格式混亂問題，利用Llama.cpp配合GBNF語法，強制模型嚴格遵循JSON結構，提升自動化腳本穩定性。 https://whynoideas.dev/blog/2026-11-01-dgx-spark-gb10-ai.html https://whynoideas.dev/blog/2026-11-01-dgx-spark-gb10-ai.html Sat, 31 Oct 2026 16:00:00 +0000 NVIDIA DGX Spark GB10 登場，128GB 統一記憶體打造桌面 AI 工作站。本文詳解規格、支援模型、安裝步驟及實測速度。 https://whynoideas.dev/blog/2026-11-01-cloudflare-workers-ai.html https://whynoideas.dev/blog/2026-11-01-cloudflare-workers-ai.html Sat, 31 Oct 2026 16:00:00 +0000 唔使養Server，用官方免費額度零成本起步。專為非技術人員設計，簡單步驟即可部署自動化回應。 https://whynoideas.dev/blog/2026-11-01-chatgpt.html https://whynoideas.dev/blog/2026-11-01-chatgpt.html Sat, 31 Oct 2026 16:00:00 +0000 怕客戶資料洩漏被禁AI？採用兩難分類法，敏感資料本地處理，創意工作雲端進行。實測Ollama部署本地模型，平衡合規與效率。 https://whynoideas.dev/blog/2026-11-01-browser-use-ai.html https://whynoideas.dev/blog/2026-11-01-browser-use-ai.html Sat, 31 Oct 2026 16:00:00 +0000 想叫 AI 幫你填表搜料？介紹開源工具 Browser Use，教你用自然語言自動化網頁操作，避開常見陷阱，提升工作效率。 https://whynoideas.dev/blog/2026-11-01-ai.html https://whynoideas.dev/blog/2026-11-01-ai.html Sat, 31 Oct 2026 16:00:00 +0000 買電腦跑唔郁大模型？本文拆解容量、頻寬、量化同生態四步決策法，教你避開常見陷阱，確保每一分錢都花在刀口上。 https://whynoideas.dev/blog/2026-11-01-ai-vllm-ollama-llama-cpp.html https://whynoideas.dev/blog/2026-11-01-ai-vllm-ollama-llama-cpp.html Sat, 31 Oct 2026 16:00:00 +0000 想喺自己部機跑AI但怕麻煩？拆解VLLM、Ollama同llama.cpp三大引擎，睇邊個最適合你嘅需求同硬件配置。 https://whynoideas.dev/blog/2026-11-01-ai-skill-md.html https://whynoideas.dev/blog/2026-11-01-ai-skill-md.html Sat, 31 Oct 2026 16:00:00 +0000 唔使每次重覆Prompt！透過Skill.md將工作流程寫入AI，支持按需加載，簡化操作，提升效率。 https://whynoideas.dev/blog/2026-11-01-ai-nvidia-spark.html https://whynoideas.dev/blog/2026-11-01-ai-nvidia-spark.html Sat, 31 Oct 2026 16:00:00 +0000 唔使再估！實測不同硬件跑AI模型嘅每秒字數，教你點樣揀啱配置，避免等等到心碎。 https://whynoideas.dev/blog/2026-11-01-ai-mixtral-small.html https://whynoideas.dev/blog/2026-11-01-ai-mixtral-small.html Sat, 31 Oct 2026 16:00:00 +0000 唔想電腦變磚頭？教你分辨Mixtral同Small 3.1嘅記憶體需求，用Ollama一行指令本地部署，私隱有保障。 https://whynoideas.dev/blog/2026-11-01-ai-hugging-face.html https://whynoideas.dev/blog/2026-11-01-ai-hugging-face.html Sat, 31 Oct 2026 16:00:00 +0000 網絡差或者斷網時，AI模型總卡在下載？透過設定HF_HUB_OFFLINE同HF_HOME兩個環境變數，即可強制使用本地緩存，實現穩定離線部署。 https://whynoideas.dev/blog/2026-11-01-ai-cloudflare-workers.html https://whynoideas.dev/blog/2026-11-01-ai-cloudflare-workers.html Sat, 31 Oct 2026 16:00:00 +0000 唔使養伺服器，用Cloudflare Workers AI免費起步，快速部署智能客服，新手亦易上手。 https://whynoideas.dev/blog/2026-11-01-ai-agent.html https://whynoideas.dev/blog/2026-11-01-ai-agent.html Sat, 31 Oct 2026 16:00:00 +0000 實戰捕捉 AI Agent 三大盲點：TTS 假成功、內容複製貼上、記憶錯亂。睇點樣加守門機制解決，唔使再靠運氣搵問題。 https://whynoideas.dev/blog/2026-11-01-ai-agent-n8n.html https://whynoideas.dev/blog/2026-11-01-ai-agent-n8n.html Sat, 31 Oct 2026 16:00:00 +0000 唔使寫Code，用n8n Agent Builder建立智能助手。支援Self-hosting、混合模型及Human-in-the-loop，告別重複工作，提升自動化效率。 https://whynoideas.dev/blog/2026-11-01-ai-3.html https://whynoideas.dev/blog/2026-11-01-ai-3.html Sat, 31 Oct 2026 16:00:00 +0000 系統沒報錯不代表功能正常。本文拆解配音消失、文章重複及上下文污染三大隱蔽故障，教你用斷言驗證內容有效性，避免被狀態碼200誤導。 https://whynoideas.dev/blog/2026-11-01-70b.html https://whynoideas.dev/blog/2026-11-01-70b.html Sat, 31 Oct 2026 16:00:00 +0000 想喺屋企玩Llama 3.3 70B？本文詳解42.5GB檔案對記憶體嘅要求，比較Mac同NVIDIA配置，並提供避坑貼士同安裝指令。 https://whynoideas.dev/blog/gb10-111gb-qwen38-benchmark.html https://whynoideas.dev/blog/gb10-111gb-qwen38-benchmark.html Fri, 30 Oct 2026 19:20:00 +0000 一部 NVIDIA GB10（128GB 統一記憶體）跑 Qwen3.8-Flash-Next UD-Q4_K_XL 的真實數據：111GB / 4 shard、生成 22 tok/s、prompt 520 tok/s、冷載入 3 分 24 秒，含 GGML_RPC_NO_RDMA=1 踩坑修正。 https://whynoideas.dev/blog/why-local-ai-not-cloud.html https://whynoideas.dev/blog/why-local-ai-not-cloud.html Fri, 30 Oct 2026 16:00:00 +0000 比較本地同雲端 AI 喺私隱、成本、性能、合規四個維度嘅分別，幫律師行決定本地部署定係用 API 方案更適合。 https://whynoideas.dev/blog/rogue-ai-hk-sme-impact.html https://whynoideas.dev/blog/rogue-ai-hk-sme-impact.html Fri, 30 Oct 2026 16:00:00 +0000 FTC 調查 OpenAI 同 Anthropic 的 rogue AI agents 問題，對香港中小企有咩啟示？比較金管局 genAI sandbox，分析本地部署 vs 雲端 API 嘅安全隱患。