tag:blogger.com,1999:blog-48774873207817679522026-06-13T22:25:50.910+08:00小狐狸事務所小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.comBlogger6345125tag:blogger.com,1999:blog-4877487320781767952.post-28249276163245450932026-06-13T17:41:26.850+08:002026-06-13T17:42:05.792+08:00Mapleboard MP510-50 測試 (四十四) : 設定 PPPoE 自動撥接上網<div>上周可能因為雷雨使市電瞬斷, 鄉下老家的 Mapleboard 主機應該有重開機, 它是用網路線直接連線光世代數據機, 以撥接方式連線上網, 印象中當時設定 pppoe 時有設開機自動連線, 不知為何這次失靈. 早上花了一點時間才重新設定好, 以下紀錄這次查修過程. 之前的設定參考 :</div><div><br /></div><div>#&nbsp;<a href="https://yhhuang1966.blogspot.com/2023/02/mapleboard-mp510-50-nmcli-pppoe.html" target="_blank">Mapleboard MP510-50 測試 (十) : 用 nmcli 設定 PPPoE 撥接上網</a></div><div><br /></div><div>本系列全部文章參考 :</div><div><br /></div><div>#&nbsp;<a href="https://yhhuang1966.blogspot.com/2023/02/mapleboard-mp510-50.html" target="_blank">Mapleboard MP510-50 測試筆記索引</a></div><div><br /></div><div>首先檢查網路連線狀態 :&nbsp;</div><div><br /></div><div><div>tony1966@LX2438:~$ <b><span style="color: #2b00fe;">ifconfig</span></b>&nbsp;&nbsp;</div><div>eth0: flags=4163&lt;UP,BROADCAST,RUNNING,MULTICAST&gt;&nbsp; mtu 1500</div><div>&nbsp; &nbsp; &nbsp; &nbsp; inet 192.168.1.102&nbsp; netmask 255.255.255.0&nbsp; broadcast 192.168.1.255</div><div>&nbsp; &nbsp; &nbsp; &nbsp; inet6 fe80::66c7:6bea:e03e:ef60&nbsp; prefixlen 64&nbsp; scopeid 0x20&lt;link&gt;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; ether 16:72:2e:da:e7:3c&nbsp; txqueuelen 1000&nbsp; (Ethernet)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; RX packets 12&nbsp; bytes 1961 (1.9 KB)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; RX errors 0&nbsp; dropped 0&nbsp; overruns 0&nbsp; frame 0</div><div>&nbsp; &nbsp; &nbsp; &nbsp; TX packets 192&nbsp; bytes 19933 (19.9 KB)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; TX errors 0&nbsp; dropped 0 overruns 0&nbsp; carrier 0&nbsp; collisions 0</div><div>&nbsp; &nbsp; &nbsp; &nbsp; device interrupt 14&nbsp;&nbsp;</div><div><br /></div><div>lo: flags=73&lt;UP,LOOPBACK,RUNNING&gt;&nbsp; mtu 65536</div><div>&nbsp; &nbsp; &nbsp; &nbsp; inet 127.0.0.1&nbsp; netmask 255.0.0.0</div><div>&nbsp; &nbsp; &nbsp; &nbsp; inet6 ::1&nbsp; prefixlen 128&nbsp; scopeid 0x10&lt;host&gt;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; loop&nbsp; txqueuelen 1000&nbsp; (Local Loopback)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; RX packets 356&nbsp; bytes 27782 (27.7 KB)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; RX errors 0&nbsp; dropped 0&nbsp; overruns 0&nbsp; frame 0</div><div>&nbsp; &nbsp; &nbsp; &nbsp; TX packets 356&nbsp; bytes 27782 (27.7 KB)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; TX errors 0&nbsp; dropped 0 overruns 0&nbsp; carrier 0&nbsp; collisions 0</div></div><div><br /></div><div>沒有出現&nbsp;ppp0, 表示並未撥接上網, 可能是因為非預期斷電導致 NetworkManager 服務毀損, 設定檔異常, 或是原本設定的 autoconnect (自動連線) 屬性跑掉, 先重啟 NetworkManager 服務 :&nbsp;</div><div><br /></div><div><div>tony1966@LX2438:~$ <b><span style="color: #2b00fe;">sudo systemctl restart NetworkManager</span></b>&nbsp; &nbsp;</div><div><br /></div><div>檢查服務狀態, 確認 NetworkManager 顯示為 active (running) 狀態 :</div><div><br /></div><div>tony1966@LX2438:~$ <b><span style="color: #2b00fe;">sudo systemctl status NetworkManager</span></b>&nbsp; &nbsp;</div><div>● NetworkManager.service - Network Manager</div><div>&nbsp; &nbsp; &nbsp;Loaded: loaded (/lib/systemd/system/NetworkManager.service; enabled; vendor preset: enabled)</div><div>&nbsp; &nbsp; &nbsp;Active: <span style="background-color: #fcff01;">active (running)</span> since Thu 2026-06-11 14:34:45 CST; 15s ago</div><div>&nbsp; &nbsp; &nbsp; &nbsp;Docs: man:NetworkManager(8)</div><div>&nbsp; &nbsp;Main PID: 33100 (NetworkManager)</div><div>&nbsp; &nbsp; &nbsp; Tasks: 4 (limit: 4213)</div><div>&nbsp; &nbsp; &nbsp;Memory: 3.4M</div><div>&nbsp; &nbsp; &nbsp; &nbsp; CPU: 295ms</div><div>&nbsp; &nbsp; &nbsp;CGroup: /system.slice/NetworkManager.service</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;└─33100 /usr/sbin/NetworkManager --no-daemon</div><div><br /></div></div><div>用 NetworkManager 的指令工具 nmcli 指令來檢視之前建立的 PPPoE 連線 :&nbsp;</div><div><br /></div><div><div>tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection show</span></b>&nbsp; &nbsp;</div><div>NAME&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;UUID&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; TYPE&nbsp; &nbsp; &nbsp; DEVICE&nbsp;</div><div>Ifupdown (eth0)&nbsp; 681b428f-beaf-8932-dce4-687ed5bae28e&nbsp; ethernet&nbsp; eth0&nbsp; &nbsp;</div><div>EDIMAX-tony&nbsp; &nbsp; &nbsp; ddebdded-8805-4631-aa85-f08c412404bf&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;</div><div><span style="background-color: #fcff01;">hinet</span>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; a857fb4d-b2c2-43e2-a43f-a30a30def74f&nbsp; pppoe&nbsp; &nbsp; &nbsp;--&nbsp; &nbsp; &nbsp;</div><div><span style="background-color: #fcff01;">hinet</span>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 62250a6d-dc03-44b9-be71-7179ce1ec6c9&nbsp; pppoe&nbsp; &nbsp; &nbsp;--&nbsp; &nbsp; &nbsp;</div><div>TonyNote8&nbsp; &nbsp; &nbsp; &nbsp; 37db6693-372b-4190-a4d0-c71aef40bc38&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;</div><div><br /></div><div>其中 hinet 就是之前用來撥接上網的 PPPoE 連線,&nbsp;這兩個連線的名稱雖然一樣, 但它們的 UUID (唯一識別碼) 完全不同,&nbsp;這很有可能就是無法自動撥接上網的元兇, 因為當系統開機準備自動連線時, 它看到兩個都叫 hinet 的設定檔, NetworkManager 可能會不知道該用哪一個, 或者其中一個是壞掉的舊設定, 系統卻偏偏去讀到壞的那一個導致自動撥接失敗.&nbsp;</div><div><br /></div><div>由於不知哪個 UUID 才是可正常連線的設定, 所以最乾淨的做法是把這兩個同名為 hinet 的連線設定都刪除, 重新建立一個乾淨的連線設定 :&nbsp;</div><div><br /></div><div>ony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection delete uuid a857fb4d-b2c2-43e2-a43f-a30a30def74f</span></b>&nbsp; &nbsp;</div></div><div>連線「hinet」 (a857fb4d-b2c2-43e2-a43f-a30a30def74f) 已成功刪除。</div><div><br />tony1966@LX2438:~$ nmcli connection show&nbsp;<br />NAME&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;UUID&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; TYPE&nbsp; &nbsp; &nbsp; DEVICE&nbsp;<br />Ifupdown (eth0)&nbsp; 681b428f-beaf-8932-dce4-687ed5bae28e&nbsp; ethernet&nbsp; eth0&nbsp; &nbsp;<br />EDIMAX-tony&nbsp; &nbsp; &nbsp; ddebdded-8805-4631-aa85-f08c412404bf&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;<br />hinet&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 62250a6d-dc03-44b9-be71-7179ce1ec6c9&nbsp; pppoe&nbsp; &nbsp; &nbsp;--&nbsp; &nbsp; &nbsp;<br />TonyNote8&nbsp; &nbsp; &nbsp; &nbsp; 37db6693-372b-4190-a4d0-c71aef40bc38&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;</div><div><br /></div><div>刪除第二個 hinet 連線設定 :&nbsp;</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection delete uuid 62250a6d-dc03-44b9-be71-7179ce1ec6c9</span></b>&nbsp;&nbsp;<br />連線「hinet」 (62250a6d-dc03-44b9-be71-7179ce1ec6c9) 已成功刪除。</div><div><br /></div><div>檢視連線清單已無 hinet 了 :&nbsp;</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection show</span></b>&nbsp;&nbsp;<br />NAME&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;UUID&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; TYPE&nbsp; &nbsp; &nbsp; DEVICE&nbsp;<br />Ifupdown (eth0)&nbsp; 681b428f-beaf-8932-dce4-687ed5bae28e&nbsp; ethernet&nbsp; eth0&nbsp; &nbsp;<br />EDIMAX-tony&nbsp; &nbsp; &nbsp; ddebdded-8805-4631-aa85-f08c412404bf&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;<br />TonyNote8&nbsp; &nbsp; &nbsp; &nbsp; 37db6693-372b-4190-a4d0-c71aef40bc38&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;</div><div><br /></div><div>建立新的 PPPoE 撥接連線 :</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">sudo nmcli connection add type pppoe con-name "hinet" ifname eth0 username "光世代帳號@ip.hinet.net" password "老家電話號碼"</span></b>&nbsp; &nbsp;<br />[sudo] tony1966 的密碼：&nbsp;<br />連線「hinet」 (edf0fefa-d122-4d62-88bd-5a9cc085830a) 已成功新增。</div><div><br /></div><div>檢視連線清單新的 hinet 連線已 :&nbsp;&nbsp;</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection show</span></b>&nbsp;&nbsp;<br />NAME&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;UUID&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; TYPE&nbsp; &nbsp; &nbsp; DEVICE&nbsp;<br />Ifupdown (eth0)&nbsp; 681b428f-beaf-8932-dce4-687ed5bae28e&nbsp; ethernet&nbsp; eth0&nbsp; &nbsp;<br />EDIMAX-tony&nbsp; &nbsp; &nbsp; ddebdded-8805-4631-aa85-f08c412404bf&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;<br /><span style="background-color: #fcff01;">hinet</span>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; edf0fefa-d122-4d62-88bd-5a9cc085830a&nbsp; pppoe&nbsp; &nbsp; &nbsp;--&nbsp; &nbsp; &nbsp;<br />TonyNote8&nbsp; &nbsp; &nbsp; &nbsp; 37db6693-372b-4190-a4d0-c71aef40bc38&nbsp; wifi&nbsp; &nbsp; &nbsp; --&nbsp; &nbsp; &nbsp;</div><div><br /></div><div>用下列指令設定開機自動用此連線設定撥接上網 :&nbsp;</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection modify id "hinet" connection.autoconnect yes</span></b></div><div><br /></div><div>手動用 hinet 連線撥接上網 :&nbsp;</div><div><br />tony1966@LX2438:~$ <b><span style="color: #2b00fe;">nmcli connection up id "hinet"</span></b>&nbsp; &nbsp;<br />連線已成功啟用（D-Bus 啟用路徑：/org/freedesktop/NetworkManager/ActiveConnection/4）<br />tony1966@LX2438:~$ ifconfig<br />eth0: flags=4163&lt;UP,BROADCAST,RUNNING,MULTICAST&gt;&nbsp; mtu 1500<br />&nbsp; &nbsp; &nbsp; &nbsp; ether 16:72:2e:da:e7:3c&nbsp; txqueuelen 1000&nbsp; (Ethernet)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX packets 95413&nbsp; bytes 73056382 (73.0 MB)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX errors 0&nbsp; dropped 0&nbsp; overruns 0&nbsp; frame 0<br />&nbsp; &nbsp; &nbsp; &nbsp; TX packets 490867&nbsp; bytes 54678967 (54.6 MB)<br />&nbsp; &nbsp; &nbsp; &nbsp; TX errors 0&nbsp; dropped 0 overruns 0&nbsp; carrier 0&nbsp; collisions 0<br />&nbsp; &nbsp; &nbsp; &nbsp; device interrupt 14&nbsp;&nbsp;<br /><br />lo: flags=73&lt;UP,LOOPBACK,RUNNING&gt;&nbsp; mtu 65536<br />&nbsp; &nbsp; &nbsp; &nbsp; inet 127.0.0.1&nbsp; netmask 255.0.0.0<br />&nbsp; &nbsp; &nbsp; &nbsp; inet6 ::1&nbsp; prefixlen 128&nbsp; scopeid 0x10&lt;host&gt;<br />&nbsp; &nbsp; &nbsp; &nbsp; loop&nbsp; txqueuelen 1000&nbsp; (Local Loopback)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX packets 262801&nbsp; bytes 19744911 (19.7 MB)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX errors 0&nbsp; dropped 0&nbsp; overruns 0&nbsp; frame 0<br />&nbsp; &nbsp; &nbsp; &nbsp; TX packets 262801&nbsp; bytes 19744911 (19.7 MB)<br />&nbsp; &nbsp; &nbsp; &nbsp; TX errors 0&nbsp; dropped 0 overruns 0&nbsp; carrier 0&nbsp; collisions 0<br /><br /><span style="background-color: #fcff01;">ppp0</span>: flags=4305&lt;UP,POINTOPOINT,RUNNING,NOARP,MULTICAST&gt;&nbsp; mtu 1492<br />&nbsp; &nbsp; &nbsp; &nbsp; inet <span style="background-color: #fcff01;">220.133.1x.1yy</span>&nbsp; netmask 255.255.255.255&nbsp; destination 168.95.98.254<br />&nbsp; &nbsp; &nbsp; &nbsp; inet6 2001:b011:c002:44b:20dd:96c1:295d:93a9&nbsp; prefixlen 64&nbsp; scopeid 0x0&lt;global&gt;<br />&nbsp; &nbsp; &nbsp; &nbsp; inet6 fe80::4948:3225:47ee:c1ee&nbsp; prefixlen 64&nbsp; scopeid 0x20&lt;link&gt;<br />&nbsp; &nbsp; &nbsp; &nbsp; inet6 2001:b011:c002:44b:ca70:4579:c7c2:10f9&nbsp; prefixlen 64&nbsp; scopeid 0x0&lt;global&gt;<br />&nbsp; &nbsp; &nbsp; &nbsp; inet6 fe80::2eab:a3ae:c3d2:2ad5&nbsp; prefixlen 64&nbsp; scopeid 0x20&lt;link&gt;<br />&nbsp; &nbsp; &nbsp; &nbsp; ppp&nbsp; txqueuelen 3&nbsp; (Point-to-Point Protocol)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX packets 121&nbsp; bytes 40452 (40.4 KB)<br />&nbsp; &nbsp; &nbsp; &nbsp; RX errors 0&nbsp; dropped 0&nbsp; overruns 0&nbsp; frame 0<br />&nbsp; &nbsp; &nbsp; &nbsp; TX packets 157&nbsp; bytes 26014 (26.0 KB)<br />&nbsp; &nbsp; &nbsp; &nbsp; TX errors 0&nbsp; dropped 0 overruns 0&nbsp; carrier 0&nbsp; collisions 0<br /><br /></div><div>出現 ppp0 網路連線表示已連網成功, 開啟瀏覽器可正常瀏覽網頁, 重新啟動系統確定重開機會自動撥接上網, 終於讓主機與網站重新恢復運作.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-33167894600729264072026-06-12T20:08:58.675+08:002026-06-12T22:35:22.337+08:00MSI 電競桌機維修 (二)<div>前天<a href="https://yhhuang1966.blogspot.com/2026/06/msi.html" target="_blank">去建國路德源電腦檢修 MSI 主機</a>, 更換了新的 ADATA 1TB 系統碟後, 昨天晚上重灌 Win11 作業系統. 雖然當初購買主機時店家有附一張 Win11 Home 原版光碟, 但這不僅要找一台外接光碟機, 重灌後還會回復到 2024 年初的 Win11, 這樣會有大量的 Windows Update 要做, 會花很多時間. 德源工程師說可以到微軟下載最新 Win11 燒到 USB 裡, 用 USB 重灌系統最省事. 因為兩年來, 微軟修正了許多 Win11 對於 Intel 第 12 代 CPU 以及 RTX 3060 顯卡的排程優化與 Bug, 直接安裝最新版, 系統運作會比兩年前的舊版本穩定非常多.&nbsp;</div><div><br /></div><div>在進行 Win11 重灌之前, 先打開機箱, 把資料碟 D 碟 (ACER 2TB) 電源與資料線暫時拔掉, 等 Win11 安裝完後再插回, 這樣可避免過程中選擇要重灌到哪一個 SSD 時不小心選錯目標為 D 碟那就糟了 .</div><div><br /></div><div>以下是 Gemini 建議的 Win11 重灌程序 (但新版 Win11 在 3 與 4 步驟畫面已有所不同) :</div><div><ol style="text-align: left;"><li>隨身碟插上主機, 開機後狂按 <span style="background-color: #fcff01;">F11</span>&nbsp;(會進入 Boot 選單)</li><li>選取 UEFI: [隨身碟名稱] 按 Enter (預設就只有這個 USB)</li><li>進入畫面後點「立即安裝」, 勾選「我沒有產品金鑰」, 勾選「Windows 11 家用版」</li><li>安裝類型請選第二個 「自訂 : 唯獨安裝 Windows (進階)」</li><li>這時候畫面上應該只會看到一個乾淨完全沒有分割過的硬碟 (寫著磁碟 0 未分配空間, 容量約 900 多 GB), 直接點它後按右下角的 「下一步」</li><li>電腦會開始跑百分比, 跑完後會自動重啟 (重啟時就可以把 USB 隨身碟拔掉了), 接著就會進入 Windows 11 新機開航的設定畫面 (設定語系, 網路, 微軟帳號)</li></ol></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 建立 Win11 安裝媒體 :</span></b></div><div><br /></div><div>拿一顆 8GB 以上的 USB 隨身碟插到一台 Win11 電腦的 USB, 然後到微軟官網下載 MediaCreationTool 軟體 (Windows11InstallationAssistant.exe), 按立即下載鈕 :</div><div><br /></div><div># <a href="https://www.microsoft.com/zh-tw/software-download/windows11" target="_blank">https://www.microsoft.com/zh-tw/software-download/windows11</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhYo9W85wPivqWmy9AmTM_cofezqj6mZD9idrFywk3XncobD-eiBkVDx_WSvhsUEQamGr7SC3hRk8Hf4eSIOoxrCut4oVs83rXZbQMcTa0jjVqV8dAm0AsYr5Wc3axEfSG2H9YpqlP3nF1z_wSbsWkToOhkz9wZn4Zk6easIQbkSuNFWvrgfUJV53QdjVbr/s1557/MediaCreationToolW11-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="739" data-original-width="1557" height="152" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhYo9W85wPivqWmy9AmTM_cofezqj6mZD9idrFywk3XncobD-eiBkVDx_WSvhsUEQamGr7SC3hRk8Hf4eSIOoxrCut4oVs83rXZbQMcTa0jjVqV8dAm0AsYr5Wc3axEfSG2H9YpqlP3nF1z_wSbsWkToOhkz9wZn4Zk6easIQbkSuNFWvrgfUJV53QdjVbr/s320/MediaCreationToolW11-1.jpg" width="320" /></a></div><br /><div>執行&nbsp;Windows11InstallationAssistant.exe, 接受授權 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhC8kUc1KO1OPnP7K4586AFmfWWF5PnKiFQbXgt4AkPIHPNfcapIWMhsJ6uW9VY6plgBhRYgi8tadzN0xa_gFVYRPPuGqaUogVj7VxymW6aYGL79fD7C97FdXF7kN5iN_y0oixqE93XO4TvGzOGGW_4Ztzlytx5Vg98379EYkvAz-4Hchocn73GbRUmsdFS/s1181/MediaCreationToolW11-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="968" data-original-width="1181" height="262" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhC8kUc1KO1OPnP7K4586AFmfWWF5PnKiFQbXgt4AkPIHPNfcapIWMhsJ6uW9VY6plgBhRYgi8tadzN0xa_gFVYRPPuGqaUogVj7VxymW6aYGL79fD7C97FdXF7kN5iN_y0oixqE93XO4TvGzOGGW_4Ztzlytx5Vg98379EYkvAz-4Hchocn73GbRUmsdFS/s320/MediaCreationToolW11-2.jpg" width="320" /></a></div><div><br /></div><div><br /></div>按下一步 :&nbsp;<div><br /></div><div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjY0mIUBgJTSuurTbGtBqOcp9bBaphivCHfHyoirwvSY6w1LzyWEc8Cl4Z8twyPSKHPCsUlD6oe9hrrHWr5SpQIfWR6zj5iw6-neNN9tovPOG80w4GJ7kH6zfXJaPK046tj92Jq6CpwZNEj3rK5LaiuT26XcEYMeSVix6So6TWFk08tOBjKpBzU3QXUx4I5/s1191/MediaCreationToolW11-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="975" data-original-width="1191" height="262" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjY0mIUBgJTSuurTbGtBqOcp9bBaphivCHfHyoirwvSY6w1LzyWEc8Cl4Z8twyPSKHPCsUlD6oe9hrrHWr5SpQIfWR6zj5iw6-neNN9tovPOG80w4GJ7kH6zfXJaPK046tj92Jq6CpwZNEj3rK5LaiuT26XcEYMeSVix6So6TWFk08tOBjKpBzU3QXUx4I5/s320/MediaCreationToolW11-3.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>勾選 USB 快閃磁碟機 :&nbsp;</div><div><br /></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjzpl5WqfNFFkF4fgW5sGpN_Wg68q5z-CLw7KL0hbCU3JmgR1K7H_jIOeigBY86uqO4m3Z4oSpgubDaLpLgK8W0ZQpGkSS8OEEeYAuHLQiI2o4P0ec0vWR9yRB5uyA3wT8hajQQ00t1BkeTY3gwLKdHf60YGi9ffj3nLKIR63Oe-It1BN1Vq3McJuLM6JGG/s1190/MediaCreationToolW11-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="976" data-original-width="1190" height="262" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjzpl5WqfNFFkF4fgW5sGpN_Wg68q5z-CLw7KL0hbCU3JmgR1K7H_jIOeigBY86uqO4m3Z4oSpgubDaLpLgK8W0ZQpGkSS8OEEeYAuHLQiI2o4P0ec0vWR9yRB5uyA3wT8hajQQ00t1BkeTY3gwLKdHf60YGi9ffj3nLKIR63Oe-It1BN1Vq3McJuLM6JGG/s320/MediaCreationToolW11-4.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>按下一步會將 Win11 寫入隨身碟 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgRJg3PUxxYosG_r6tnr-lQ1svPvlACwyFTaWAg1QJlpcJzdGLyes9jD-dgi6LT1IHdGxyZtaq5jgMdeXdsn2yjFHbSvf62bcXkFExF3nQrIGRU_Y35BtAHf3e_LiQYUTiP1uJzYXOOfNxvirK6Qrryw3Wpym_D4s3AhiecUPR5C1bGVdddSA-FNLY9kxky/s1189/MediaCreationToolW11-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="972" data-original-width="1189" height="262" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgRJg3PUxxYosG_r6tnr-lQ1svPvlACwyFTaWAg1QJlpcJzdGLyes9jD-dgi6LT1IHdGxyZtaq5jgMdeXdsn2yjFHbSvf62bcXkFExF3nQrIGRU_Y35BtAHf3e_LiQYUTiP1uJzYXOOfNxvirK6Qrryw3Wpym_D4s3AhiecUPR5C1bGVdddSA-FNLY9kxky/s320/MediaCreationToolW11-5.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI6PkxhcURBtsYXDv1aLLp4NnOppmjqgu7hB9ne2KQ2eNCeSWFiYU3ODsQYFDbImIBfVSJ3ibq5UezOxGlFGnD9iDinBxgSlAFFtWyS5qhhzRZnOWFbMLhWNVHNOcRpf9gTb17Tqh9tkR0LRnh9y6AsQdY0m9Ovp465jpBY9F4hIQS6une8Vhb4hoY16uC/s1189/MediaCreationToolW11-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="977" data-original-width="1189" height="263" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI6PkxhcURBtsYXDv1aLLp4NnOppmjqgu7hB9ne2KQ2eNCeSWFiYU3ODsQYFDbImIBfVSJ3ibq5UezOxGlFGnD9iDinBxgSlAFFtWyS5qhhzRZnOWFbMLhWNVHNOcRpf9gTb17Tqh9tkR0LRnh9y6AsQdY0m9Ovp465jpBY9F4hIQS6une8Vhb4hoY16uC/s320/MediaCreationToolW11-6.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhnJSB7POo6GMomNR4vLRWASr4XxizqHm4nRTeXqenuOk2dRpLYcIuh7asDRKJFeA44-dtjimABsaoYYBIg0zj45iAgUm6cTQTorrSr5158uTalcM6epe-HoVGJryHBx5GJUfx3JuSpGNHyeQnSLMJnLw9UTl8jYjXuXnNmaRFy-YEFr0aIS4AEKlHYAbHH/s1184/MediaCreationToolW11-7.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="971" data-original-width="1184" height="262" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhnJSB7POo6GMomNR4vLRWASr4XxizqHm4nRTeXqenuOk2dRpLYcIuh7asDRKJFeA44-dtjimABsaoYYBIg0zj45iAgUm6cTQTorrSr5158uTalcM6epe-HoVGJryHBx5GJUfx3JuSpGNHyeQnSLMJnLw9UTl8jYjXuXnNmaRFy-YEFr0aIS4AEKlHYAbHH/s320/MediaCreationToolW11-7.jpg" width="320" /></a></div><br /><div><br /></div><div>完成後取下隨身碟, 插到 MSI 電競桌機 USB (最好是主機後面的 USB 槽).&nbsp;</div><div><br /></div><div><br /></div><div><span style="background-color: #fce5cd; color: #990000; font-weight: 700;">2. 安裝 Win11 :</span></div><div><br /></div><div>MSI 桌機開機後狂按 <span style="background-color: #fcff01;">F11</span>&nbsp;會進入 Boot 選單, 目前只有 USB 隨身碟 (UEFL) 一個可開機選項, 直接按 Enter :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEijcnfP1HIhTTqG_qjgw4AFUMscZ_UYXPGKItjqgjLwhLdh2qbE1EVA0KjYJ_pOIoQdn7P9gnId1Qu-3Fo-w6rDlQ58gvwynXdhEDMoK-aF9nMmyqzs0MB7Wzg91JnrrzfhM6Lodz3z6WDL402-IqBR6bkmYbNv5pgyqB6D8PzdKOrBqsIewJMzetsOdwfI/s1376/MSI-Win11-install-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="663" data-original-width="1376" height="154" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEijcnfP1HIhTTqG_qjgw4AFUMscZ_UYXPGKItjqgjLwhLdh2qbE1EVA0KjYJ_pOIoQdn7P9gnId1Qu-3Fo-w6rDlQ58gvwynXdhEDMoK-aF9nMmyqzs0MB7Wzg91JnrrzfhM6Lodz3z6WDL402-IqBR6bkmYbNv5pgyqB6D8PzdKOrBqsIewJMzetsOdwfI/s320/MSI-Win11-install-1.jpg" width="320" /></a></div><br /><div><br /></div><div>預設繁體中文, 按下一步 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi66AVJhPItZ519R_7zYb5pYJgd74DEJLinnHqvPlvO1qpLduQFtub08lF1sn6QSJtYpaSzhe9_RamLHEQhM_FwLvf5uo5KsQpa4GGEugdxpVUmTi2Vp4cVzmSQY3Kz5wyWMuHnjfiGq7FdhqZUR1Sni06a55mLHLR9eNmUd90EdQ5u1GJDgEE0sdXkeGpu/s1157/MSI-Win11-install-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="878" data-original-width="1157" height="243" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi66AVJhPItZ519R_7zYb5pYJgd74DEJLinnHqvPlvO1qpLduQFtub08lF1sn6QSJtYpaSzhe9_RamLHEQhM_FwLvf5uo5KsQpa4GGEugdxpVUmTi2Vp4cVzmSQY3Kz5wyWMuHnjfiGq7FdhqZUR1Sni06a55mLHLR9eNmUd90EdQ5u1GJDgEE0sdXkeGpu/s320/MSI-Win11-install-2.jpg" width="320" /></a></div><div><br /></div><div><br /></div>鍵盤預設微軟注音, 按下一步 :&nbsp;</div><div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi_BbyxneDpvUNdeDYlsOqPn84ba7KS9472bPYuK6vkpsSRc2EDG52Lp0Z6inU_wza9YMwtfdP5zktnbOYlbc2SJpG3U0CiSmGmmPSvf5Wu_5DHeLLTI_MFGXTLKR-3O0tZA4amXuutWmeM0LpFNR-QmnFSmA55obnUlD-0lC2I1AQhubLQrsWF8R0wQcZO/s1156/MSI-Win11-install-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="881" data-original-width="1156" height="244" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi_BbyxneDpvUNdeDYlsOqPn84ba7KS9472bPYuK6vkpsSRc2EDG52Lp0Z6inU_wza9YMwtfdP5zktnbOYlbc2SJpG3U0CiSmGmmPSvf5Wu_5DHeLLTI_MFGXTLKR-3O0tZA4amXuutWmeM0LpFNR-QmnFSmA55obnUlD-0lC2I1AQhubLQrsWF8R0wQcZO/s320/MSI-Win11-install-5.jpg" width="320" /></a></div><br /><div><br /></div><div>選取安裝 Windows 11 :&nbsp;</div><div><br /></div><div><br /></div></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgWNET_7Xee_WxDuZC8hcs5eXph7pH8G2o31yNSinl4UHjDeZ9GK3mbRCUje-bUKn_Qc3gOrH3ntnwLiA8Zk2rz0H1nHhr7hrvYOidrPi5XXHGFmxcICj8a7C-pLyuWVEV3qJ7a2MVrhuF_6WVEISIO5lzP3dP9kOfqS0kzpfekf3ajbW3MqLz68phzZv5_/s1154/MSI-Win11-install-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="882" data-original-width="1154" height="245" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgWNET_7Xee_WxDuZC8hcs5eXph7pH8G2o31yNSinl4UHjDeZ9GK3mbRCUje-bUKn_Qc3gOrH3ntnwLiA8Zk2rz0H1nHhr7hrvYOidrPi5XXHGFmxcICj8a7C-pLyuWVEV3qJ7a2MVrhuF_6WVEISIO5lzP3dP9kOfqS0kzpfekf3ajbW3MqLz68phzZv5_/s320/MSI-Win11-install-6.jpg" width="320" /></a></div><div><br /></div><div><br /></div>輸入 Win11 金鑰 :&nbsp;<div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhJUcmdES_wzy0fOakuBILhDe9A955lPO1QNf_RnmGzhgS-bgtYAQX2x3jNRdktrgTMmTZeFSJS2BSDQ_QP_I8IdXan0tQPkmH8CvRdH426K_7mjy_4w8HCqxOV5IrcQ5RcTWh1hh_fXCPmAFSj4dmrzMtV68NxfDWHdYC_PySxhZiMFcP1uvcWUjtvmJn2/s1194/MSI-Win11-install-7.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="902" data-original-width="1194" height="242" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhJUcmdES_wzy0fOakuBILhDe9A955lPO1QNf_RnmGzhgS-bgtYAQX2x3jNRdktrgTMmTZeFSJS2BSDQ_QP_I8IdXan0tQPkmH8CvRdH426K_7mjy_4w8HCqxOV5IrcQ5RcTWh1hh_fXCPmAFSj4dmrzMtV68NxfDWHdYC_PySxhZiMFcP1uvcWUjtvmJn2/s320/MSI-Win11-install-7.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>授權條款按接受 :&nbsp;</div><br /><div><br /></div></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEha6kYv2ryBLT4dfleFSvKY3WUwPmPZEsBOTIVtWjnaOeXhQNvoK0LliuxkMt8lbu7NC9uaFMt2jBiind_JxDLVFoWMiDHoZwPNWOXHzzWznTSymAscqgJXv0-nQczRhx1BN4b7ikK7uwL41bDbd3kJ7Z3hCMrNM04Yl3HTsbddjxGsMg_6IU7FjHpISYMK/s1251/MSI-Win11-install-8.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="927" data-original-width="1251" height="237" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEha6kYv2ryBLT4dfleFSvKY3WUwPmPZEsBOTIVtWjnaOeXhQNvoK0LliuxkMt8lbu7NC9uaFMt2jBiind_JxDLVFoWMiDHoZwPNWOXHzzWznTSymAscqgJXv0-nQczRhx1BN4b7ikK7uwL41bDbd3kJ7Z3hCMrNM04Yl3HTsbddjxGsMg_6IU7FjHpISYMK/s320/MSI-Win11-install-8.jpg" width="320" /></a></div><div><br /></div><div><br /></div>安裝位置點選磁碟 0 (新的 ADATA 1TB SSD) :<div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhCPI7zWMQYdZFIfv59PyjnR-K-QqPIcsJ6GQlNgr0qmmu5jgy12AyRp67dMLIt3TkIM1nVDjPkzSiQ81040kYOJ2DYPt3bJKgr2u2omBmEOE_zHpvuhAO67YulenI7zKsKn9M5gt6hWXmv2msvITdP1_zSljdPoPMaVC3ztjML7Jp26dd3CesJ6bT-KzhF/s1240/MSI-Win11-install-9.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="956" data-original-width="1240" height="247" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhCPI7zWMQYdZFIfv59PyjnR-K-QqPIcsJ6GQlNgr0qmmu5jgy12AyRp67dMLIt3TkIM1nVDjPkzSiQ81040kYOJ2DYPt3bJKgr2u2omBmEOE_zHpvuhAO67YulenI7zKsKn9M5gt6hWXmv2msvITdP1_zSljdPoPMaVC3ztjML7Jp26dd3CesJ6bT-KzhF/s320/MSI-Win11-install-9.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>兩個都勾選按安裝 :</div><div><br /></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjWG6DsPLdWhw3ACbV0YuC2kN8iJQjz-oFfiZZwQh4C7iKqR4Tzah88BTcIcjwIlINT826bCh1PFN6k0xcdmuLoddKo_-aACerSv23drpgJZRLS3JEQaEzHXQPE09Mb4tBqQvPvRjrBJqFJP7ArY9LVdybUMdh6moKKx-qI_sSvFQwPVdWZd8BFAFGeN1Gz/s1163/MSI-Win11-install-10.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="863" data-original-width="1163" height="237" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjWG6DsPLdWhw3ACbV0YuC2kN8iJQjz-oFfiZZwQh4C7iKqR4Tzah88BTcIcjwIlINT826bCh1PFN6k0xcdmuLoddKo_-aACerSv23drpgJZRLS3JEQaEzHXQPE09Mb4tBqQvPvRjrBJqFJP7ArY9LVdybUMdh6moKKx-qI_sSvFQwPVdWZd8BFAFGeN1Gz/s320/MSI-Win11-install-10.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEir99hNI5QdND71u0jL3uD_MglCUzTBOaHYW7DUFLDNBfsLPDIJbLGn4V9z-O4_gF0S-tGlDfeE8D3TVu0NwnDQNO_MgMzqs29eEpseDJwY2xca-xxh3ZcpHf70Q3RAUsQH9iXOfxorTlpW_dUlfrCBN5SbmLBzKQve05gew02anA4cVkUEqgfLbz2pRhyphenhyphenn/s1295/MSI-Win11-install-11.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="654" data-original-width="1295" height="162" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEir99hNI5QdND71u0jL3uD_MglCUzTBOaHYW7DUFLDNBfsLPDIJbLGn4V9z-O4_gF0S-tGlDfeE8D3TVu0NwnDQNO_MgMzqs29eEpseDJwY2xca-xxh3ZcpHf70Q3RAUsQH9iXOfxorTlpW_dUlfrCBN5SbmLBzKQve05gew02anA4cVkUEqgfLbz2pRhyphenhyphenn/s320/MSI-Win11-install-11.jpg" width="320" /></a></div><br /><div><br /></div><div>安裝完成後出現國家 (台灣) 確認畫面, 按是 :</div><div><br /></div><div><br /></div></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjSHJChFAGGKdqn3OEzH6mV8V3Uj42Nzi8VtIbWz6WGWJFJ3k4-n4DiJwuTiJskaAJN0dczgy2xIhUjJc8elDhxE-8eN4IOshzKe3kmJmnXIjQQW3YJx_qaN0bnm6iLmN_k3wRobsLxJmukhWwY8rzRGQWoFHv8PHFY2Oi_hFw6_TjJTwQkaxQj9yOSBWCA/s1240/MSI-Win11-install-12.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="805" data-original-width="1240" height="208" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjSHJChFAGGKdqn3OEzH6mV8V3Uj42Nzi8VtIbWz6WGWJFJ3k4-n4DiJwuTiJskaAJN0dczgy2xIhUjJc8elDhxE-8eN4IOshzKe3kmJmnXIjQQW3YJx_qaN0bnm6iLmN_k3wRobsLxJmukhWwY8rzRGQWoFHv8PHFY2Oi_hFw6_TjJTwQkaxQj9yOSBWCA/s320/MSI-Win11-install-12.jpg" width="320" /></a></div><br /><div><br /></div><div>確認輸入法為微軟注音 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjxp3C9fbB9i32dvaKfoBVXEtb-3Ks1JytGTDzOCBs7hVr8tfzp8eJeSSgJ7vP3uz5FULOOxpraQIyb-heGz7OdRZjAB5oLzVXTJ2KFNazg2jq8X5csUbkuEPGfd-kmNHSQm6Z4LTmxlW1lt-3hqul_4epRmNWSFJOMwASj92ub2Wpb0bx7PGDSo1z61ewk/s1235/MSI-Win11-install-13.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="805" data-original-width="1235" height="209" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjxp3C9fbB9i32dvaKfoBVXEtb-3Ks1JytGTDzOCBs7hVr8tfzp8eJeSSgJ7vP3uz5FULOOxpraQIyb-heGz7OdRZjAB5oLzVXTJ2KFNazg2jq8X5csUbkuEPGfd-kmNHSQm6Z4LTmxlW1lt-3hqul_4epRmNWSFJOMwASj92ub2Wpb0bx7PGDSo1z61ewk/s320/MSI-Win11-install-13.jpg" width="320" /></a></div><br /><div><br /></div><div>第二鍵盤按跳過 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiDBmqlwbonIF5XlsUYiXFTFi_kwQjJ-d6yHXfarZ16sGzqc0t5VNoDQZlRaoGQ6EJhzBqS4YEbwiA3865Rn70vZyDFMKJAAk-_x0YpyF8YZmlSn1FwWnphGOALj6qSGXbPnHfhVjHtVkthUyUklcHOZ1b1fK9Nc89HOhYlYwSHda3tnlsrSHO46iLuYwiL/s1244/MSI-Win11-install-14.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="804" data-original-width="1244" height="207" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiDBmqlwbonIF5XlsUYiXFTFi_kwQjJ-d6yHXfarZ16sGzqc0t5VNoDQZlRaoGQ6EJhzBqS4YEbwiA3865Rn70vZyDFMKJAAk-_x0YpyF8YZmlSn1FwWnphGOALj6qSGXbPnHfhVjHtVkthUyUklcHOZ1b1fK9Nc89HOhYlYwSHda3tnlsrSHO46iLuYwiL/s320/MSI-Win11-install-14.jpg" width="320" /></a></div><br /><div><br /></div><div>命名電腦為 MSI_RTX3060 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiFHObR6IebZ5YyDqwQtApPNIHbZ6M-KPx4UKGEywWCfVzrJmsH0RG7PiG6Fce9UNwYZG0-8Xw6TmteHUTGldEYSKLaA3Rv8CcgPPdQIXp5MqAPatLsk3KdxPMn53ZCxRFFsniMFQQg8uzVk23ojmwFSJkDtngA-c72mV1WYDcMOoy27K3-QEUqYmQuze8m/s1217/MSI-Win11-install-15.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="759" data-original-width="1217" height="200" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiFHObR6IebZ5YyDqwQtApPNIHbZ6M-KPx4UKGEywWCfVzrJmsH0RG7PiG6Fce9UNwYZG0-8Xw6TmteHUTGldEYSKLaA3Rv8CcgPPdQIXp5MqAPatLsk3KdxPMn53ZCxRFFsniMFQQg8uzVk23ojmwFSJkDtngA-c72mV1WYDcMOoy27K3-QEUqYmQuze8m/s320/MSI-Win11-install-15.jpg" width="320" /></a></div><br /><div><br /></div><div>登入微軟帳號 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh6m8X-9L0fm6cjI0tEUBw08PmQoOk3Z-McTENZ8n6fYiDJc4OfRAkndKVTvRa4qJzWhFXAvghwldAVlrwTOU1alDOx5EA10QzZXULIoVpz-EM3RHEAkwFuIKuPEMcJ278JA_dQ1405KVo3xqmgV4dDXP1wAusx_n7WdN0foEv2tqfyt63u8Egnh1G04pLM/s1277/MSI-Win11-install-16.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="815" data-original-width="1277" height="204" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh6m8X-9L0fm6cjI0tEUBw08PmQoOk3Z-McTENZ8n6fYiDJc4OfRAkndKVTvRa4qJzWhFXAvghwldAVlrwTOU1alDOx5EA10QzZXULIoVpz-EM3RHEAkwFuIKuPEMcJ278JA_dQ1405KVo3xqmgV4dDXP1wAusx_n7WdN0foEv2tqfyt63u8Egnh1G04pLM/s320/MSI-Win11-install-16.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh7ncoS6Y4ER1m1GWv-3glxuyaqT7itNBgBmleVahWEp6KNb5sfkWOTXmW1mc6BNDmoazRIj5Nm0XQwC_jeyEtkRrnCckHw8jXNTHVokY_6C8v3TImQ4lS3mvFTVwpnekTVOj54Bi_R0SnwKd6izXjvEBXqlPHJUjtLHvR8EE7Gyw06wFpT_5YswjQUGRgB/s1256/MSI-Win11-install-17.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="727" data-original-width="1256" height="185" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh7ncoS6Y4ER1m1GWv-3glxuyaqT7itNBgBmleVahWEp6KNb5sfkWOTXmW1mc6BNDmoazRIj5Nm0XQwC_jeyEtkRrnCckHw8jXNTHVokY_6C8v3TImQ4lS3mvFTVwpnekTVOj54Bi_R0SnwKd6izXjvEBXqlPHJUjtLHvR8EE7Gyw06wFpT_5YswjQUGRgB/s320/MSI-Win11-install-17.jpg" width="320" /></a></div><br /><div><br /></div><div>設定主機登入密碼 (PIN=akf) :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLs10e6Y_j7Svq0eRmvOLExA-kePAXXKi9EcyNmxTH2lxKqd76inqb3bwE20TmvWpe3ztqOfRT_d-h0naWmQmAxFf301i9r5uI9sSsCfcx9V6DNE8QsLNLWquabN5sQs55lNfkWOqrrwdMF4C8QSsfSiC9_7FYi1R9WogzqA41gOvqRJmVmWFZv9axM-RM/s1192/MSI-Win11-install-19.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="768" data-original-width="1192" height="206" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLs10e6Y_j7Svq0eRmvOLExA-kePAXXKi9EcyNmxTH2lxKqd76inqb3bwE20TmvWpe3ztqOfRT_d-h0naWmQmAxFf301i9r5uI9sSsCfcx9V6DNE8QsLNLWquabN5sQs55lNfkWOqrrwdMF4C8QSsfSiC9_7FYi1R9WogzqA41gOvqRJmVmWFZv9axM-RM/s320/MSI-Win11-install-19.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhLZzUZE2zYJOp48XZFjW_fUhgWUsmg6UgpOWsB_yj0hAC1uj2xdo4FJSZtGqmV4Nfrsv5ZTRMhajFuLczksPAYvr98-KsPi6EdjQUfJdgLi0HGMzLEQ4sCkh7ahzsMp1tlgOxKmN5F6NwPO6I_uDUHJKh6E8h1N9o3S-4G5FFVESv5NM4NDPgBw7hCYArs/s1241/MSI-Win11-install-20.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="797" data-original-width="1241" height="206" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhLZzUZE2zYJOp48XZFjW_fUhgWUsmg6UgpOWsB_yj0hAC1uj2xdo4FJSZtGqmV4Nfrsv5ZTRMhajFuLczksPAYvr98-KsPi6EdjQUfJdgLi0HGMzLEQ4sCkh7ahzsMp1tlgOxKmN5F6NwPO6I_uDUHJKh6E8h1N9o3S-4G5FFVESv5NM4NDPgBw7hCYArs/s320/MSI-Win11-install-20.jpg" width="320" /></a></div><br /><div><br /></div><div>然後系統重開機後, 就用此 PIN 登入即可, Win11 會自動進行更新.&nbsp;</div><div><br /></div><div><br /></div><div><span style="background-color: #fce5cd; color: #990000;"><b>3. 接回 D 碟 :</b></span></div><div><br /></div><div>關機後拔掉主機後面電源線, 把 2TB ACER SSD 的電源線與 SATA 資料線接回去, 再接回主機電源線後開機, 開啟檔案總管預期會看到 D 碟才對, 但是卻沒有, 我關機換插另一排電源插頭, 重插 SATA 資料線後再開機終於順利看到 D 碟了, 裡面的資料都還在.&nbsp;</div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">4. 安裝主機板驅動程式 :</span></b></div><div><br /></div><div>Win11 更新完後自動跳出 MSI Utility Installation 程式, 裡面列出建議安裝的工具程式與驅動程式, 但不要照單全收, 只勾選檔名有 drivers 的即可, 其餘全部不要安裝 (尤其是防毒軟體 Norton), 甚至微星自家的&nbsp;MSI Center 軟體也不需要, 它是主機的控制台軟體, 提供風扇與散熱控制 (含 RGB 調光), 以及電腦狀態監控等,&nbsp;但 540 MB 的體積很臃腫, 會占據部分主機資源.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-75461858104900472342026-06-11T15:30:26.955+08:002026-06-11T16:05:06.091+08:00好書 : 實用的 Gemini API 開發點子書<div>昨天去河堤還書時順路去逛明儀, 找到這本好書 :&nbsp;</div><div><br /></div><div>#&nbsp;<a href="https://www.books.com.tw/products/0011051943" target="_blank">實用的 Gemini API 開發點子書</a> (紙本)</div><div>#&nbsp;<a href="https://readmoo.com/book/210474109000101" target="_blank">實用的 Gemini API 開發點子書</a> (電子書)</div><div><br /></div><div>回來查市圖與母校書館都沒進此書, 殘念~~</div><div><br /></div><div>但用 HyRead 查詢發現市圖有買電子書, 馬上借閱樂讀中.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-2725650015623480122026-06-11T09:19:53.122+08:002026-06-11T09:25:32.409+08:00MSI 電競桌機維修 (一)<div>我的 MSI 主機六月初當機後就無法開機, 經表弟協助拆掉顯卡後 VGA 燈仍亮影像無輸出, 原以為是主機板內顯有問題, 但查詢 AI 才知道這主機所搭載的 CPU i5-12400F 是沒有內顯的, 編號以 F 結尾的 CPU 無內顯, 廠商在設計主機板時會讓有無內顯的 CPU 均可用, 此主機板 (MSI H610M-E) 上的 HDMI 與 VGA 埠是為有內顯的 CPU 晶片而設. 對 i5-12400F CPU 而言是無作用的, 所以初步看來 RTX3060 顯卡出包的嫌疑最大.&nbsp;</div><div><br /></div><div>這幾天梅雨不斷無法送修, 昨天雨停了, 下班後將主機送到建國路 MSI 原廠維修中心檢修 :</div><div><br /></div><div>#&nbsp;807高雄市三民區港新里建國二路41號 (地點在順發斜對面)</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjXAYgY_t-yvQNWkWNEY8lGyTvVwRiI6nsZYKBaQ95kTMxWoow6PuIo77hKrntOSBDqbaLqbi79rmzKs5yBGtykEfEpSQSQU36WrF5QkBdfnmwCv0DqMXy4lvg35l49JPFd_-YA36jhhvUE1jd5pWV3h1Ne33oH6yFYDHnovBfRHM8r9hyphenhyphenrN-72d4UTlby3/s2306/%E9%AB%98%E9%9B%84%20MSI%20%E7%B6%AD%E4%BF%AE%E4%B8%AD%E5%BF%83.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1018" data-original-width="2306" height="141" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjXAYgY_t-yvQNWkWNEY8lGyTvVwRiI6nsZYKBaQ95kTMxWoow6PuIo77hKrntOSBDqbaLqbi79rmzKs5yBGtykEfEpSQSQU36WrF5QkBdfnmwCv0DqMXy4lvg35l49JPFd_-YA36jhhvUE1jd5pWV3h1Ne33oH6yFYDHnovBfRHM8r9hyphenhyphenrN-72d4UTlby3/s320/%E9%AB%98%E9%9B%84%20MSI%20%E7%B6%AD%E4%BF%AE%E4%B8%AD%E5%BF%83.jpg" width="320" /></a></div><br /><div><br /></div><div>此主機是 2024-03-28 於 momo 購買的, 主機板與顯卡都有三年保固, 參考 :</div><div><br /></div><div>#&nbsp;<a href="https://yhhuang1966.blogspot.com/2024/03/momo.html" target="_blank">momo 購買微星電競桌機</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiodxiW4Sm7HHz6iw2zrGnIYSzYi57inh76mkU8J88KbjGSX9aH3Cmns04YjbpLhDFq6Tz_vCKMy4NlKQHaEyrXwlmd0EmQ04XReu9i6dN_GpT-PmBS1yy1M61ZzPntO9Ichtp0EBXokwYEBM5Zw1iAVSKDw7vPfKPt8-CF_kDHeYZjZo4cFij54H3Io-mc/s1764/2024-03-28%20%E6%96%BC%20momo%20%E8%B3%BC%E8%B2%B7%E5%BE%AE%E6%98%9F%E9%9B%BB%E7%AB%B6%E4%B8%BB%E6%A9%9F.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="921" data-original-width="1764" height="167" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiodxiW4Sm7HHz6iw2zrGnIYSzYi57inh76mkU8J88KbjGSX9aH3Cmns04YjbpLhDFq6Tz_vCKMy4NlKQHaEyrXwlmd0EmQ04XReu9i6dN_GpT-PmBS1yy1M61ZzPntO9Ichtp0EBXokwYEBM5Zw1iAVSKDw7vPfKPt8-CF_kDHeYZjZo4cFij54H3Io-mc/s320/2024-03-28%20%E6%96%BC%20momo%20%E8%B3%BC%E8%B2%B7%E5%BE%AE%E6%98%9F%E9%9B%BB%E7%AB%B6%E4%B8%BB%E6%A9%9F.jpg" width="320" /></a></div><br /><div><br /></div><div>MSI 原廠維修中心的工程師看了一下機身, 說這台算是半 DIY 主機, 不是微星原廠套裝主機, 所以只能幫我檢修其中的 MSI 主板與 RTX3060 顯卡, 但因為不是套裝, 無法幫我拆機 (之前有過幫客戶拆機導致的糾紛), 推薦我去附近一家德源電腦請他們幫我檢測確認問題所在, 如果問題出在主板或顯卡, 請他們拆下後拿回來 MSI 送修, 這兩塊都在保固內.&nbsp;</div><div><br /></div><div>我想說順發比較近, 且之前有在那邊維修過老電腦, 就背過去看看, 結果那邊有一大堆電腦待修, 那個女工程師也沒有之前的男工程師態度好, 所以又走去加油站對面的德源電腦, 因 MSI 原廠工程師有先打電話詢問檢測費, 所以說明來意後便馬上幫我檢修, 工程師拆掉 RTX3060 換上其它顯卡, 開機還是黑屏亮 VGA 燈 (RTX3060 嫌疑退散~), 接著先後拔兩排記憶體 (16GB + 32GB) 也是一樣, 最後拔掉 ADATA 1TB 系統碟就可順利進 BIOS, VGA 燈也熄滅了, 確認是系統 SSD 出包. 工程師說 ADATA SSD 還在保固內, 叫我拿去原廠 (就在隔壁) 換新.&nbsp;</div><div><br /></div><div>ADATA 服務也不錯, 小姐問明情況登記姓名電話後馬上拿全新盒裝給我. 德源工程師換上後檢測 OK 叫我回家自己安裝 Win11. 呵呵, 主板機與顯卡都沒問題真是好佳在, 只花了 500 元檢修費就把主機就回來了, 真是感謝親切的德源工程師, 以後若有電腦方面需求就找德源 (老闆也非常隨和, 工程師一邊檢修, 老闆就一邊跟我瞎聊). 走回公司路經 MSI 原廠維修中心, 我進去跟工程師講一下檢修結果, 並謝謝他介紹我去德源, 馬上就搞定這問題.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-14112968209216627452026-06-10T15:19:47.851+08:002026-06-10T15:22:58.492+08:00Vibe coding 時代的全端開發技術堆疊<div>今天在與 Gemini 討論目前最夯的網頁應用程式開發的技術堆疊, 它認為&nbsp;Next.js (App Router) + Tailwind + shadcn/ui 技術堆疊是開發者的黃金三角, 其中 Next.js 是以 React 為基礎的全端框架, Tailwind 負責網頁的視覺與化妝 (樣式); shadcn/ui 是 UI 元件庫, 負責處理前端最麻煩的無障礙互動與複雜動畫邏輯, 而 Next.js 則是整個專案的總架構師, 在 App Router 模式下它同時管理了前端與後端 (內建 Node.js 伺服器).&nbsp;</div><div><br /></div><div>這麼說來, 後端不是用不到 Python 了嗎? Gemini 的回答是 :&nbsp;在很多中小型專案或 SaaS 產品中確實不再需要 Python 了, Next.js 自己就能搞定全端, 以下類型的專案完全可以拋棄 Python, 只用 Next.js 一條龍開發即可 :</div><div><ul style="text-align: left;"><li>全端 SaaS 應用, 工具型網站</li><li>電商網站, 品牌官網, 部落格</li><li>企業內部管理系統, Dashboard 儀表板</li><li>需要處理會員登入, 權限控管, CRUD 的標準 Web 應用</li></ul></div><div>Next.js 內建的 Route Handlers (API 路由) 和 Server Actions (伺服器函數) 本質上就是運行在 Node.js 環境下的後端, 它可以直接連接 PostgreSQL 或 MongoDB 等資料庫, 處理金流 (Stripe, 綠界), 與發送 Email 等.&nbsp;</div><div><br /></div><div>但在特定的企業級或大數據, AI 等場景下, Python 依然無可替代, 如果專案涉及下列應用, 則後端絕對還是 Python 的天下 :&nbsp;</div><div><ul style="text-align: left;"><li>人工智慧/大語言模型 (AI/LLM/Agent) :<br />雖然可以用 JavaScript 呼叫 OpenAI 的 API, 但如果要自己做 RAG (檢索增強生成), 處理向量資料庫 (Vector Database), 微調模型 (Fine-tuning), 或是使用 LangChain/CrewAI 搭建複雜的 AI Agent 工作流, Python 的生態系 (PyTorch, Hugging Face) 是壓倒性的強大.&nbsp;</li><li>資料科學/大數據分析/網絡爬蟲 :<br />如果網站需要處理巨量數據分析, 自動化圖表生成, 或需要用 Scrapy/Selenium 去網絡上抓取大量資料, Python 的 Pandas 與 NumPy 依然是行業標準, Node.js 在這方面的效能和工具庫完整度遠遠落後.&nbsp;</li><li>影音處理與圖形計算 :<br />需要做大規模的圖片去背, 影片剪輯, 音訊轉文字 (本地端執行 Whisper 模型安等重度消耗 CPU/GPU 的任務時, Python 呼叫底層 C++ 庫的效率與生態更為成熟.&nbsp;</li></ul></div><div>所以現代全端開發的終極架構可以做如下分工 :</div><div><ul style="text-align: left;"><li>Next.js + Tailwind + shadcn/ui&nbsp;: <br />扮演對外的門面與大腦, 負責處理所有的前端 UI (Tailwind + shadcn), 路由跳轉, 使用者登入, 購物車邏輯, 並提供極致的 SEO.&nbsp;</li><li>Python : <br />扮演秘密研究室, 不直接面對使用者, 它靜靜地躲在 Next.js 後面, 被包成一個個高速的 FastAPI 介面, 當 Next.js 需要進行 AI 分析或大數據計算時再從後台發請求給 Python, 算完把結果丟回給 Next.js 呈現.&nbsp;</li></ul></div><div>結論是 : 我的 Django 書可以收起來了, 但 FastAPI 可以涉獵一下.&nbsp;</div><div><br /></div><div>雖然在 Vibe Coding 時代不懂程式技術的小白也可以靠 AI 完成一個專案, 但小白與熟練工程師的差別是過程 : 百轉千迴 vs 一步到位. <span style="background-color: #fcff01;">未來的軟工要培養像王語嫣那樣的能力, 對武林各門各派武功聊如指掌, 能指點如何運氣使力發功, 即使本身一點武功都沒有也無妨</span>.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-32563071686611305992026-06-09T10:02:28.240+08:002026-06-09T10:12:13.823+08:00momo 購買小米除蟎機 2 Pro<div>之前在<a href="https://yhhuang1966.blogspot.com/2025/12/blog-post_25.html" target="_blank">小米裕誠店購買的小米除蟎機 Pro</a> 放在鄉下使用, 想說高雄家也放一個免得帶來帶去不方便, 為了攢 momo 幣, 這回不去小米店買了 (反正價錢都一樣) :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh1-9nRKCWlw1APCQ8k6f09mnD9BZe3iPzwv-H77rlM-sxW7SKWEJqZ8p9_PHXvuIJcgjSghzYm-c9b0eQ-fxDWGM5CXcEJo3nlKie97ddZz8wBzAhjKIwGb_mBAJprb7rTkDqwYJNVVqvRu2Wpgr29F-jKMbStVrjhCgfmF3gVFMRDHvMBGDom9sKUhe_5/s1776/momo%20%E8%B3%BC%E8%B2%B7%E5%B0%8F%E7%B1%B3%E9%99%A4%E8%9F%8E%E6%A9%9F%202%20Pro-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="960" data-original-width="1776" height="173" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh1-9nRKCWlw1APCQ8k6f09mnD9BZe3iPzwv-H77rlM-sxW7SKWEJqZ8p9_PHXvuIJcgjSghzYm-c9b0eQ-fxDWGM5CXcEJo3nlKie97ddZz8wBzAhjKIwGb_mBAJprb7rTkDqwYJNVVqvRu2Wpgr29F-jKMbStVrjhCgfmF3gVFMRDHvMBGDom9sKUhe_5/s320/momo%20%E8%B3%BC%E8%B2%B7%E5%B0%8F%E7%B1%B3%E9%99%A4%E8%9F%8E%E6%A9%9F%202%20Pro-2.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEguukbD0BOJSJQpgz1OHHYcmIF6c5-crxdBudjDTfjxWGOVX0WvRkfFVqSPylGPrcF4I9T-OkifAdqkbK8D_VUvPaWBh3YLNnCpcvdytkDrE56BhDikQyJILIQ0zpjN74_WlCUDBPE_A_QmCdVdrkegtzLgiFlgJgpuoSxFtvKikrweMTi0zZcZzTf5ejBj/s1703/momo%20%E8%B3%BC%E8%B2%B7%E5%B0%8F%E7%B1%B3%E9%99%A4%E8%9F%8E%E6%A9%9F%202%20Pro-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="579" data-original-width="1703" height="109" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEguukbD0BOJSJQpgz1OHHYcmIF6c5-crxdBudjDTfjxWGOVX0WvRkfFVqSPylGPrcF4I9T-OkifAdqkbK8D_VUvPaWBh3YLNnCpcvdytkDrE56BhDikQyJILIQ0zpjN74_WlCUDBPE_A_QmCdVdrkegtzLgiFlgJgpuoSxFtvKikrweMTi0zZcZzTf5ejBj/s320/momo%20%E8%B3%BC%E8%B2%B7%E5%B0%8F%E7%B1%B3%E9%99%A4%E8%9F%8E%E6%A9%9F%202%20Pro-1.jpg" width="320" /></a></div><br /><div><br /></div><div>momo 幣剩 1 元也要用掉, 避免過期浪費了, 實付 1998 元.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-88343095251425695482026-06-08T15:02:06.781+08:002026-06-08T15:11:26.804+08:00取消 booking.com 岡山格蘭比亞預定<div>7 月下旬要帶阿中去岡山採水蜜桃, 最早是在 booking.com 上預定岡山車站旁的格蘭比亞酒店, 後來水某同事推薦 JR 經營的 VIA INN, 不僅價格更便宜, 地點就在車站樓上超方便, 預定時都選 7/20 才扣款, 今天上 booking.com 把格蘭比亞取消掉 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhlNjdAu-qB06GFBi1dcp28EuFUtP7euozVxZxjtDB8MPywDiBzGoHykON5BQ0S0vEQmXTxtsq4FD4IElXEf4VA7P-tT6QQ0ITqBTP89kKT1m1WMb-BoUQM1-dqLaE52STrNNU1_ppqnKGMrblDO2TP0gRJKqTVKSdTDfs9my_HN-XCKwjqSN6i6MohcMK5/s1046/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1046" data-original-width="689" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhlNjdAu-qB06GFBi1dcp28EuFUtP7euozVxZxjtDB8MPywDiBzGoHykON5BQ0S0vEQmXTxtsq4FD4IElXEf4VA7P-tT6QQ0ITqBTP89kKT1m1WMb-BoUQM1-dqLaE52STrNNU1_ppqnKGMrblDO2TP0gRJKqTVKSdTDfs9my_HN-XCKwjqSN6i6MohcMK5/s320/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-1.jpg" width="211" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEij0z2KU2Pj8DUEeqdJA41vM8dVT0HawGC-uVkZOBv_81zlEZikHchHgrwKA0ZkIKBctsd_sTjBn36rrkbQMk2_CXtFdB2JKFwhtL6GQ5Qq13-_J7Wvww_a_v8kBzKLMMGeojYRPijPUoqtkP8iyPCyRLsE-70ea9svd5HY7ZaIJD9HrZ6acGKCvDmZXgWa/s2147/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="911" data-original-width="2147" height="136" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEij0z2KU2Pj8DUEeqdJA41vM8dVT0HawGC-uVkZOBv_81zlEZikHchHgrwKA0ZkIKBctsd_sTjBn36rrkbQMk2_CXtFdB2JKFwhtL6GQ5Qq13-_J7Wvww_a_v8kBzKLMMGeojYRPijPUoqtkP8iyPCyRLsE-70ea9svd5HY7ZaIJD9HrZ6acGKCvDmZXgWa/s320/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-2.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh1U0g8xbjogNZrJj4KInadtXKLv4QNX14evYnDHdyCVqQA43_rZe6vFFWKUxqb8o3Hc7Rlj0hJVrAH5lXTrM58L3Mpf034dQHseCqZU3mwhVpFcKUUozKVkr7wrwPMiBJWJ6xN4S6Zdi_Aet7c7BAfOnJqg8-97_sckPW5u57t_4Qs-eW0j_HgP7lYjy83/s1860/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="777" data-original-width="1860" height="134" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh1U0g8xbjogNZrJj4KInadtXKLv4QNX14evYnDHdyCVqQA43_rZe6vFFWKUxqb8o3Hc7Rlj0hJVrAH5lXTrM58L3Mpf034dQHseCqZU3mwhVpFcKUUozKVkr7wrwPMiBJWJ6xN4S6Zdi_Aet7c7BAfOnJqg8-97_sckPW5u57t_4Qs-eW0j_HgP7lYjy83/s320/%E5%8F%96%E6%B6%88booking.com%E6%A0%BC%E8%98%AD%E6%AF%94%E4%BA%9E%E9%85%92%E5%BA%97%E9%A0%90%E5%AE%9A%E6%88%BF-3.jpg" width="320" /></a></div><br /><div><br /></div><div>延後扣款雖然價格會高些, 但好處是扣款日前隨時可 0 元取消. 但嚴格來說不是前一日取消都可以, 我查 VIA INN 的取消政策是必須至少兩天前取消才是零元, 例如 7/24 入住, 必須在 7/21 結束前取消才免費 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjU_oe5BVRE0MECWe6B97-MQIUD3X_i_874g7Jrvgp9aa7SfY3A_Q_xaYtrcsybOIKhmOJQDReq3YG0MTomPx9CAb_Rc-MvZjCp2Hm2kn52-oBohWGVNBWEov_9MbuyjZtauovqE6EanuR0OoOpn1rP-i4AUgP4HsfZYhlAP1-nnCPvOyb-_HsoAza7Mz0j/s1396/VIA-INN-%E8%A8%82%E6%88%BF%E5%8F%96%E6%B6%88%E6%94%BF%E7%AD%96.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="769" data-original-width="1396" height="176" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjU_oe5BVRE0MECWe6B97-MQIUD3X_i_874g7Jrvgp9aa7SfY3A_Q_xaYtrcsybOIKhmOJQDReq3YG0MTomPx9CAb_Rc-MvZjCp2Hm2kn52-oBohWGVNBWEov_9MbuyjZtauovqE6EanuR0OoOpn1rP-i4AUgP4HsfZYhlAP1-nnCPvOyb-_HsoAza7Mz0j/s320/VIA-INN-%E8%A8%82%E6%88%BF%E5%8F%96%E6%B6%88%E6%94%BF%E7%AD%96.jpg" width="320" /></a></div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-57251194047668243742026-06-08T10:12:25.871+08:002026-06-08T10:12:25.871+08:00好站 : 輕鬆製作可開機的 USB 磁碟機<div>這個周末把水某十二年前的 ASUS 舊筆電 (X550CC) 成功地利用 ISO 檔從 Win8 升級為 Win10, 在這過程中找到一個介紹 Rufus 工具軟體的網站 :&nbsp;</div><div><br /></div><div>#&nbsp;<a href="https://rufus.ie/zh_TW/" target="_blank">輕鬆製作可開機的 USB 磁碟機</a></div><div><br /></div><div>它除了可用來製作開機 USB 外, 若將來要將 X550CC 的系統硬碟換成 SSD, 也可以用它來燒錄 ISO 檔灌 Win10.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-39632087742343463102026-06-08T08:13:15.267+08:002026-06-08T08:13:15.268+08:00補記 : 5/21 Google雲端實驗室線上課程<div>今天整理筆電瀏覽器頁籤, 發現 5/21 參加 Google 線上課程的網頁, 當日語內訓課程衝堂, 所以也沒仔細聽, 雖有錄影但尚未有時間回看, 先把連結記錄下來再說 :&nbsp;</div><div><div><br /></div><div># <a href="https://cloudonair.withgoogle.com/events/labs-onair-ai-build-a-dynamic-website-with-gemini-cli-may-2026" target="_blank">使用 Gemini CLI 建立動態網站</a></div><div># <a href="https://cloud.google.com/resources/apac-labs-onair#upcoming-labs" target="_blank">Welcome to Google Cloud Labs: OnAir</a></div><div># <a href="https://cloudonair.withgoogle.com/events/labs-onair-ai-use-model-context-protocol-mcp-tools-with-adk-agents-may-2026" target="_blank">Use Model Context Protocol (MCP) Tools with ADK Agents</a></div><div># <a href="https://cloudonair.withgoogle.com/events/labs-onair-ai-build-a-dynamic-website-with-gemini-cli-may-2026" target="_blank">Build a Dynamic Website with Gemini CLI</a></div><div># <a href="https://cloudonair.withgoogle.com/events/labs-onair-ai-use-model-context-protocol-mcp-tools-with-adk-agents-may-2026" target="_blank">將模型上下文協定 (MCP) 工具與 ADK 代理一起使用</a></div><div><br /></div></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-77608071905954034492026-06-07T22:47:45.290+08:002026-06-12T09:16:29.020+08:002026 年第 22 周記事<div>上週都在測試 Ollama, 本來都跑得順順的, 哪知周一 MSI 電競桌機突然當機, 螢幕先是出現 MSI 便停住, 問 Gemini 可能原因, 我照其建議排查, 發現主機板上四個 debug 燈的 VGA 亮燈, 表弟小安則懷疑可能是主機板有問題, 說週六他會回鄉下, 叫我把主機帶回鄉下讓他瞧瞧, 要拆顯卡來確定是否這 VGA 是指顯卡還是內顯. 週六下午他來我家測試, 發現拆顯卡 VGA 燈還是亮, 覺得可能是主板有問題, 我下周拿去微星請工程師檢測看看.&nbsp;</div><div><br /></div><div>由於 Ollama 測試暫停, 本周主要是閱讀阿亮老師的 OpenClaw 書, 發現 GitHub CodeSpaces 也可以養龍蝦, 甚至樹莓派 Pi 4 也可以.這樣以後可以用龍蝦來管理物聯網控制器 (Node), 我以前用爬蟲來蒐集資料進行監控的做法以後都能用龍蝦取代, AI 真是大大改變了整個系統運作方式啊!&nbsp;</div><div><br /></div><div>爸現在只有周二是上午下午去社區, 最近他時間感開始混亂, 不是週二的下午也在午睡起來後著裝準備等社區車子來接, 我在想那乾脆每天都上下午都去好了, 但與江小姐聯繫後, 她建議先 2,4, 慢慢改為 2,3,4, 最後再 1,2,3,4,5, 漸進式調整較好. 下</div><div><br /></div><div>今天傍晚拿了長竿去後院與路邊採芒果, 採了一大桶. 今年芒果雖然只套了 200 顆, 但高處套不到的很多, 估計超過 200 顆, 這些果子成熟後從高處墜落都會摔傷半邊, 本來計畫在樹下張網接芒果, 今早經過小漢進去問, 一捲網子要 400 元, 至少要買 3~4 捲才夠, 為此需投資千元並不划算, 最後想想還是算了.&nbsp;</div><div><br /></div><div>下周待辦事項 :&nbsp;</div><div><ul style="text-align: left;"><li>載爸去理髮</li><li>清理庫房 + 安裝衣櫥</li><li>整理頂樓樹葉</li></ul></div><div>學研 to-do 清單 :</div><div><ul style="text-align: left;"><li>安裝 Antigravity</li><li>測試 Claude API</li><li>安裝 OpenClaw : Pi4 &amp; GitHub Codespaces</li><li>建立新的 LINE 官方帳號&nbsp;</li></ul></div><div>Hint : 要做的事情很多, 不要浪費時間在看 Reel 爛劇上.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-20915870801075762672026-06-07T12:32:33.836+08:002026-06-07T17:05:35.864+08:00ASUS X550C 筆電升版為 Win10<div>這台 ASUS X550C 艷紅色筆電是以前水某做臨床研究時我陪她去燦坤買的 (2014/5/30), 至今超過 12 年仍正常運作且非常耐用, 除了電池外其餘功能均正常, 連燒錄機都還能用 (今天還拿來將 MSI 電競桌機的 Win11 安裝光碟備份為 ISO 檔). 但那時搭載的作業系統是 Win8, 後來忘記在更新截止期限前升級為 Win10.&nbsp;</div><div><div><br /></div><div>今天備份完光碟, 想說把之前安裝的 Thonny 4.0.1 升版為 5.0.0, 結果刪除舊版後安裝新版才知道 Win8 太舊了, 無法安裝 Thonny 5, 所以只好先問 AI 這種情況下還有辦法將 Win8 提升為 Win10 嗎? 它建議到微軟官網去下載 MediaCreationTool 軟體來升版, 但我執行後出現如下錯誤 :</div></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg3wdOm9WWBfqjamyUnXT8IH3-Smvm1kWKckAAx5eJI7paI-lU_-67O_ksX1moi5rfVXKbtO5dNYHO8fPnup7ucmhWQKhZ96oNwxXuaneYUFMul4hXbNZPJVCuOLeGW4Fnxbwao_8P0pxTKGM8FpZxj0xRn215tg4pT5Bep14uTIKZe34PN6PlDwSMAhmjP/s478/media_creation_tool_error.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="164" data-original-width="478" height="110" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg3wdOm9WWBfqjamyUnXT8IH3-Smvm1kWKckAAx5eJI7paI-lU_-67O_ksX1moi5rfVXKbtO5dNYHO8fPnup7ucmhWQKhZ96oNwxXuaneYUFMul4hXbNZPJVCuOLeGW4Fnxbwao_8P0pxTKGM8FpZxj0xRn215tg4pT5Bep14uTIKZe34PN6PlDwSMAhmjP/s320/media_creation_tool_error.jpg" width="320" /></a></div><br /><div><br /></div><div>看來已無法從微軟官網更新, 問 Gemini 只有從網路社區下載 ISO 檔這方法了.&nbsp;</div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 下載 Windows 10 光碟映像檔 (ISO) :</span></b>&nbsp;</div><div><br /></div><div>試過 Gemini 建議的幾個下載 Win10 ISO 檔的方法都失敗後, Gemini 拿出壓箱寶, 叫我試試從下列網站下載, 如果這台 Win8 無法出現下載頁就用其他台電腦下載後再複製過來 :</div><div><br /></div><div>#&nbsp;<a href="http://massgrave.dev" target="_blank">massgrave.dev</a>&nbsp;&nbsp;</div><div><br /></div><div>按上方 Download Windows/Office :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEir0lCJmaCcr2Mge0bi7hGg-ZsW1Bx-M4cKZevG73y0MJ52RAjOvu6rxZXwRdCACkX3bcvJg-DxoeFz2jGoKBsgmAo3YQwHU536eNmTkX1aiHmJxAOH1yiqvRerrWgtTKpYrPE9j98K4t2eKdfrCUgkuyIbS5ojoNuVeqWVwAITfYIQLdF4vIBz1Rc0ewHZ/s913/download_win10_22h2_iso_3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="569" data-original-width="913" height="199" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEir0lCJmaCcr2Mge0bi7hGg-ZsW1Bx-M4cKZevG73y0MJ52RAjOvu6rxZXwRdCACkX3bcvJg-DxoeFz2jGoKBsgmAo3YQwHU536eNmTkX1aiHmJxAOH1yiqvRerrWgtTKpYrPE9j98K4t2eKdfrCUgkuyIbS5ojoNuVeqWVwAITfYIQLdF4vIBz1Rc0ewHZ/s320/download_win10_22h2_iso_3.jpg" width="320" /></a></div><br /><div>按 Windows10 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhWc5kefLvyo6wSO5DTkEh0gjHgIIGi_XjOVOZKBNWAeOjBXzMgMcQP2MHC1gKlSfB2NQZ6l6Y0S43Tmi7mvHGQkZo4zQTk9iNZhag3NtZ2MfM36fIilt_PF9KhFAizJfCgj_-8IKw1xbSJnnAUlcbnPKPjbubBcytCtHIxprRbKf8f7Pv6VUNIsxSBvtEm/s917/download_win10_22h2_iso_4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="606" data-original-width="917" height="211" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhWc5kefLvyo6wSO5DTkEh0gjHgIIGi_XjOVOZKBNWAeOjBXzMgMcQP2MHC1gKlSfB2NQZ6l6Y0S43Tmi7mvHGQkZo4zQTk9iNZhag3NtZ2MfM36fIilt_PF9KhFAizJfCgj_-8IKw1xbSJnnAUlcbnPKPjbubBcytCtHIxprRbKf8f7Pv6VUNIsxSBvtEm/s320/download_win10_22h2_iso_4.jpg" width="320" /></a></div><br /><div><br /></div><div>按 Windows 10 Custermer 22H2, 拉到頁面最底下, 按倒數第二個的 X64 版 Traditional Chinese 的連結 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgtOvw3an88gKw4ODgjnbj4sTNRKJYY2iZySPzfRkmT2LEAkPTbDfJ0wNiVg3e3aZDs461JTo7sWCXIJqC8pBfSWOFek851QDKjBQHl_mUJHd_HzYu1Zdp9mr-gKiBV2LPCTVFMujmzFb1oTxSJT9hp7TJBlA_J_pi5ZZPF3pKz4_FO-uGPSyLpLLApKy09/s1228/download_win10_22h2_iso_5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="647" data-original-width="1228" height="169" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgtOvw3an88gKw4ODgjnbj4sTNRKJYY2iZySPzfRkmT2LEAkPTbDfJ0wNiVg3e3aZDs461JTo7sWCXIJqC8pBfSWOFek851QDKjBQHl_mUJHd_HzYu1Zdp9mr-gKiBV2LPCTVFMujmzFb1oTxSJT9hp7TJBlA_J_pi5ZZPF3pKz4_FO-uGPSyLpLLApKy09/s320/download_win10_22h2_iso_5.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhPne3pXcsU1nTi8DruQAGcXQqtkSlVsCaR-GqFZ8-7GutzJFGK2p04Xa0NBVmAZb_EkTJa82Ef6sHggxGuuKquCF5_M6_YmjvLFd6ntsk-ET0_CZR2SEcXMpsO2y5_Bat8TRZteosgriC32LqW5CpPiyfEPAmMkgQH41ERheCd9RAwbNPnK9ACBEdBc6Vg/s1301/download_win10_22h2_iso_6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="672" data-original-width="1301" height="165" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhPne3pXcsU1nTi8DruQAGcXQqtkSlVsCaR-GqFZ8-7GutzJFGK2p04Xa0NBVmAZb_EkTJa82Ef6sHggxGuuKquCF5_M6_YmjvLFd6ntsk-ET0_CZR2SEcXMpsO2y5_Bat8TRZteosgriC32LqW5CpPiyfEPAmMkgQH41ERheCd9RAwbNPnK9ACBEdBc6Vg/s320/download_win10_22h2_iso_6.jpg" width="320" /></a></div><br /><div><br /></div><div>這會開啟&nbsp;BUZZHEAVIER 的下載頁面, 按其中的 Server1 或 Server2 即可下載 ISO 檔 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgpqOp5eZksVyXd9KQZnPJaAxcaYta6aNJllKjJkjzCVlUZddTsT_N_Un3uDlS9-QIaxDQk63V6sy4HfsnyTkS69uilN2zA7QpKJHkS9D4UUxppD3r1gI76N-s3SH3VyNCMzlH1RYCGXcUuRh7-00e9vI_ApflRD7l_q-x3fiTHG0KmivcIIlXbdZ6YghCr/s1199/download_win10_22h2_iso_7.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="654" data-original-width="1199" height="175" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgpqOp5eZksVyXd9KQZnPJaAxcaYta6aNJllKjJkjzCVlUZddTsT_N_Un3uDlS9-QIaxDQk63V6sy4HfsnyTkS69uilN2zA7QpKJHkS9D4UUxppD3r1gI76N-s3SH3VyNCMzlH1RYCGXcUuRh7-00e9vI_ApflRD7l_q-x3fiTHG0KmivcIIlXbdZ6YghCr/s320/download_win10_22h2_iso_7.jpg" width="320" /></a></div><br /><div><br /></div><div>但重點來了, 不要以為放著等他下載完畢, 這個網站通常會在下載幾分鐘後降速, 接著出現網路連線錯誤, 好在此網站有續傳功能, 這時按下繼續鈕就會恢復下載, 整個過程要盯著看下載狀態, 一出現網路連線錯誤就要按繼續 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgdICfCLVO-YDdI7SVOKXjRceGgOsrz3c59vGSOxow1n47XSYa7aI-hehheHODWrkaKFm-z0i3xSEgn5YLSw-zM3dw1KjSGRb_UFmU10GMZgHBWToG6OH4BS-P5o2Wt-q-XgUWXwP9hIeor2Cz8-lTilEH1TBTS75ePpC-hzw8wMaA7pz2MNHWha8N6HAZ4/s413/download_win10_22h2_iso_1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="272" data-original-width="413" height="211" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgdICfCLVO-YDdI7SVOKXjRceGgOsrz3c59vGSOxow1n47XSYa7aI-hehheHODWrkaKFm-z0i3xSEgn5YLSw-zM3dw1KjSGRb_UFmU10GMZgHBWToG6OH4BS-P5o2Wt-q-XgUWXwP9hIeor2Cz8-lTilEH1TBTS75ePpC-hzw8wMaA7pz2MNHWha8N6HAZ4/s320/download_win10_22h2_iso_1.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiclk-vzLLPDfQOosBHlVmYvrhc77YrOAEQdzhk9bxk4SeqKFM6zbvjEbe31oD4TfhNZChsxrxELtaYHyurCZzj9Z1tdi1Z8vUGPrgYZxd00ywJY7thTg7Lq0VEtR9xrXuGy-ls8B6T9JtN1wT5WqAZSsV14ywIJXba3fXXA3FMc1txqnVn7E5wRKIIvscp/s437/download_win10_22h2_iso_2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="326" data-original-width="437" height="239" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiclk-vzLLPDfQOosBHlVmYvrhc77YrOAEQdzhk9bxk4SeqKFM6zbvjEbe31oD4TfhNZChsxrxELtaYHyurCZzj9Z1tdi1Z8vUGPrgYZxd00ywJY7thTg7Lq0VEtR9xrXuGy-ls8B6T9JtN1wT5WqAZSsV14ywIJXba3fXXA3FMc1txqnVn7E5wRKIIvscp/s320/download_win10_22h2_iso_2.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div><div><b><span style="background-color: #fce5cd; color: #990000;">2. 安裝 Windows 10 :</span></b>&nbsp;</div><div><br /></div></div><div>完成 ISO 檔下載後用 Bandizip 等軟體將 ISO 檔解壓縮, 將滑鼠移到 setup.exe 檔按右鍵點選 "以系館管理員身分執行" :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZLe11ntJMXfHMidaJX7PL5eYtaZFkSYWfzRzUKzgkMlGK-0VLOygAid0SwmQ3u5BCqfRZyQa9gioKUjZUov-6k7gZ4kGTtwTslFpqJRSYCcwWI_93VLtqDsUbGE2-8G4jxbXeT1f-c3VukxxTIsCOR6sCe9B2M28ev96Y7j0nGf9bz1MBY0x29Y4-R2fI/s663/Win10-22H2-install-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="348" data-original-width="663" height="168" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZLe11ntJMXfHMidaJX7PL5eYtaZFkSYWfzRzUKzgkMlGK-0VLOygAid0SwmQ3u5BCqfRZyQa9gioKUjZUov-6k7gZ4kGTtwTslFpqJRSYCcwWI_93VLtqDsUbGE2-8G4jxbXeT1f-c3VukxxTIsCOR6sCe9B2M28ev96Y7j0nGf9bz1MBY0x29Y4-R2fI/s320/Win10-22H2-install-1.jpg" width="320" /></a></div><br /><div><br /></div><div>按 "變更 Windows 安裝程式下載更新的方式" :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiNrFWFo7S1929-qsqcTun0QAsXuD5wpkhTD7cBoiSiXGQo0vNybVdIR87q08rKuKxqx-1_Iy6poKTJ9u7lsZdvwAEmvj7AhJyw4xQWXUCmckVRR0O8wcJCMVmm8DUaNE5wBTcOnwf3qXgM63jIY9PePcHsI-z_rMt7VlF88MTiRqUunpwgO0VeUvD9Fa2H/s712/Win10-22H2-install-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="559" data-original-width="712" height="251" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiNrFWFo7S1929-qsqcTun0QAsXuD5wpkhTD7cBoiSiXGQo0vNybVdIR87q08rKuKxqx-1_Iy6poKTJ9u7lsZdvwAEmvj7AhJyw4xQWXUCmckVRR0O8wcJCMVmm8DUaNE5wBTcOnwf3qXgM63jIY9PePcHsI-z_rMt7VlF88MTiRqUunpwgO0VeUvD9Fa2H/s320/Win10-22H2-install-2.jpg" width="320" /></a></div><br /><div><br /></div><div>勾選 "現在不要" :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxqvZhRxMOUd4yQ50IvRz8oHCUISU6KZv_fVNmNsTAuuBE7QIHpfocEIAcv-qtbbsL5hdB7iD9QS9I_5w3yE6Ohyphenhyphene3F1uJhPTJ-eXtXikMtoA4ndwZK6u_w0nTScZG0G4hyphenhyphena0LEqByAwdovCpNJrKyxGyPOrkSleuQAu56XXDM_i2C_cqHXWSt-RBo6qX-/s710/Win10-22H2-install-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="556" data-original-width="710" height="251" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgxqvZhRxMOUd4yQ50IvRz8oHCUISU6KZv_fVNmNsTAuuBE7QIHpfocEIAcv-qtbbsL5hdB7iD9QS9I_5w3yE6Ohyphenhyphene3F1uJhPTJ-eXtXikMtoA4ndwZK6u_w0nTScZG0G4hyphenhyphena0LEqByAwdovCpNJrKyxGyPOrkSleuQAu56XXDM_i2C_cqHXWSt-RBo6qX-/s320/Win10-22H2-install-3.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>按 "接受" :</div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgBvFexkUxI0iLGjMxZsBe6Pjqc86b2D3inxfqyYTLKrUyiwfgVN7wCXO4UxQjvWXoREcH3zBUdpAPnWJAVJNeDyYR-DXJag_65P-n45Mm9CxqgF-JFRLC3uEyCf2U4c8hKcWUVibHxxqkFldM87brg0bp8KlH78coslfKe2jWJTNwpoRjhcLAguJRSfFkE/s710/Win10-22H2-install-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="557" data-original-width="710" height="251" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgBvFexkUxI0iLGjMxZsBe6Pjqc86b2D3inxfqyYTLKrUyiwfgVN7wCXO4UxQjvWXoREcH3zBUdpAPnWJAVJNeDyYR-DXJag_65P-n45Mm9CxqgF-JFRLC3uEyCf2U4c8hKcWUVibHxxqkFldM87brg0bp8KlH78coslfKe2jWJTNwpoRjhcLAguJRSfFkE/s320/Win10-22H2-install-4.jpg" width="320" /></a></div><div><br /></div><div><br /></div>確認 "安裝 Windows 10 家用版" 與 "保留個人檔案與應用程式" 都有勾選後按 "安裝" :<div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgU3R7kbCJsOkSO40r-KHyijpZ49XmkggZ5paBdIVu0CQAHN8VyvzkljhJefOyZ_m22ZBsgu285pR6TJFf5QSyKBDnfYWoAZY_Phf3Ny_xW785TbHn2fXyI_Z2gPe3Hao44gSh4mBhPnF317RCmGFaIxTkhtJhyphenhyphen7eJm546q2Ivne6vn7-Z-PX19oJtvpWrP/s710/Win10-22H2-install-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="555" data-original-width="710" height="250" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgU3R7kbCJsOkSO40r-KHyijpZ49XmkggZ5paBdIVu0CQAHN8VyvzkljhJefOyZ_m22ZBsgu285pR6TJFf5QSyKBDnfYWoAZY_Phf3Ny_xW785TbHn2fXyI_Z2gPe3Hao44gSh4mBhPnF317RCmGFaIxTkhtJhyphenhyphen7eJm546q2Ivne6vn7-Z-PX19oJtvpWrP/s320/Win10-22H2-install-5.jpg" width="320" /></a></div><br /><div><br /></div><div>安裝完成後檢視系統資訊, 確認已升版到 Win 10 無誤 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiRxpERAZLce62Rn3w7OqQGt3QgepUh9mjfb2VaFQ0PtH7bXaoNVrS3uOvqEx9_fnisoLqCu1FVLuixtY2M1mXHWdv0rBgJecgJJlHBVTGU1cWlnNdhInw2UWiTCZRw9VNOS4DQce5qcU2c6DVTqmf1euFqZrP90na78W8ApYY1LzA2oLb55KdpB3dnvFXp/s791/Win10-22H2-install-8.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="352" data-original-width="791" height="142" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiRxpERAZLce62Rn3w7OqQGt3QgepUh9mjfb2VaFQ0PtH7bXaoNVrS3uOvqEx9_fnisoLqCu1FVLuixtY2M1mXHWdv0rBgJecgJJlHBVTGU1cWlnNdhInw2UWiTCZRw9VNOS4DQce5qcU2c6DVTqmf1euFqZrP90na78W8ApYY1LzA2oLb55KdpB3dnvFXp/s320/Win10-22H2-install-8.jpg" width="320" /></a></div><br /><div><br /></div><div>這樣就可以順利安裝 Thonny 5 了 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgaFp0TChxFkkCOz-KrjazHcrB70dNjp5J711SoQ3zh5Dlk_KSUBVFbcV1H9-GhwEpSL0q3IzjkyB8HjOhrJA_oCE1-4tz2yQUA7TwX4gRfr6VSSDT3GcQKxAjzfgiMyJTZ5mHk3GHisSOkg3mtvSob3E6F0vlVAaTQtqRWh5_mkwrTORqjL8qOlRl7Gr1N/s598/Win10-22H2-install-9.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="461" data-original-width="598" height="247" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgaFp0TChxFkkCOz-KrjazHcrB70dNjp5J711SoQ3zh5Dlk_KSUBVFbcV1H9-GhwEpSL0q3IzjkyB8HjOhrJA_oCE1-4tz2yQUA7TwX4gRfr6VSSDT3GcQKxAjzfgiMyJTZ5mHk3GHisSOkg3mtvSob3E6F0vlVAaTQtqRWh5_mkwrTORqjL8qOlRl7Gr1N/s320/Win10-22H2-install-9.jpg" width="320" /></a></div><div><br /></div></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">3. 備份系統映像 :</span></b></div><br />為了避免萬一 C 碟掛掉又要從 ISO 重灌系統, 接下來先清理 C 碟, 將資料移往 D 碟儲存,下載與文件, 圖片, 影片, 以及桌面的垃圾都清乾淨, 讓 C 碟盡量瘦身後, 進行系統映像備份作業.&nbsp;<div><br /></div><div>首先按左下角 "系統/設定", 搜尋 "控制台" 中開啟控制台, 將右上角的檢視方式改為大圖示 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhpQdjoCIjxyK1c1E2TVMtY7Bi6WyVqYBYUsMIuUKSt1EsMMde6W5-lPKG0XGDTGDW-5lqedyIosqbK3b9OyqRC9yrGZ2QZFc5X7vB-VHT3OXy_AbCLJgNXOpcLflS7900DBo9JmuTGRs3C8HNCDgiKc_a9FJUDRCdgXlnk1WQmfStcuaHDoBSCei13Sezb/s810/ASUS-X550CC-backup-2.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="350" data-original-width="810" height="138" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhpQdjoCIjxyK1c1E2TVMtY7Bi6WyVqYBYUsMIuUKSt1EsMMde6W5-lPKG0XGDTGDW-5lqedyIosqbK3b9OyqRC9yrGZ2QZFc5X7vB-VHT3OXy_AbCLJgNXOpcLflS7900DBo9JmuTGRs3C8HNCDgiKc_a9FJUDRCdgXlnk1WQmfStcuaHDoBSCei13Sezb/s320/ASUS-X550CC-backup-2.jpg" width="320" /></a></div><div><br /></div><div><br /></div>點選 "備份與還原 (Windows 7)" :&nbsp;<div><br /><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjqGd2DocVru62UdSqnAAQu9m_yZKvyo0J4Nc-y8Du_-ytWGIzetQkMlt3PLR7h8TDh7XzKUXCcc0HWc2mdkQ9OL8O4CyrMad7riO8pgqNxeGACx1SCvmVaPTcoXKdfhLUakAqmxhhQYojdKeMV2is9oJrQslpKcHKDngXarght4N2zdpLmhsAi6emOtvZa/s1013/ASUS-X550CC-backup-3.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="517" data-original-width="1013" height="163" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjqGd2DocVru62UdSqnAAQu9m_yZKvyo0J4Nc-y8Du_-ytWGIzetQkMlt3PLR7h8TDh7XzKUXCcc0HWc2mdkQ9OL8O4CyrMad7riO8pgqNxeGACx1SCvmVaPTcoXKdfhLUakAqmxhhQYojdKeMV2is9oJrQslpKcHKDngXarght4N2zdpLmhsAi6emOtvZa/s320/ASUS-X550CC-backup-3.jpg" width="320" /></a></div><br /><div><br /></div><div>按左上角的 "系統備份" :&nbsp;</div><div><br /></div><div><br /></div><div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiUhgGLT2O4Wmu2yi4QSUmzp6WFhz0RiMakcw5qdXUmbxwkWlGSmeJMvnrRcfAM6ed1HIPFpeYAtn86ezumzs0gMOzxLGXFeUkkFk7JyGKIRa0fOPZEClPm1tAhukI36MRiH672rsRsqmwgoaZEJmtCo_u1EJoS8Th-Vo4WcNbSsjgjY_MkKGTFykAGbm2X/s854/ASUS-X550CC-backup-4.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="455" data-original-width="854" height="170" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiUhgGLT2O4Wmu2yi4QSUmzp6WFhz0RiMakcw5qdXUmbxwkWlGSmeJMvnrRcfAM6ed1HIPFpeYAtn86ezumzs0gMOzxLGXFeUkkFk7JyGKIRa0fOPZEClPm1tAhukI36MRiH672rsRsqmwgoaZEJmtCo_u1EJoS8Th-Vo4WcNbSsjgjY_MkKGTFykAGbm2X/s320/ASUS-X550CC-backup-4.jpg" width="320" /></a></div><div><br /></div><br />系統會先掃描, 找到可放置備份映像的 D 碟, 按下一步 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj320HtDuAH9Er2laYLURo90psqjtdH1ih7wu38sHWFtZuN6Bcv4Ee9m40vHt3KKaYVkcXoQcGDRMHHcGeWQQjN29ZIx13FhvNxykVicMZkVID_smvh9qqsWK4YYcX-uURdwJ_-TimskvP9_gvLkzAKAKDV3ZB6Fp7V3a34Fj3K3AvcdcjEejaXeEpNEvMf/s618/ASUS-X550CC-backup-5.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="556" data-original-width="618" height="288" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEj320HtDuAH9Er2laYLURo90psqjtdH1ih7wu38sHWFtZuN6Bcv4Ee9m40vHt3KKaYVkcXoQcGDRMHHcGeWQQjN29ZIx13FhvNxykVicMZkVID_smvh9qqsWK4YYcX-uURdwJ_-TimskvP9_gvLkzAKAKDV3ZB6Fp7V3a34Fj3K3AvcdcjEejaXeEpNEvMf/s320/ASUS-X550CC-backup-5.jpg" width="320" /></a></div><br /><div><br /><br />按開始備份即可 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjKlwb2uv5OJyKbizN5EQqgw7UdyofWAO3wDB2IwGe9iB_9CDWSJhhT7mBB0KMo_t4RgCoV1hX3oSo6OPmrZIGI3oOg8Bh8gMZGqbBXZoaEyTsLwpCkBkgFidJ5sW84Wts3WuKrXRk6vLEft1y6p4BO68c5oBLronhNJpG5X3veL25VRxUy0cHplN3WCszf/s620/ASUS-X550CC-backup-6.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="576" data-original-width="620" height="297" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjKlwb2uv5OJyKbizN5EQqgw7UdyofWAO3wDB2IwGe9iB_9CDWSJhhT7mBB0KMo_t4RgCoV1hX3oSo6OPmrZIGI3oOg8Bh8gMZGqbBXZoaEyTsLwpCkBkgFidJ5sW84Wts3WuKrXRk6vLEft1y6p4BO68c5oBLronhNJpG5X3veL25VRxUy0cHplN3WCszf/s320/ASUS-X550CC-backup-6.jpg" width="320" /></a></div><div><br /><br />這裡有提示備份 C 碟映像會佔去 D 碟 110GB 容量, 我在備份前有檢查 D 碟還有 211GB :<br /><div><b><span style="background-color: #fce5cd; color: #990000;"><br /></span></b></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgbL6siaPCnrw-gzx20z5PSbZrbGv8B9kzT-LMqX2EsoTul_hEVvoat6yZKmrgxdBG5S9shogRk55GTAR0PZkeRZzArYjb-B0DjPzhbDALhgymkyWQ1YclOLuQTbUEmZtpgmzIj3adJQlrU3m0ckgSzaJtRcai2fxmnDN8NOGppWc8QxR2hXJRl1JRiTM72/s762/ASUS-X550CC-backup-7.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="306" data-original-width="762" height="129" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgbL6siaPCnrw-gzx20z5PSbZrbGv8B9kzT-LMqX2EsoTul_hEVvoat6yZKmrgxdBG5S9shogRk55GTAR0PZkeRZzArYjb-B0DjPzhbDALhgymkyWQ1YclOLuQTbUEmZtpgmzIj3adJQlrU3m0ckgSzaJtRcai2fxmnDN8NOGppWc8QxR2hXJRl1JRiTM72/s320/ASUS-X550CC-backup-7.jpg" width="320" /></a></div><br /><div><br /></div><div>大約 40 分鐘後備份完成, 再次檢查 D 碟容量, 剩下 110GB :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh3YYpXcQL13MeVBSCWLpId-O2pYAhotPPq1GOIR4H1Z1MjHAV6wacrvtQs-pMP3n50hYySMSzt0ASduYpVlPVea0UCkZBr641Y1HkuUztHMymnYixVoKHUVFRNsAo0PSSzzQjgJLWH0NfPTk0d-9ItuFUBvAO0teWgE_inqfub7lOlvlrlXkN3qHWYvrzv/s715/ASUS-X550CC-backup-8.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="292" data-original-width="715" height="131" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh3YYpXcQL13MeVBSCWLpId-O2pYAhotPPq1GOIR4H1Z1MjHAV6wacrvtQs-pMP3n50hYySMSzt0ASduYpVlPVea0UCkZBr641Y1HkuUztHMymnYixVoKHUVFRNsAo0PSSzzQjgJLWH0NfPTk0d-9ItuFUBvAO0teWgE_inqfub7lOlvlrlXkN3qHWYvrzv/s320/ASUS-X550CC-backup-8.jpg" width="320" /></a></div><br /><div><br /></div><div>備份完成會詢問是否要製作 Win10 修護光碟, 我去找了一張空白 DVD 讓它燒錄 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgnClEBFYgYtjZP2DDxDPevabq5439tyQW5sD0JovXV_3-AkSaBlUsG5mUdIEA8PDn7RUXnJMVaiq7_11RQhxSYkcLu1dwwTsZDZjCC2p90XdLl-ExrUxR3xd41Gx873CuExsR88trQXUOQfXa1voUNpxy9popWcOB39PHw5A8kT3AjcpAuzDNqFsYeavVi/s540/ASUS-X550CC-backup-9.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="289" data-original-width="540" height="171" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgnClEBFYgYtjZP2DDxDPevabq5439tyQW5sD0JovXV_3-AkSaBlUsG5mUdIEA8PDn7RUXnJMVaiq7_11RQhxSYkcLu1dwwTsZDZjCC2p90XdLl-ExrUxR3xd41Gx873CuExsR88trQXUOQfXa1voUNpxy9popWcOB39PHw5A8kT3AjcpAuzDNqFsYeavVi/s320/ASUS-X550CC-backup-9.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI2qaclsRGD79xuFSPch1BHxomwamGJ22AjVSQmJILUaBlfn8_pY5f2cKKYJ-PkD5GO64_XrVlB3Doh0g85etAaDEEu3OvYssLMHvOOxi-YqbzS8Hhkj7bRxTH2VEo35kJuXW85z0CLyg5ahOHeV5uEcK_NW57H8eJppQZ4bV-WGgmmP27i0Cif2t2tWJB/s511/ASUS-X550CC-backup-10.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="315" data-original-width="511" height="197" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI2qaclsRGD79xuFSPch1BHxomwamGJ22AjVSQmJILUaBlfn8_pY5f2cKKYJ-PkD5GO64_XrVlB3Doh0g85etAaDEEu3OvYssLMHvOOxi-YqbzS8Hhkj7bRxTH2VEo35kJuXW85z0CLyg5ahOHeV5uEcK_NW57H8eJppQZ4bV-WGgmmP27i0Cif2t2tWJB/s320/ASUS-X550CC-backup-10.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjOdEGrWLLaF7jiQ8jgVoFTYzAow9GSpqJyUAqF2oPr4KTo0QpXOmu4PAUIhOmT2QJ7OuAHGwXTIv9BbNQL0nsm_BToJ6eos1hWIuWl2mnMbDGW4X4x4BQwIWWZsnuhHgoYI_1hIrGIW_Pipw8sSqlNJnEd225n-s9D8Utn9W2fkeYeOJ1x5mFPuXHxB8s_/s511/ASUS-X550CC-backup-11.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="314" data-original-width="511" height="197" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjOdEGrWLLaF7jiQ8jgVoFTYzAow9GSpqJyUAqF2oPr4KTo0QpXOmu4PAUIhOmT2QJ7OuAHGwXTIv9BbNQL0nsm_BToJ6eos1hWIuWl2mnMbDGW4X4x4BQwIWWZsnuhHgoYI_1hIrGIW_Pipw8sSqlNJnEd225n-s9D8Utn9W2fkeYeOJ1x5mFPuXHxB8s_/s320/ASUS-X550CC-backup-11.jpg" width="320" /></a></div><br /><div><br /></div><div>這樣就順利完成 Win10 升版後的維護了.&nbsp;</div><div><br /></div><div>這台舊筆電在母親於 2014 年最後一次化療住院時被我借用拿去病房陪病時寫程式, 所以有一種特殊的感情. 升版完成後整理硬碟, 看到當時留下的舊資料令我感到物是人非.&nbsp;</div><div><br /></div><div><br /></div><div>2026-06-07 補充 :</div><div><br /></div><div>其實這台筆電真正型號是 X550CC, 是 X550C 款式較晚期的子型號 (沒有 SD 卡槽, 改成 2 個左側 USB 槽), 配備清單可按 Win + R 後輸入&nbsp;msinfo32 按 Enter 即可看到 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgFDigVvjsMVkJO14vjB1sE4yZfW3DH3O813_zss7G3HNaqHg212FMfwzgaMPuQyU88Bxbah4iaRG_I2t-Kj4j6g5Czf3oileRuO40BA-yk7UlIvcIpYgpL12BC9q1qjiosggNsUyW2CFRY-ylo8SE19NDPWkL0ttroKO0LIufEnEs8Zlj-6-MjW4xaZiZ1/s892/ASUS-X550CC.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="892" data-original-width="683" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgFDigVvjsMVkJO14vjB1sE4yZfW3DH3O813_zss7G3HNaqHg212FMfwzgaMPuQyU88Bxbah4iaRG_I2t-Kj4j6g5Czf3oileRuO40BA-yk7UlIvcIpYgpL12BC9q1qjiosggNsUyW2CFRY-ylo8SE19NDPWkL0ttroKO0LIufEnEs8Zlj-6-MjW4xaZiZ1/s320/ASUS-X550CC.jpg" width="245" /></a></div><div><br /></div><br /><div>CPU 是帶有 U 字尾的省電版處理器 i5-3337U, 它只支援低電壓記憶體, 此筆電主機板有一個 DRAM 擴充槽, 可買一條 8GB 或 4GB DDR3L 1600 或 PC3L-12800 (筆記型電腦專用) DRAM 補上去, 開多網頁時會比較順暢.&nbsp;</div><div><br /></div><div>但卡頓的關鍵其實是硬碟, AI 建議買一個 2.5吋 SATA3 SSD (厚度為 7mm 規格) 來替換硬碟, 例如美光 (Crucial) MX500 系列或金士頓 (Kingston) KC600 系列, 這兩款都帶有快取記憶體, 當作 Windows 10 的系統碟速度最快, 壽命也最穩定. 換 SSD 的話 Win10 要重灌, 序號因為是綁主機板所以可以延續.&nbsp;</div><div><br /></div></div><div class="separator" style="clear: both; text-align: left;">目前維持現狀即可, 此筆電目前就是擔任救援角色與光碟相關操作而已.&nbsp;</div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-81876861391036421562026-06-05T22:06:03.857+08:002026-06-06T08:13:45.809+08:00樹莓派學習筆記 : 在 Bulleye 上安裝 google-genai 套件<div>今天把當機已一個月的高雄 Pi 3B 主機重灌 Bulleye, 安裝爬蟲與資料科學相關套件後馬上製作映像檔, 以免下次 TF 卡異常又要重來. 完成後想說來安裝 Gemini 新版 API 套件 google-genai 看看(舊版是 google-generativeai 谷歌已不再維護), 雖然安裝成功, 但出現相依性問題 :</div><div><br /></div><div><div>pi@kaopi3:~ $ <b><span style="color: #2b00fe;">pip install google-genai</span></b>&nbsp;&nbsp;</div><div>Looking in indexes: https://pypi.org/simple, https://www.piwheels.org/simple</div><div>Collecting google-genai</div><div>&nbsp; Downloading https://www.piwheels.org/simple/google-genai/google_genai-1.47.0-py3-none-any.whl (241 kB)</div></div><div>... (略) ...</div><div><div>Installing collected packages: <span style="background-color: #fcff01;">websockets</span>, tenacity, google-genai</div><div>&nbsp; Attempting uninstall: websockets</div><div>&nbsp; &nbsp; Found existing installation: websockets 12.0</div><div>&nbsp; &nbsp; Uninstalling websockets-12.0:</div><div>&nbsp; &nbsp; &nbsp; Successfully uninstalled websockets-12.0</div><div>ERROR: <span style="background-color: #fcff01;">pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.</span></div><div><span style="background-color: #fcff01;">gradio-client 1.3.0 requires websockets&lt;13.0,&gt;=10.0, but you have websockets 15.0.1 which is incompatible</span>.</div><div>Successfully installed google-genai-1.47.0 tenacity-9.1.2 websockets-15.0.1</div></div><div><br /></div><div>我把錯誤提交給 Gemini 分析, 原來 google-genai 需要最新版的環境所以把 websockets 從 12.0 版升級到了 15.0.1 版,&nbsp;原本安裝的 gradio-client (1.3.0) 是一個比較挑剔的套件, 它要求 websockets 的版本必須在 10.0 到 13.0 之間 (不含 13), 現在提升 15.0.1 版了, 可能會讓 gradio 無法運作. 如果在樹莓派 Pi 3 上不會用到 gradio, 那這個錯誤訊息可以無視 (安裝 gradio 是因為它有一拖拉庫的相依套件例如 matplotlib, 並不是真的會用到 gradio).&nbsp;</div><div><br /></div><div>實測一下看看能否正常使用 :</div><div><br /></div><div><div>pi@kaopi3:~ $ <b><span style="color: #2b00fe;">python</span></b>&nbsp;&nbsp;</div><div>Python 3.9.2 (default, May 18 2026, 18:09:43)&nbsp;</div><div>[GCC 10.2.1 20210110] on linux</div><div>Type "help", "copyright", "credits" or "license" for more information.</div></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">from google import genai</span></b>&nbsp; &nbsp;</div><div>/home/pi/.local/lib/python3.9/site-packages/google/auth/__init__.py:54: FutureWarning: You are using a Python version 3.9 past its end of life. Google will update google-auth with critical bug fixes on a best-effort basis, but not with any other fixes or features. Please upgrade your Python version, and then update google-auth.</div><div>&nbsp; warnings.warn(eol_message.format("3.9"), FutureWarning)</div><div>/home/pi/.local/lib/python3.9/site-packages/google/oauth2/__init__.py:40: FutureWarning: You are using a Python version 3.9 past its end of life. Google will update google-auth with critical bug fixes on a best-effort basis, but not with any other fixes or features. Please upgrade your Python version, and then update google-auth.</div><div>&nbsp; warnings.warn(eol_message.format("3.9"), FutureWarning)</div><div><br /></div><div>此警告只是說 Python 3.9 實在太舊了而已.&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">from dotenv import dotenv_values</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">config=dotenv_values('.env')</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">gemini_api_key=config.get('GEMINI_API_KEY')</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">client=genai.Client(api_key=gemini_api_key)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">response=client.models.generate_content(</span></b>&nbsp; &nbsp;</div><div>...&nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">model='gemini-2.5-flash',</span></b>&nbsp;</div><div>...&nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">contents='你是誰?'</span></b>&nbsp;</div><div>...&nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">)</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(response.text)&nbsp;</span></b>&nbsp;</div><div>我是一个大型语言模型，由 Google 训练。</div></div><div><br /></div><div>可見此新版 Gemini API 在 Pi 3 上使用沒問題.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-38590437727270469212026-06-04T16:33:17.249+08:002026-06-06T08:16:12.444+08:00Claude Code 學習筆記 : 儲值與建立 API 金鑰 <div>最近幾天因為 MSI 電競主機故障待修, 於是轉而閱讀阿亮老師的大作 "養成你的 AI 龍蝦管家", 書中介紹三大 AI 自主代理平台 (三王工具) :&nbsp;</div><div><ul style="text-align: left;"><li>Goolge Antigravity</li><li>Claude Code</li><li>OpenClaw</li></ul></div><div>Google Antigravity 是一個多 Agent, 可預覽測試網頁的可視化 AI 程式開發工具; Claude Code 是終端機特助, 可用來管理專案資料夾, 進行 Git 版本控制與微調設定檔等. OpenClaw 則是一個能透過多通道 (LINE, Telegram) 自主完成任務的 AI 管家 (秘書), 有了 Antigravity 與 Claude Code 幫助, 才能充分發揮 OpenClaw 功能.&nbsp;</div><div><br /></div><div>本來打算先學完 Gemini CLI 後再來付月費學 Claude Code, 但看了阿亮老師的書發現其實可以先用 Pay As You Go 方式儲值購買 API 使用額度 (credits) 來用, 反正只是測試學習又不是做大專案, 吃 token 不會太兇, 不夠再除值即可.&nbsp;</div><div><br /></div><div>本系列前一篇文章參考 :</div><div><br /></div><div>#&nbsp;<a href="https://yhhuang1966.blogspot.com/2025/12/windows-claude-code.html" target="_blank">在 Windows 上安裝原生版 Claude Code</a></div><div><br /></div>我在 2023 年就已註冊 Claude AI 帳號, 所以只要到 Anthropic Console 網站登入即可 :<div><br /></div><div>#&nbsp;<a href="https://platform.claude.com/" target="_blank">https://platform.claude.com/</a></div><div><br /></div><div>我使用 Hinet 信箱登入, Claude 帳號不使用密碼, 而是每次登入都將認證碼寄到信箱 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgrV8tHvaQpRFkSIIFfDCeqVkR5IS3h24bE2Nr7HkEJQ_0RYCTyT8t9ZLqK8K-W3DuZFPJ85dAQ_gqmKjH2-6tsrfa7gny4s0rmrFq9SjHkkYT5YfNBpDHi4jJBJlmm6Tn15RwO5UsUFPpBPLAfqFjP9EAVuyywcbEgCxMIlSuTPurbmmy-YdVPcM19vG-r/s1060/claude-code-api-recharge-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="954" data-original-width="1060" height="288" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgrV8tHvaQpRFkSIIFfDCeqVkR5IS3h24bE2Nr7HkEJQ_0RYCTyT8t9ZLqK8K-W3DuZFPJ85dAQ_gqmKjH2-6tsrfa7gny4s0rmrFq9SjHkkYT5YfNBpDHi4jJBJlmm6Tn15RwO5UsUFPpBPLAfqFjP9EAVuyywcbEgCxMIlSuTPurbmmy-YdVPcM19vG-r/s320/claude-code-api-recharge-1.jpg" width="320" /></a></div><br /><div><br /></div><div>到信箱收取認證碼, 填入後按 Verify Email address 鈕 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiM58GJwsRzSdIFZYkgaWAi7QaCnG0WCIC2JTe3vHMuGsQQQ0eAAUja6gBxvaKIAxuEmH6YSugwZVmBLaqZj-YkmVGasNV0MyGbWtly9kj6B_xwFsloa99wrNcrO9710ExBrJGjr7PBB_u9ZeJS5Wt5jVf_HWcDOn6jgyyQBTO7AjnNQKBLiIENgzaaFWyK/s1026/claude-code-api-recharge-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="882" data-original-width="1026" height="275" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiM58GJwsRzSdIFZYkgaWAi7QaCnG0WCIC2JTe3vHMuGsQQQ0eAAUja6gBxvaKIAxuEmH6YSugwZVmBLaqZj-YkmVGasNV0MyGbWtly9kj6B_xwFsloa99wrNcrO9710ExBrJGjr7PBB_u9ZeJS5Wt5jVf_HWcDOn6jgyyQBTO7AjnNQKBLiIENgzaaFWyK/s320/claude-code-api-recharge-2.jpg" width="320" /></a></div><br /><div><br /></div><div>登入成功後會轉至 Dashboard 頁面, 按 Add Fund 即可儲值 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhUnGQdKgnFNKZ2ZDbAjk7Pg_C2RGoaKip3jtUUhFNqN2T7ncJa8vLjCBGSpk1pSiv2QAbHFz7kwF0Sl9_CiF-TWgslSZ9D8pYWQckgFGUKPSHk6u6k9aCJO6uT0mk7OelDSb4SG3BTfTorT41Vk_tdMVdd53cpUaXD8k4lRRBs_G-wyKNfYWcGk_8FsBTP/s2498/claude-code-api-recharge-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1122" data-original-width="2498" height="144" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhUnGQdKgnFNKZ2ZDbAjk7Pg_C2RGoaKip3jtUUhFNqN2T7ncJa8vLjCBGSpk1pSiv2QAbHFz7kwF0Sl9_CiF-TWgslSZ9D8pYWQckgFGUKPSHk6u6k9aCJO6uT0mk7OelDSb4SG3BTfTorT41Vk_tdMVdd53cpUaXD8k4lRRBs_G-wyKNfYWcGk_8FsBTP/s320/claude-code-api-recharge-3.jpg" width="320" /></a></div><br /><div><br /></div><div>既定金額有 20, 100, 500 美元, 想儲其它金額例如 10 美元可按右邊的 Others :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi8ItQ18rIgbfbn0o_-yKuYTXqKjycUdWdTORhxPXBvGMx0KKfKE9wqqke1E1A5XPFHX9Z3khWlHVFy6I-Pm4KwwHHGXIKmC1V1QAtPSFv24Xdg8phUsm8IZ-FG5BCPQbllbGJa08Kl3dOZ4AZKLUNpJIRDsian2zMeIFmzXm4qtwi1UoSIKt23qY5Vl79H/s1160/claude-code-api-recharge-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1160" data-original-width="1099" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi8ItQ18rIgbfbn0o_-yKuYTXqKjycUdWdTORhxPXBvGMx0KKfKE9wqqke1E1A5XPFHX9Z3khWlHVFy6I-Pm4KwwHHGXIKmC1V1QAtPSFv24Xdg8phUsm8IZ-FG5BCPQbllbGJa08Kl3dOZ4AZKLUNpJIRDsian2zMeIFmzXm4qtwi1UoSIKt23qY5Vl79H/s320/claude-code-api-recharge-4.jpg" width="303" /></a></div><br /><div><br /></div><div><br /></div><div>輸入 10 美元, 填寫 Billling Address 的欄位 (例如全名, 國家, 郵遞區號等) 與信用卡資訊 (JCB 卡較不容易通過綁定, 建議用 VISA 或 MASTER) :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgMl_A-QbTsZ7-ydNSM-g0YZwgAUnDLlpyV_bGkLH5xxu_I0KwrZjZyuPKp1_Xu6PbTZLPfftLbJFxgPWCoVEcDguoNm8UPD0WH-KUaF-zbwvCRrwUqz-yg7EUNSo00Q-CdR6GYt5ztGtEqql-QvzRKyXD3ctv9ZtTjk80TxiIebSKk6avVYmtAgq9hWhmf/s1147/claude-code-api-recharge-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1147" data-original-width="1096" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgMl_A-QbTsZ7-ydNSM-g0YZwgAUnDLlpyV_bGkLH5xxu_I0KwrZjZyuPKp1_Xu6PbTZLPfftLbJFxgPWCoVEcDguoNm8UPD0WH-KUaF-zbwvCRrwUqz-yg7EUNSo00Q-CdR6GYt5ztGtEqql-QvzRKyXD3ctv9ZtTjk80TxiIebSKk6avVYmtAgq9hWhmf/s320/claude-code-api-recharge-5.jpg" width="306" /></a></div><br /><div><br /></div><div>雖然資料都有填好 (Business tax ID 統編欄位不要填), 但底下 tax 與 total due 無資料, "Buy USD 10 of credits" 鈕是灰色無法按 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgEdTW45MSgRnZu1VuplPZB64-3zkrHwZ5DXNu4LLuhRc2JVKKFwRnzQ_Cy8Q7AcZKlwQVUhWawMsjNoaUjQNr7yq-cPvFERhJJa4F_9O4qQgdagiyMSurK6ayugF576nUagKjPdYEU9HMbBgTuY8TQbvGtiS3Hov3qSdEJca-r29MBN7DdDeInKpaA-D39/s1080/claude-code-api-recharge-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="820" data-original-width="1080" height="243" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgEdTW45MSgRnZu1VuplPZB64-3zkrHwZ5DXNu4LLuhRc2JVKKFwRnzQ_Cy8Q7AcZKlwQVUhWawMsjNoaUjQNr7yq-cPvFERhJJa4F_9O4qQgdagiyMSurK6ayugF576nUagKjPdYEU9HMbBgTuY8TQbvGtiS3Hov3qSdEJca-r29MBN7DdDeInKpaA-D39/s320/claude-code-api-recharge-6.jpg" width="320" /></a></div><br /><div><br /></div><div>我問 Gemini 原因, 它認為是 Anthropic 的前端對台灣這個國家的地址/稅率判定卡住了, 建議改選美國, 例如免消費稅的奧勒岡州 (Oregon), 城市填 Portlan (郵遞區號 97201), 街道地址可填 1234 SW Morrison St., 這樣就能通過檢查, 底下的購買鈕變成黑色可按 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgRjAOFuKoVS0O3RQLsbWc_dWqGIS9-LCNd8QqvHuma63e1_RQWqPEIqWcp0YeB8BVpis0QHkmyRYlkOK4Na5j4FEmX34EwJ0g-vi-PkuskNFzNxQJ25q5XxsLgDSdACfLEmWyGOPIxLBcCgoE-G28GNA_jm7psRShHu65KLrZqujui2u8KdYIZ6zIkt2VY/s1103/claude-code-api-recharge-8.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1101" data-original-width="1103" height="319" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgRjAOFuKoVS0O3RQLsbWc_dWqGIS9-LCNd8QqvHuma63e1_RQWqPEIqWcp0YeB8BVpis0QHkmyRYlkOK4Na5j4FEmX34EwJ0g-vi-PkuskNFzNxQJ25q5XxsLgDSdACfLEmWyGOPIxLBcCgoE-G28GNA_jm7psRShHu65KLrZqujui2u8KdYIZ6zIkt2VY/s320/claude-code-api-recharge-8.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZ0K98nE_N47WdUM4z9wvC4epPLguFpKtFmo1q6b0gZ5XIAwT6YMI6RRxktGoMWci6T5gu3-b3eZZHuGd5cjNFPCfSGWqNLtZGb9dfdAhC3q-XdhWZbDd1_b9wgkMatOCn9-tO2JSg64jdvtq8ZDD5qlHDfNXtr0MElnhSg75YKMlpOTcA1iiXr3di6N5U/s1104/claude-code-api-recharge-9.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1104" data-original-width="1090" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZ0K98nE_N47WdUM4z9wvC4epPLguFpKtFmo1q6b0gZ5XIAwT6YMI6RRxktGoMWci6T5gu3-b3eZZHuGd5cjNFPCfSGWqNLtZGb9dfdAhC3q-XdhWZbDd1_b9wgkMatOCn9-tO2JSg64jdvtq8ZDD5qlHDfNXtr0MElnhSg75YKMlpOTcA1iiXr3di6N5U/s320/claude-code-api-recharge-9.jpg" width="316" /></a></div><div><br /></div><br /><div>按購買鈕後會出現詢問是否要自動儲值 (Auto-reload), 按底下的 Skip for now 鈕跳過 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEilKfXhEi0OzchaLVRqdrUkGJNWII4BSMwUhmJqwk5ogQTlT1TXPOBHOLCKvqczsmdPqqVZF79wVxuuIcsYNS1DKfWviKSSpoCis1Ykxg2HkfdY2eRMcks-P55KKvDHR9IZmypBqZIQdVQqFUtDYCwy3TanuiZPN7vRk-DF66oqU6GXY_eRXm28ZWiT4zbc/s1096/claude-code-api-recharge-10.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="832" data-original-width="1096" height="243" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEilKfXhEi0OzchaLVRqdrUkGJNWII4BSMwUhmJqwk5ogQTlT1TXPOBHOLCKvqczsmdPqqVZF79wVxuuIcsYNS1DKfWviKSSpoCis1Ykxg2HkfdY2eRMcks-P55KKvDHR9IZmypBqZIQdVQqFUtDYCwy3TanuiZPN7vRk-DF66oqU6GXY_eRXm28ZWiT4zbc/s320/claude-code-api-recharge-10.jpg" width="320" /></a></div><br /><div><br /></div><div><br /></div><div>這樣便完成儲值程序, 按 Create API Key 建立金鑰 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgucA-oYNZjeJKr7c8HC0XJXdqe_wXuZ3o0ZTaeFkqf1dmV72jgZfwy21_cOoOHJsoRPLyXOKqi2mbsDEKNmRFuFQNHXpk1LNoJ3DEpE-y4avNJMGl9RgHtRPUYWGLEKQTSoWj0Oc4fGM30v7xRROscRxncSLRhz5fzwiMzztbp8ONH3CL5qMAqL9WeMhw-/s1497/claude-code-api-recharge-12.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1125" data-original-width="1497" height="240" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgucA-oYNZjeJKr7c8HC0XJXdqe_wXuZ3o0ZTaeFkqf1dmV72jgZfwy21_cOoOHJsoRPLyXOKqi2mbsDEKNmRFuFQNHXpk1LNoJ3DEpE-y4avNJMGl9RgHtRPUYWGLEKQTSoWj0Oc4fGM30v7xRROscRxncSLRhz5fzwiMzztbp8ONH3CL5qMAqL9WeMhw-/s320/claude-code-api-recharge-12.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>這兩個 2024 年測試用的金鑰應該已過時可刪除, 按 Create Key 鈕新增 :</div><div><br /></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEip3Q-57DitCqVPDTXnhrOdBGrSptiUYyd6pdm3ho7Xxvxtlp6ZPc_Ow5VH2hqIy528jSFH3Fahnp1j8sfcBI-vr-jugNX7zKVCsTELpeUuKjrDE7bB_e8U1NDmC8zp5IzxU3Ibfy07oo3Cr9Lt-PbMwZgXtjOucF93fOdqeOmnUnU6FNyF5ib6t0SVFl7-/s1924/claude-code-api-recharge-13.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="616" data-original-width="1924" height="102" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEip3Q-57DitCqVPDTXnhrOdBGrSptiUYyd6pdm3ho7Xxvxtlp6ZPc_Ow5VH2hqIy528jSFH3Fahnp1j8sfcBI-vr-jugNX7zKVCsTELpeUuKjrDE7bB_e8U1NDmC8zp5IzxU3Ibfy07oo3Cr9Lt-PbMwZgXtjOucF93fOdqeOmnUnU6FNyF5ib6t0SVFl7-/s320/claude-code-api-recharge-13.jpg" width="320" /></a></div><br /><div><br /></div><div>選擇 Workspace 輸入 Name, 按 Add 鈕新增 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhWxM0QVUD295Bwdy7BBrIkImhc-ptNrU4OoCMKfYV-W-ZfMeq_vFBdyzWFZRVRP3s3kkvUh7KfiMHD31uxjHFsHaTEmNpNmBvDGuyBI73KmbSGQJM9C98sG0vi2mjN1XphQGYKDhborjjVzVL5vRQC9FmrJs3SENP1pznOTlpxmhPDNrZ52BpMerwlyVu2/s1100/claude-code-api-recharge-14.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="547" data-original-width="1100" height="159" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhWxM0QVUD295Bwdy7BBrIkImhc-ptNrU4OoCMKfYV-W-ZfMeq_vFBdyzWFZRVRP3s3kkvUh7KfiMHD31uxjHFsHaTEmNpNmBvDGuyBI73KmbSGQJM9C98sG0vi2mjN1XphQGYKDhborjjVzVL5vRQC9FmrJs3SENP1pznOTlpxmhPDNrZ52BpMerwlyVu2/s320/claude-code-api-recharge-14.jpg" width="320" /></a></div><div><br /></div><div><br /></div>先按 Copy key 鈕複製金鑰, 儲存到檔案 (例如 .env) 後按 Done 即可 :&nbsp;<br /><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhP5HJCXtNYofEpwJIapUn4hD6vHDR2T2-HBMPLXhoakIFWvD23A2aXImfSVyCz1PAguIte_gEtTKt5kXtJ_Uy2hMipDqSo5bLRSQtNBctrnTsSf3KXUUg39igwTtEaoXkqIxyKVQb56jS7PX_h2_1kFlPGAT2fojF9ebTa0GPXvPHCFCzMHnTE6ABh6Dpe/s1102/claude-code-api-recharge-15.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="577" data-original-width="1102" height="168" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhP5HJCXtNYofEpwJIapUn4hD6vHDR2T2-HBMPLXhoakIFWvD23A2aXImfSVyCz1PAguIte_gEtTKt5kXtJ_Uy2hMipDqSo5bLRSQtNBctrnTsSf3KXUUg39igwTtEaoXkqIxyKVQb56jS7PX_h2_1kFlPGAT2fojF9ebTa0GPXvPHCFCzMHnTE6ABh6Dpe/s320/claude-code-api-recharge-15.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiPmWDo9x_Xc9S3ncVEKtfXBfv0YRiqklVNsrOkLPPBt0aUiFPqYObXE8Uih8OK_qF8q5dEo6w0qSOIMu18zZhhNrtpVb7Ma0i7Ulmd1huFHgjirlhFsRUk1PjCUc8Uc39eOKFSmJAONWoyJQVpet2Bc8jYdfanvXDp2ERMlhcw41xdDGDs6RgH9Wn4BFwu/s1933/claude-code-api-recharge-16.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="760" data-original-width="1933" height="126" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiPmWDo9x_Xc9S3ncVEKtfXBfv0YRiqklVNsrOkLPPBt0aUiFPqYObXE8Uih8OK_qF8q5dEo6w0qSOIMu18zZhhNrtpVb7Ma0i7Ulmd1huFHgjirlhFsRUk1PjCUc8Uc39eOKFSmJAONWoyJQVpet2Bc8jYdfanvXDp2ERMlhcw41xdDGDs6RgH9Wn4BFwu/s320/claude-code-api-recharge-16.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>回 Dashboard 確認已儲值 10 美元無誤 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgl2lSAV6EKTfI8aK6pHLbutN5SlFHTO-PrW5lQnxxnH5oRnPpmFVWkmPNthcaEWE7l9CWVntkfkdEs1ZSt0shOJbVaFQT_FQkMLElmxLkHyWjneGgxnNL32MouLVajcUaZ17stS7wR0jQ7usxm1pVph2maq9F6uOAwQfpeIFUeZfX3cKqdZDo_fXReZ5Yh/s1970/claude-code-api-recharge-17.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="972" data-original-width="1970" height="158" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgl2lSAV6EKTfI8aK6pHLbutN5SlFHTO-PrW5lQnxxnH5oRnPpmFVWkmPNthcaEWE7l9CWVntkfkdEs1ZSt0shOJbVaFQT_FQkMLElmxLkHyWjneGgxnNL32MouLVajcUaZ17stS7wR0jQ7usxm1pVph2maq9F6uOAwQfpeIFUeZfX3cKqdZDo_fXReZ5Yh/s320/claude-code-api-recharge-17.jpg" width="320" /></a></div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-5973781746417652652026-06-02T10:13:16.990+08:002026-06-02T10:13:16.990+08:00momo 購買功率計<div>之前買的功率計遍尋不著, 只好上 momo 再買一個 :</div><div><br /></div><div>#&nbsp;<a href="https://www.momoshop.com.tw/goods/GoodsDetail.jsp?i_code=11062152" target="_blank">測冷氣耗電量 瓦數插座 高精密電流傳感器 功率計</a> $481</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgZOjLLIKmLV48bKMennMxRkEkSSnBwm2ab1FRCNBKfUwNzrBdAYx-fbPD23A5LJqQU_GNDMZ_pyHR1w2Rdx7WVuHUUwUYLejwH1OY_GRT4runpg-QB5gO6n5zJNXdXQr2Ao9WVGw4rP16u2ZVKwEI5ojFsveHFXbOpgAKxdfBan970gLj_oFW3_HsEpF3o/s1572/momo%20%E8%B3%BC%E8%B2%B7%E5%8A%9F%E7%8E%87%E8%A8%88.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="656" data-original-width="1572" height="134" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgZOjLLIKmLV48bKMennMxRkEkSSnBwm2ab1FRCNBKfUwNzrBdAYx-fbPD23A5LJqQU_GNDMZ_pyHR1w2Rdx7WVuHUUwUYLejwH1OY_GRT4runpg-QB5gO6n5zJNXdXQr2Ao9WVGw4rP16u2ZVKwEI5ojFsveHFXbOpgAKxdfBan970gLj_oFW3_HsEpF3o/s320/momo%20%E8%B3%BC%E8%B2%B7%E5%8A%9F%E7%8E%87%E8%A8%88.jpg" width="320" /></a></div><br /><div><br /></div><div>用掉 momo 幣與 momo 點, 實付 26 元.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-47731229973831316012026-05-31T23:37:19.743+08:002026-06-03T08:36:21.845+08:002026 年第 21 周記事<div>不知不覺已快到年中了, 回顧年初到現在, 學習速度雖然慢, 但一步一腳印, 堅持寫下學習紀錄, 有時回看也覺得有點佩服自己, 居然能有這毅力. 本周仍是專注於 Ollama 的學習, 之前有測試過 OpenAI API 的經驗, Ollama 的 API 其實更簡單, 很容易上手.&nbsp;</div><div><br /></div><div>上週去高雄家附近的家樂福超市買貓罐罐, 得知月底租約到期要關了, 衝到貓飼料架發現家樂福自有品牌的大小包貓糧都沒貨了, 店員說賣完為止不再進貨. 週四去澄清湖受訓, 下課後直衝大順店搶購, 買到兩包大的, 九包小的, 週五下班又去搬回四包大的, 應該夠貓咪吃半年了.&nbsp;</div><div><br /></div><div>週三爸從社區回來突然說以後不去了, 因為早上剛到社區在量血壓時, 一位工作人員打開它的餐具包翻找, 翻出一條運動帶, 說這個是社區的, 不要帶回去, 他因此覺得好像是偷拿東西一樣, 感覺很不好. 我詢問管事的江小姐, 她查詢後得知工作人員是要幫爸找乘車證, 沒有先打招呼就翻找引起誤會, 希望我向爸說明並道歉. 週三晚上照例回鄉下, 跟爸解釋後似乎理解了, 週四早上我要出發上班前他也說吃過早餐會在曬穀場等社區的車子, 但等我到高雄撥電話回家, 他又說昨天那件事, 決定以後不去社區了, 我從監視器也看到他在門口跟司機搖手說不去了, 我只好跟江小姐說那這兩天就改為中午送餐, 後續再努力勸他回社區. 結果周末爸又問今天社區的車怎還沒來, 哈哈, 有時會記得有時又會忘記.&nbsp;</div><div><br /></div><div><br /></div><div>2026-06-03 補充 :</div><div><br /></div><div>周一開始爸又恢復去社區了, 因為周末兩天偶而他想起被搜包包事件而抱怨時, 我說這是誤會啦, 工作人員已經被江小姐告誡了, 而且你沒搭車的話, 司機就沒業績, 無法獲得政府補助, 這招似乎有效. 上周四他突然想起這件事不去社區, 剛好我那天早上要去澄清湖受儲備幹部訓練, 經過仁武北極殿前等紅燈, 正在苦惱這件事時, 望見這小巧的廟宇便心中向神明祈求能順利化解, 果真本周便有轉機, 剛好今天阿中送我一盒他表弟從美國帶回來的 See's Candy, 下班後便帶此貴森森的巧克力與路邊買的鳳梨去北極殿答謝, 此廟供奉玄天上帝與註生娘娘, 廟雖小但內部繪畫雕刻極精美, 與廟祝閒聊約莫一炷香後才回家. 以後有路過再進去燒香.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-60524424091559001482026-05-31T20:05:40.468+08:002026-05-31T20:05:40.468+08:00Ollama 學習筆記 : 本地模型的函式呼叫<div>本篇旨在測試 ollama 套件的函式呼叫功能. 函式呼叫機制是透過請求訊息中的 tools 參數告訴模型, 如果需要存取外部資料時 (call-out) 時可呼叫那些函式. Ollama 官網中的模型如果帶有 tools 標籤, 表示該模型支援函式呼叫, 例如 gemma4 與 qwen3 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh8BzURauhH5JgXwn7fH4bX5NkY74qskgonE9gO0GM_UtmX4TAxpXRy8LO0qmvcg877URU8YLGh50csbvf5RckyZKvfQN6rcuamHGcjZGsr5KCQRer4Y5wt3B4_FQ-9A4qlcRD1d70LgnsTQzSrtRs-LfI3AAlsg3sh9Bh2_crKiNmBEBRfjX0ATrw48JYp/s1679/ollama-gemma4-tools.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="786" data-original-width="1679" height="150" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh8BzURauhH5JgXwn7fH4bX5NkY74qskgonE9gO0GM_UtmX4TAxpXRy8LO0qmvcg877URU8YLGh50csbvf5RckyZKvfQN6rcuamHGcjZGsr5KCQRer4Y5wt3B4_FQ-9A4qlcRD1d70LgnsTQzSrtRs-LfI3AAlsg3sh9Bh2_crKiNmBEBRfjX0ATrw48JYp/s320/ollama-gemma4-tools.jpg" width="320" /></a></div><br /><div><br /></div><div>本篇旨在測試本地模型的函式呼叫.&nbsp;</div><div><br /></div><div><div><div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div><div><br /></div><div>關於雲端模型函式呼叫可參考 OpenAI API 測試筆記 :</div><div><br /></div></div></div><div><div># <a href="https://yhhuang1966.blogspot.com/2025/04/openai-api-function-calling.html" target="_blank">OpenAI API 學習筆記 : 用 Function calling 讓模型開外掛 (一)</a></div><div># <a href="https://yhhuang1966.blogspot.com/2025/04/openai-api-function-calling_28.html" target="_blank">OpenAI API 學習筆記 : 用 Function calling 讓模型開外掛 (二)</a></div><div># <a href="https://yhhuang1966.blogspot.com/2025/05/openai-api-function-calling.html" target="_blank">OpenAI API 學習筆記 : 用 Function calling 讓模型開外掛 (三)</a></div><div># <a href="https://yhhuang1966.blogspot.com/2025/05/openai-api-function-calling_3.html" target="_blank">OpenAI API 學習筆記 : 用 Function calling 讓模型開外掛 (四)</a></div><div># <a href="https://yhhuang1966.blogspot.com/2025/05/openai-api-function-calling_4.html" target="_blank">OpenAI API 學習筆記 : 用 Function calling 讓模型開外掛 (五)</a>&nbsp;&nbsp;</div></div><div><br /></div><div>從之前的 OpenAI API 測試紀錄可知,&nbsp;函式呼叫時需傳入一個 tools 參數給 API, 其值為一個描述函式呼叫介面 (函式名稱與參數結構) 的字典串列 (稱為 JSON Schema), Ollama 的 ollama 套件也是如此, 但早期它在 ollama._utils 模組中提供了一個便利的 convert_function_to_tool(func) 函式可自動去解析函式的名稱, 型別標註以及 Docstring (函式說明文件), 然後將其轉換為 Ollama 核心看得懂的 JSON Schema.&nbsp;</div><div><br /></div><div>不過自從 ollama 0.4.0 版本之後已進一步優化, 不需要手動呼叫 convert_function_to_tool(func) 函式了,&nbsp;只要用標準的 Python 語法加上型別標註與 Google 風格的 Docstring 來撰寫函式, 然後將函式放進 tools 串列, ollama 套件會在幕後自動執行 convert_function_to_tool(), 毋須顯式呼叫它.&nbsp;</div></div><div><br /></div><div>先用 pip show 檢視目前安裝的 ollama 套件版本 :</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">pip show ollama</span></b>&nbsp;&nbsp;</div><div>Name: ollama</div><div>Version: <span style="background-color: #fcff01;">0.6.2</span></div><div>Summary: The official Python client for Ollama.</div><div>Home-page: https://ollama.com</div><div>Author:</div><div>Author-email: hello@ollama.com</div><div>License-Expression: MIT</div><div>Location: C:\Users\USER\AppData\Local\Programs\Python\Python312\Lib\site-packages</div><div>Requires: httpx, pydantic</div><div>Required-by:</div></div><div><br /></div><div>版本式最新的 0.6.2, 所以支援自動呼叫&nbsp;convert_function_to_tool() 功能.&nbsp;</div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 單一函式呼叫 :</span></b>&nbsp;</div><div><br /></div><div>下面是透過 tools 參數指定以函式呼叫來計算房屋總價的範例 :</div><div><br /></div><div>先匯入 ollama 套件 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import ollama</span></b>&nbsp;</div><div><br /></div><div>定義一個加上型別標註的房屋總價計算函式&nbsp;calculate_house_price(), 兩個傳入參數型別都是 float, 傳回值型別也是 float, 開頭為 Docstring 函式描述 (說明傳入參數意義) :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <span style="color: #2b00fe;"># 加上型別與說明文字的標準 Python 函式</span></div><div><b><span style="color: #2b00fe;">def calculate_house_price(ping: float, price_per_ping: float) -&gt; float:</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; """</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; 計算房屋總價（坪數 * 每坪單價）</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; Args:</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ping: 房屋的坪數大小（單位：坪）</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; price_per_ping: 每坪的單價（單位：萬元）</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; """</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; return round(ping * price_per_ping, 2)</span></b></div><div><br /></div><div>注意, Docstring 在函式呼叫中的角色很重要, 可讓模型看懂函式的參數意義, 確保能精準的對齊語意. 然後在呼叫 ollama.chat 時將函式名稱&nbsp;calculate_house_price 放進串列傳給 tools 參數即可, ollama 會在背後隱式地呼叫&nbsp;convert_function_to_tool() 來建立函式呼叫的 JSON schema :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=ollama.chat(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model='gemma4:e4b',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; messages=[{'role': 'user', 'content': '幫我算 30 坪, 每坪 60 萬的房子總價'}],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; tools=[calculate_house_price] # ollama 會呼叫 convert_function_to_tool()</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(reply)</span></b>&nbsp;&nbsp;</div></div><div>ChatResponse(model='gemma4:e4b', created_at='2026-05-31T08:56:18.1535995Z', done=True, done_reason='stop', total_duration=8857022300, load_duration=5518332400, prompt_eval_count=138, prompt_eval_duration=70437200, eval_count=223, eval_duration=2999849600, <span style="background-color: #fcff01;">message</span>=Message(role='assistant', <span style="background-color: #fcff01;">content=''</span>, thinking='The user wants to calculate the total price of a house given its size (30 ping) and the price per ping (60 million).\nThe available tool is `calculate_house_price`.\nThis tool requires two arguments: `ping` (the area in ping) and `price_per_ping` (the price per ping).\n\n1.&nbsp; **Identify `ping`**: The user stated "30 坪", so `ping` = 30.\n2.&nbsp; **Identify `price_per_ping`**: The user stated "每坪 60 萬". The tool description for `price_per_ping` specifies the unit is "萬元" (ten thousand NT dollars). Since 60萬 = 60 (in units of 萬元), `price_per_ping` = 60.\n\nI should call the `calculate_house_price` tool with these values.', images=None, tool_name=None, <span style="background-color: #fcff01;">tool_calls</span>=[<span style="background-color: #fcff01;">ToolCall</span>(function=Function(name='calculate_house_price', arguments={'ping': 30, 'price_per_ping': 60}))]), logprobs=None)</div><div><br /></div><div>觀察 API 傳回的&nbsp;ChatResponse 物件, 關鍵訊息放在 message 屬性中, 回應內容 content 為空字串,&nbsp; 這是因為模型認為現在的第一要務是去叫 Python 函式做計算而非聊天, 所以它沒有吐出任何回應. thinking 參數則紀錄了 gemma4 模型的思考鏈推理過程, 它理解使用者想計算房屋總價且手邊有 calculate_house_price 工具可用, 然後核對 Docstring 的兩個參數 (坪數與單價), 得到思考結論 :&nbsp; 應該用這組參數去呼叫工具. 然後它把函式打包成 ToolCall 物件, 而且參數都依照函式參數的型別標示轉成數值 (避免了型別錯誤).&nbsp;&nbsp;</div><div><br /></div><div>所以, 函式呼叫的第一階段目標就是取得回應訊息中的 message 鍵的 tool_calls 鍵 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(reply.message.tool_calls)</span></b>&nbsp;&nbsp;</div><div>[ToolCall(function=Function(name='calculate_house_price', arguments={'ping': 30, 'price_per_ping': 60}))]</div></div><div><br /></div><div>第二階段應用程式要根據&nbsp;tool_calls 鍵之值來研判是否要呼叫函式, 要的話就從裡面取出要執行的函式名稱與參數後呼叫它 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <span style="color: #2b00fe;"># 判斷是否有工具呼叫需求</span></div><div><b><span style="color: #2b00fe;">if reply.message.tool_calls:</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; for tool in reply.message.tool_calls:</span></b></div><div><span style="color: #2b00fe;"><b>&nbsp; &nbsp; &nbsp; &nbsp; func_name=tool.function.name&nbsp; &nbsp; &nbsp; </b>&nbsp;# 直接用 . 欄位名稱取得函式名稱</span></div><div><span style="color: #2b00fe;"><b>&nbsp; &nbsp; &nbsp; &nbsp; func_args=tool.function.arguments&nbsp; </b># 取得參數字典 {'ping': 30, 'price_per_ping': 60}</span><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; print(f"命令確定！即將執行地端函式: {func_name}")&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; if func_name == "calculate_house_price":</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; # 直接用 ** 語法解包傳入參數執行函式呼叫</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; final_ans=calculate_house_price(**func_args)</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; print(f"實體計算答案: {final_ans} 萬元")</span></b></div><div><br /></div><div>執行結果如下 :</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div><span style="color: #2b00fe;">命令確定！即將執行地端函式: calculate_house_price</span></div><div><span style="color: #2b00fe;">實體計算答案: 1800 萬元</span></div></div><div><br /></div><div>完整程式碼如下 :</div><div><br /></div><div><div># ollama_function_call_1.py</div><div>import ollama</div><div><br /></div><div># 加上型別與說明文字的標準 Python 函式</div><div>def calculate_house_price(ping: float, price_per_ping: float) -&gt; float:</div><div>&nbsp; &nbsp; """</div><div>&nbsp; &nbsp; 計算房屋總價（坪數 * 每坪單價）</div><div>&nbsp; &nbsp; Args:</div><div>&nbsp; &nbsp; &nbsp; &nbsp; ping: 房屋的坪數大小（單位：坪）</div><div>&nbsp; &nbsp; &nbsp; &nbsp; price_per_ping: 每坪的單價（單位：萬元）</div><div>&nbsp; &nbsp; """</div><div>&nbsp; &nbsp; return round(ping * price_per_ping, 2)</div><div><br /></div><div># 呼叫 ollama.chat() 傳入 tools 參數&nbsp;</div><div>reply=ollama.chat(</div><div>&nbsp; &nbsp; model='gemma4:e4b',</div><div>&nbsp; &nbsp; messages=[{'role': 'user', 'content': '幫我算 30 坪, 每坪 60 萬的房子總價'}],</div><div>&nbsp; &nbsp; tools=[calculate_house_price] # ollama 會呼叫 convert_function_to_tool()</div><div>&nbsp; &nbsp; )</div><div># 輸出工具呼叫需求</div><div>print(reply.message.tool_calls)</div><div># 判斷是否有工具呼叫需求</div><div>if reply.message.tool_calls:</div><div>&nbsp; &nbsp; for tool in reply.message.tool_calls:</div><div>&nbsp; &nbsp; &nbsp; &nbsp; func_name=tool.function.name&nbsp; &nbsp; &nbsp; &nbsp;# 直接用 . 欄位名稱取得函式名稱</div><div>&nbsp; &nbsp; &nbsp; &nbsp; func_args=tool.function.arguments&nbsp; # 取得參數字典 {'ping': 30, 'price_per_ping': 60}&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; print(f"命令確定！即將執行地端函式: {func_name}")&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; if func_name == "calculate_house_price":</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; # 直接用 ** 語法解包傳入參數執行函式呼叫</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; final_ans=calculate_house_price(**func_args)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; print(f"實體計算答案: {final_ans} 萬元")</div><div><br /></div><div>此範例因為只有單一個函式呼叫, 所以只需要用一個最簡單的條件判斷式即可,&nbsp;不需要建立函式對照表字典.&nbsp;</div></div><div><br /></div><div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">2. 多函式呼叫 :</span></b>&nbsp;</div><div><br /></div><div>下面是多函式呼叫的範例 :</div><div><br /></div><div><div># ollama_function_call_2.py</div><div>import ollama</div><div><br /></div><div>def add_two(a: int, b: int) -&gt; int:</div><div>&nbsp; &nbsp; """將兩數相加"""</div><div>&nbsp; &nbsp; return int(a) + int(b)</div><div><br /></div><div>def multiply_two(a: int, b: int) -&gt; int:</div><div>&nbsp; &nbsp; """將兩數相乘"""</div><div>&nbsp; &nbsp; return int(a) * int(b)</div><div><br /></div><div>available_funcs={"add_two": add_two, "multiply_two": multiply_two}</div><div>messages=[{"role": "user", "content": "請計算 123 加 456 是多少? 算完之後再將結果乘以 5."}]</div><div>print("正在將任務送給 Ollama...")</div><div><br /></div><div># 使用 while 迴圈以應付 AI 的多步思考</div><div>while True:</div><div>&nbsp; &nbsp; reply=ollama.chat(</div><div>&nbsp; &nbsp; &nbsp; &nbsp; model='gemma4:e4b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; messages=messages,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; tools=[add_two, multiply_two]&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; )</div><div>&nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; # 每次都必須把 AI 的回應（不管是想叫工具還是想說話）塞進歷史紀錄</div><div>&nbsp; &nbsp; messages.append(reply['message'])</div><div>&nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; # 檢查 AI 這輪是不是又想呼叫工具</div><div>&nbsp; &nbsp; if reply['message'].get('tool_calls'):</div><div>&nbsp; &nbsp; &nbsp; &nbsp; print("\n偵測到 AI 決定調用工具！")</div><div>&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; for tool in reply['message']['tool_calls']:</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; func_name=tool['function']['name']</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; func_args=tool['function']['arguments']</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; print(f"-&gt; AI 選擇了函式: {func_name} | 參數: {func_args}")</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; if func_name in available_funcs:</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; real_result=available_funcs[func_name](**func_args)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; print(f"[Python 執行結果]: {real_result}")</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; # 回傳工具呼叫結果，將結果存入記憶送回大腦，讓 AI 決定下一步</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; messages.append({</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 'role': 'tool',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 'name': func_name,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 'content': str(real_result)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; })</div><div>&nbsp; &nbsp; &nbsp; &nbsp; # 執行完本次工具後進入下一次迴圈，讓 AI 看看這個結果滿不滿意</div><div>&nbsp; &nbsp; &nbsp; &nbsp; continue&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp;&nbsp;</div><div>&nbsp; &nbsp; else:</div><div>&nbsp; &nbsp; &nbsp; &nbsp; # AI 不再需要呼叫任何工具，回應最終答案</div><div>&nbsp; &nbsp; &nbsp; &nbsp; print("\n最終答案整合完畢！")</div><div>&nbsp; &nbsp; &nbsp; &nbsp; print(f"AI 最終回覆: {reply['message']['content']}")</div><div>&nbsp; &nbsp; &nbsp; &nbsp; break # 跳出迴圈</div><div><br /></div></div><div>這個程式是一個典型的本地端 AI 代理透過一個 while 迴圈扮演 AI 大腦與地端 Python 函式之間的傳話筒與執行官的範例, 讓不擅長複雜數學的模型也能透過借用工具精準完成連鎖計算任務, 實現了多輪自動化工具呼叫機制.&nbsp;</div><div><br /></div><div>此處的提示詞 "請計算 123 加 456 是多少? 算完之後再將結果乘以 5" 是一個需要多步驟進行連鎖運算的複雜指令,&nbsp;此程式能讓 AI 自動拆解步驟並連續調用本地的 Python 函式來解決問題. 此程式先定義了兩個基礎數學函式 add_two() 與 multiply_two(), 並加上了型別標註與 Docstring 說明, 讓 ollama 套件能自動將其轉換為 AI 看得懂的工具規格. 然後利用 available_funcs 字典作為路由對照表, 將 AI 決定的函式名稱字串動態映射到真正的 Python 函式.&nbsp;</div><div><br /></div><div>while 迴圈為此程式的核心-多輪自動化控制,&nbsp;因為使用者的問題包含兩個步驟 (先加後乘), AI 無法一次給出最終答案, 所以必須靠迴圈來控制 :</div><div><ul style="text-align: left;"><li>第一輪迴圈 :<br />AI 辨識出需要加法, 吐出 add_two 請求, 函式執行完得到 579, 以 role: 'tool' 角色將結果存入歷史紀錄並進到下一輪.&nbsp;</li><li>第二輪迴圈 :<br />AI 拿到 579 後發現還有乘法任務, 再度吐出 multiply_two 請求, 函式執行完得到 2895 並將結果存入歷史紀錄並進到下一輪.&nbsp;</li><li>第三輪迴圈 : <br />AI 發現所有運算皆已完成不再觸發 tool_calls, 於是轉入 else 區塊生成最後回應結束程式.&nbsp;</li></ul></div><div>程式在迴圈中不斷執行 messages.append() 將 AI 的思考決定與本地 Python 函式執行結果存入對話紀錄, 這種滾動式記憶是確保 AI 大腦不會失憶, 能順利執行下一個正確步驟的關鍵.&nbsp;</div><div><br /></div><div>執行結果如下 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">%Run ollama_function_call_2.py</span></b>&nbsp;&nbsp;</div><div><div><span style="color: #2b00fe;">正在將任務送給 Ollama...</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">偵測到 AI 決定調用工具！</span></div><div><span style="color: #2b00fe;">-&gt; AI 選擇了函式: add_two | 參數: {'a': 123, 'b': 456}</span></div><div><span style="color: #2b00fe;">[Python 執行結果]: 579</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">偵測到 AI 決定調用工具！</span></div><div><span style="color: #2b00fe;">-&gt; AI 選擇了函式: multiply_two | 參數: {'a': 579, 'b': 5}</span></div><div><span style="color: #2b00fe;">[Python 執行結果]: 2895</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">最終答案整合完畢！</span></div><div><span style="color: #2b00fe;">AI 最終回覆: 123 加 456 的結果是 579。</span></div><div><span style="color: #2b00fe;">再將 579 乘以 5，結果是 2895。</span></div></div><div><br /></div><div>與 OpenAI API 的函式呼叫比起來, 不需要手刻函式呼叫的 JSON Schema 真是省事很多, 程式碼也變得非常簡潔.&nbsp;</div></div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-21705010155820424822026-05-31T07:59:13.643+08:002026-05-31T08:46:42.176+08:00Ollama 學習筆記 : 使用 OpenAI API 存取本地模型<div>在前一篇測試中, 我們使用 ollama 套件來呼叫 Ollama REST API 的 /api/chat 與 /api/generate 端點, 其參數格式與 OpenAI 的有些出入, 為了讓使用 OpenAI API 撰寫的 AI 應用程式可以不需要修改而能移植到存取本地模型的應用程式中, Ollama 內建了一個 OpenAI 相容轉接層, 提供 /v1/chat/completions 端點來處理此類請求. 本篇旨在測試此相容端點之用法.&nbsp;</div><div><br /></div><div><div><div>本系列全部測試文章索引參考 :&nbsp;&nbsp;</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div><div><br /></div></div></div><div>關於 OpenAI API 用法參考 :</div><div><br /></div><div>#&nbsp;<a href="https://yhhuang1966.blogspot.com/2025/03/openai-api_5.html" target="_blank">OpenAI API 學習筆記索引</a></div><div><br /></div><div>ollama 套件參數結構與 OpenAI API 參數結構之主要差異摘要如下 :</div><div><ul style="text-align: left;"><li>核心參數 :<br />極度相似但命名有別, model, messages, stream, role, content 這些核心參數是最基礎的對話欄位, Ollama 在設計時顯然參考了 OpenAI 的規範, 所以命名完全一樣.&nbsp;</li><li>進階參數 :<br />OpenAI 格式直接把這些模型參數 (控制隨機性與生成長度等) 扁平地放在最外層; 而 Ollama 則將其打包在一個名為 "options" 的字典子欄位裡.</li></ul></div><div>具體之差異對照如下表 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 149.011px;">&nbsp;參數功能</th> <th style="padding: 3px; width: 240.011px;">&nbsp;OpenAI 格式</th> <th style="padding: 3px; width: 240.011px;">&nbsp;Ollama 格式</th> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>參數擺放位置</b></td> <td style="padding: 3px;">&nbsp;直接平鋪在 JSON 的<b>最外層</b></td> <td style="padding: 3px;">&nbsp;內層的 <b>"options": { ... }</b> 子字典中</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>隨機性 (溫度)</b></td> <td style="padding: 3px;">&nbsp;"temperature": 0.7</td> <td style="padding: 3px;">&nbsp;"options": { "temperature": 0.7 }</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>限制輸出字數</b></td> <td style="padding: 3px;">&nbsp;"max_tokens": 100</td> <td style="padding: 3px;">&nbsp;"options": { "num_predict": 100 }</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>上下文記憶長度</b></td> <td style="padding: 3px;">&nbsp;"max_completion_tokens": 4096</td> <td style="padding: 3px;">&nbsp;"options": { "num_ctx": 4096 }</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>核心採樣 (Top P)</b></td> <td style="padding: 3px;">&nbsp;"top_p": 0.9</td> <td style="padding: 3px;">&nbsp;"options": { "top_p": 0.9 }</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>頻率懲罰 (避免重複)</b></td> <td style="padding: 3px;">&nbsp;"frequency_penalty": 1.1</td> <td style="padding: 3px;">&nbsp;"options": { "repeat_penalty": 1.1 }</td> </tr> <tr> <td style="padding: 3px;">&nbsp;<b>固定隨機種子</b></td> <td style="padding: 3px;">&nbsp;"seed": 42</td> <td style="padding: 3px;">&nbsp;"options": { "seed": 42 }</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 使用 requests 套件 :</span></b></div><div><br /></div><div>首先使用 requests 套件存取 Ollama 相容轉接層的 /v1/chat/completions 端點 :</div><div><br /></div><div>定義端點網址 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url="http://localhost:11434/v1/chat/completions"</span></b></div><div><br /></div><div>定義 HTTP 請求標頭, 這裡需在 Authorization 欄位用 Bearer 指定 OpenAI API 金鑰, 但這只是形式, 不需要去申請金鑰, 隨便填例如 ollama 即可 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">headers={</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "Content-Type": "application/json",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "Authorization": "<span style="background-color: #fcff01;">Bearer ollama</span>"&nbsp; # api_key 填任意值 (例如 ollama)</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div><div><br /></div><div>定義酬載, 依照 OpenAI API 格式, temperature 參數直接放在外層 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">payload={</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "model": "gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "messages": [</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請用 100 個字簡介量子糾纏"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; <span style="background-color: #fcff01;">"temperature": 0.3</span>,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "stream": False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div><div><br /></div><div>呼叫 requests.post() 傳入 headers 與 json 酬載提出請求, 傳回值為一個 Response 物件, 呼叫其 json() 方法將 JSON 字串轉成字典後即可在 messges 鍵的 content 鍵取得模型回應 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.post(url, headers=headers, json=payload)</span></b>&nbsp; &nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(result['choices'][0]['message']['content'])</span></b>&nbsp; &nbsp;</div><div><span style="color: #2b00fe;">量子糾纏是一種奇特的量子現象。當兩個或多個粒子發生糾纏時，它們會建立起一種超越空間的特殊連結。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">這意味著，即使將這些粒子分開到極遠的地方，只要你測量其中一個粒子的某個屬性（例如自旋），會瞬間、同步地決定另一個粒子的狀態。這種「幽靈般的超距作用」，讓它們的行為高度相關，彷彿心有靈通。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">糾纏現象是量子計算、量子通訊等尖端科技的基礎，是現代物理學最令人著迷的領域之一。</span></div></div><div><br /></div><div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">2. 使用 openai 套件 :</span></b></div></div><div><br /></div><div>使用 OpenAI API 需先安裝 openai 套件 (但不需要申請金鑰, 任意填即可) :</div><div><br /></div><div><b><span style="color: #2b00fe;">pip install openai</span></b>&nbsp; &nbsp;&nbsp;</div><div><br /></div><div>然後匯入 openai.OpenAI 類別, 傳入 base_url 與 api_key 參數建立 OpenAI 物件 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">from openai import OpenAI</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">client=OpenAI(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; base_url='http://localhost:11434/v1/',</span></b></div><div><span style="color: #2b00fe;"><b>&nbsp; &nbsp; </b><span style="background-color: #fcff01; font-weight: bold;">api_key='ollama'</span><b>&nbsp; </b># 填任意值</span></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div><br /></div><div>注意, 這裡 base_url 只要給端點的根目錄位置 http://localhost:11434/v1/ 即可, 具體要去哪個端點是由呼叫的函式決定的. 呼叫 OpenAI 物件的 chat.completions.create() 方法與指定之本地模型對話 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=client.chat.completions.create(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model="gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; messages=[</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請用 100 個字簡介量子糾纏"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; <span style="background-color: #fcff01;">temperature=0.3</span>,&nbsp; # OpenAI 格式參數直接放最外層</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; stream=False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div><br /></div><div>回應字串放在傳回物件&nbsp;ChatCompletion 的 choices[0].message.content 鍵裡 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(reply.choices[0].message.content)</span></b>&nbsp; &nbsp;</div><div><span style="color: #2b00fe;">量子糾纏是一種奇特的量子現象。當兩個或多個粒子發生糾纏後，它們會形成一種深層的量子連結。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">無論這些粒子相隔多遠，它們的物理狀態都是相互關聯的。當你測量其中一個粒子的屬性（例如自旋）時，另一個粒子的屬性會瞬間、同步地決定，彷彿它們心有靈犀一般。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">這種非定域的關聯性，超乎傳統物理學的預期。它不僅是理論概念，更是未來發展量子計算、量子通訊等尖端科技的關鍵資源。</span></div></div><div><br /></div><div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">3. 使用 ollama 套件 :</span></b></div></div><div><br /></div><div>上面測試的 Ollama 版本如下 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import ollama</span></b>&nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=ollama.chat(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model='gemma4:e4b',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; messages=[</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請用 100 個字簡介量子糾纏"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; stream=False,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; options={"temperature": 0.3}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(reply['message']['content'])</span></b>&nbsp;&nbsp;</div><div><div><span style="color: #2b00fe;">量子糾纏是一種奇特的量子現象，描述兩個或多個粒子之間存在著超越空間距離的特殊連結。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">當這些粒子發生糾纏後，它們的量子狀態會互相綁定。無論將它們分開到多遠，只要你測量其中一個粒子的屬性（例如自旋），你就能**瞬間**知道另一個粒子的狀態，即使它們相隔數光年。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">這就像一對心靈相通的「量子雙胞胎」，一個改變，另一個會同步改變。這現象是量子計算和量子通訊的基礎，曾被愛因斯坦戲稱為「鬼魅般的超距作用」。</span></div></div></div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-82315312583903531602026-05-30T22:58:35.476+08:002026-05-31T00:05:28.394+08:00Ollama 學習筆記 : REST API 用法 (三)<div>在前一篇測試中使用 requests 套件來呼叫 Ollama REST API 的端點, 我們須自行處理繁瑣的 HTTP 和資料格式解析, 本篇則是要使用 Ollama 官方的 ollama 套件來簡化處理程序, 擺脫 JSON 的序列化與反序列化麻煩. 在使用 requests 時, 我們必須自己用 json.loads() 將 JSON 字串字典, 如果開啟 stream=True 要寫迴圈, 做 decode('utf-8'), 與解析每行的 JSON 碎片, 並從巢狀結構中把字串挖出來, 非常麻煩. 使用 ollama 套件則有如下優點 :</div><div><ul style="text-align: left;"><li>非串流模式 :<br />直接回傳一個已經解析好的 Python 字典或物件.</li><li>串流模式 :<br />支援 Python 的生成器 (Generato), 直接用 for chunk in response 就能輕鬆讀取.</li></ul></div><div>本系列全部測試文章索引參考 :</div><div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div><div><br /></div><div><br /></div></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 安裝 ollama 套件 :&nbsp;</span></b></div><div><br /></div><div>ollama 為第三方套件, 需先用 pip 安裝 :</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">pip install ollama</span></b>&nbsp; &nbsp;</div><div>Collecting ollama</div><div>&nbsp; Downloading ollama-0.6.2-py3-none-any.whl.metadata (5.8 kB)</div><div>Requirement already satisfied: httpx&gt;=0.27 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from ollama) (0.27.0)</div><div>Requirement already satisfied: pydantic&gt;=2.9 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from ollama) (2.9.1)</div><div>Requirement already satisfied: anyio in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpx&gt;=0.27-&gt;ollama) (4.3.0)</div><div>Requirement already satisfied: certifi in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpx&gt;=0.27-&gt;ollama) (2024.2.2)</div><div>Requirement already satisfied: httpcore==1.* in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpx&gt;=0.27-&gt;ollama) (1.0.5)</div><div>Requirement already satisfied: idna in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpx&gt;=0.27-&gt;ollama) (3.7)</div><div>Requirement already satisfied: sniffio in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpx&gt;=0.27-&gt;ollama) (1.3.1)</div><div>Requirement already satisfied: h11&lt;0.15,&gt;=0.13 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from httpcore==1.*-&gt;httpx&gt;=0.27-&gt;ollama) (0.14.0)</div><div>Requirement already satisfied: annotated-types&gt;=0.6.0 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from pydantic&gt;=2.9-&gt;ollama) (0.7.0)</div><div>Requirement already satisfied: pydantic-core==2.23.3 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from pydantic&gt;=2.9-&gt;ollama) (2.23.3)</div><div>Requirement already satisfied: typing-extensions&gt;=4.6.1 in c:\users\user\appdata\local\programs\python\python312\lib\site-packages (from pydantic&gt;=2.9-&gt;ollama) (4.12.2)</div><div>Downloading ollama-0.6.2-py3-none-any.whl (15 kB)</div><div>Installing collected packages: ollama</div><div>Successfully installed ollama-0.6.2</div></div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">2. 常用函式 :&nbsp;</span></b></div><div><div><br /></div></div><div>用 dir() 檢視 ollama 套件內容 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import ollama</span></b>&nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">dir(ollama)</span></b>&nbsp; &nbsp;</div><div>['AsyncClient', 'ChatResponse', 'Client', 'EmbedResponse', 'EmbeddingsResponse', 'GenerateResponse', 'Image', 'ListResponse', 'Message', 'Options', 'ProcessResponse', 'ProgressResponse', 'RequestError', 'ResponseError', 'ShowResponse', 'StatusResponse', 'Tool', 'WebFetchResponse', 'WebSearchResponse', '__all__', '__builtins__', '__cached__', '__doc__', '__file__', '__loader__', '__name__', '__package__', '__path__', '__spec__', '_client', '_types', '_utils', 'chat', 'copy', 'create', 'delete', 'embed', 'embeddings', 'generate', 'list', 'ps', 'pull', 'push', 'show', 'web_fetch', 'web_search']</div></div><div><br /></div><div>ollama 套件常用函式說明如下表 :</div><div><br /></div><div><div><br /></div> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 99.011px;">&nbsp;ollama 函式</th> <th style="padding: 3px; width: 552.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;chat()</td> <td style="padding: 3px;">&nbsp;對應對話端點 /api/chat，傳入參數為 messages 串列與 stream (預設 True)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;generate()</td> <td style="padding: 3px;">&nbsp;對應文字生成端點 /api/generate，傳入參數為 prompt 字串。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;list()</td> <td style="padding: 3px;">&nbsp;列出本機已下載模型清單，對應 /api/tags 端點。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;show()</td> <td style="padding: 3px;">&nbsp;檢視模型詳情，對應 /api/show 端點，傳入參數為模型名稱。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;ps()</td> <td style="padding: 3px;">&nbsp;查看載入記憶體運行中模型，對應 /api/ps 端點。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;pull()</td> <td style="padding: 3px;">&nbsp;下載模型，對應 /api/pull 端點，傳入參數為模型名稱。若模型已存在則檢查是否為最新版。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;embed()</td> <td style="padding: 3px;">&nbsp;計算文字之嵌入向量，對應 /api/embed 端點 (支援單一或多個文字輸入)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;delete()</td> <td style="padding: 3px;">&nbsp;刪除模型釋放磁碟空間，對應 /api/delete 端點，傳入參數為模型名稱。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;create()</td> <td style="padding: 3px;">&nbsp;建立自訂模型，對應 /api/create 端點。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>先用 list() 取得已下載之模型清單 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">response=ollama.list()</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">type(response)</span></b>&nbsp; &nbsp; &nbsp;</div><div>&lt;class 'ollama._types.<span style="background-color: #fcff01;">ListResponse</span>'&gt;&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(response)</span></b>&nbsp;&nbsp;</div></div><div>models=[Model(model='gemma4:e4b', modified_at=datetime.datetime(2026, 5, 28, 22, 36, 46, 350248, tzinfo=TzInfo(+08:00)), digest='c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb', size=9608350718, details=ModelDetails(parent_model='', format='gguf', family='gemma4', families=['gemma4'], parameter_size='8.0B', quantization_level='Q4_K_M')), Model(model='llama3.2-vision:11b', modified_at=datetime.datetime(2026, 5, 25, 11, 30, 36, 32261, tzinfo=TzInfo(+08:00)), digest='6f2f9757ae97e8a3f8ea33d6adb2b11d93d9a35bef277cd2c0b1b5af8e8d0b1e', size=7816589186, details=ModelDetails(parent_model='', format='gguf', family='mllama', families=['mllama'], parameter_size='10.7B', quantization_level='Q4_K_M')), Model(model='phi4:14b', modified_at=datetime.datetime(2026, 5, 24, 17, 23, 24, 190702, tzinfo=TzInfo(+08:00)), digest='ac896e5b8b34a1f4efa7b14d7520725140d5512484457fab45d2a4ea14c69dba', size=9053116391, details=ModelDetails(parent_model='', format='gguf', family='phi3', families=['phi3'], parameter_size='14.7B', quantization_level='Q4_K_M')), Model(model='dagbs/deepseek-coder-v2-lite-instruct:q4_k_m', modified_at=datetime.datetime(2026, 5, 24, 0, 17, 21, 590078, tzinfo=TzInfo(+08:00)), digest='a6f5c73087ad25fc8666929492449eb0dc694326e4ca5b2313fef75b66645583', size=10364417401, details=ModelDetails(parent_model='', format='gguf', family='deepseek2', families=['deepseek2'], parameter_size='15.7B', quantization_level='Q4_K_M')), Model(model='mannix/deepseek-coder-v2-lite-instruct:q4_k_m', modified_at=datetime.datetime(2026, 5, 23, 21, 15, 32, 534056, tzinfo=TzInfo(+08:00)), digest='6171206208d0529a47806ebcf8ed37a88fe322859e269396dd16fdd98a56a102', size=10364432240, details=ModelDetails(parent_model='', format='gguf', family='deepseek2', families=['deepseek2'], parameter_size='15.7B', quantization_level='Q4_K_M')), Model(model='deepseek-r1:14b', modified_at=datetime.datetime(2026, 5, 23, 16, 43, 35, 552799, tzinfo=TzInfo(+08:00)), digest='c333b7232bdb521236694ffbb5f5a6b11cc45d98e9142c73123b670fca400b09', size=8988112209, details=ModelDetails(parent_model='', format='gguf', family='qwen2', families=['qwen2'], parameter_size='14.8B', quantization_level='Q4_K_M')), Model(model='qwen3:14b', modified_at=datetime.datetime(2026, 5, 22, 0, 30, 36, 339599, tzinfo=TzInfo(+08:00)), digest='bdbd181c33f2ed1b31c972991882db3cf4d192569092138a7d29e973cd9debe8', size=9276198565, details=ModelDetails(parent_model='', format='gguf', family='qwen3', families=['qwen3'], parameter_size='14.8B', quantization_level='Q4_K_M')), Model(model='gemma4:latest', modified_at=datetime.datetime(2026, 5, 20, 11, 45, 8, 247104, tzinfo=TzInfo(+08:00)), digest='c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb', size=9608350718, details=ModelDetails(parent_model='', format='gguf', family='gemma4', families=['gemma4'], parameter_size='8.0B', quantization_level='Q4_K_M'))]</div><div><br /></div><div>可見 list() 傳回的是一個&nbsp;ListResponse 物件, 可用迴圈取出關鍵資訊存入串列後轉成 DataFrame :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import pandas as pd</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">data=[]</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">for m in response.models:&nbsp;&nbsp;</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; data.append({</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; "Model Name": m.model,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; "Family": m.details.family,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; "Params": m.details.parameter_size,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; "Quant": m.details.quantization_level,</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; "Size (GB)": round(m.size / (1024 ** 3), 2)</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; })</span></b></div><div><br /></div><div>轉成 DatFrame :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">df=pd.DataFrame(data)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(df.to_string(index=False))&nbsp;</span></b> &nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;Model Name&nbsp; &nbsp; Family Params&nbsp; Quant&nbsp; Size (GB)</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;gemma4:e4b&nbsp; &nbsp; gemma4&nbsp; &nbsp;8.0B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;8.95</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; llama3.2-vision:11b&nbsp; &nbsp; mllama&nbsp; 10.7B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;7.28</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;phi4:14b&nbsp; &nbsp; &nbsp; phi3&nbsp; 14.7B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;8.43</div><div>&nbsp;dagbs/deepseek-coder-v2-lite-instruct:q4_k_m deepseek2&nbsp; 15.7B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;9.65</div><div>mannix/deepseek-coder-v2-lite-instruct:q4_k_m deepseek2&nbsp; 15.7B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;9.65</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; deepseek-r1:14b&nbsp; &nbsp; &nbsp;qwen2&nbsp; 14.8B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;8.37</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; qwen3:14b&nbsp; &nbsp; &nbsp;qwen3&nbsp; 14.8B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;8.64</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; gemma4:latest&nbsp; &nbsp; gemma4&nbsp; &nbsp;8.0B Q4_K_M&nbsp; &nbsp; &nbsp; &nbsp;8.95</div></div><div><br /></div><div>ollama.show() 傳回指定模型的資訊 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">ollama.show("gemma4:e4b")</span></b>&nbsp; &nbsp;</div><div>ShowResponse(modified_at=datetime.datetime(2026, 5, 28, 22, 36, 46, 350248, tzinfo=TzInfo(+08:00)), template='{{ .Prompt }}', modelfile='# Modelfile generated by "ollama show"\n# To build a new Modelfile based on this, replace FROM with:\n# FROM gemma4:e4b\n\nFROM D:\\OllamaModels\\blobs\\sha256-4c27e0f5b5adf02ac956c7322bd2ee7636fe3f45a8512c9aba5385242cb6e09a\nTEMPLATE {{ .Prompt }}\nRENDERER gemma4\nPARSER gemma4\nPARAMETER top_k 64\nPARAMETER top_p 0.95\nPARAMETER temperature 1\nLICENSE """&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; Apache License\n&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;Version 2.0, January 2004\n&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; http://www.apache.org/licenses/\n\n&nbsp; &nbsp;TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION\n\n&nbsp; &nbsp;1. Definitions.\n\n&nbsp; &nbsp; &nbsp; "License" shall mean the terms and conditions for use, reproduction,\n&nbsp; &nbsp; &nbsp; and distribution as defined by Sections 1 through 9 of this document.\n\n&nbsp; &nbsp; &nbsp; "Licensor" shall mean the copyright owner or entity authorized by\n&nbsp; &nbsp; &nbsp; the copyright owner that is granting the License.\n\n&nbsp; &nbsp; &nbsp; "Legal Entity" shall mean the union of the acting entity and all\n&nbsp; &nbsp; &nbsp; other entities that control, are controlled by, or are under common\n&nbsp; &nbsp; &nbsp; control with that entity. For the purposes of this definition,\n&nbsp; &nbsp; &nbsp; "control" means (i) the power, direct or indirect, to cause the\n&nbsp; &nbsp; &nbsp; direction or management of such entity, whether by contract or\n&nbsp; &nbsp; &nbsp; otherwise, or (ii) ownership of fifty percent (50%) or more of the\n&nbsp; &nbsp; &nbsp; outstanding shares, or (iii) beneficial ownership of such entity.\n\n&nbsp; &nbsp; &nbsp; "You" (or "Your") shall mean an individual or Legal Entity\n&nbsp; &nbsp; &nbsp; exercising permissions granted by this License.\n\n&nbsp; &nbsp; &nbsp; "Source" form shall mean the preferred form for making modifications,\n&nbsp; &nbsp; &nbsp; including but not limited to software source code, documentation\n&nbsp; &nbsp; &nbsp; source, and configuration files.\n\n&nbsp; &nbsp; &nbsp; "Object" form shall mean any form resulting from mechanical\n&nbsp; &nbsp; &nbsp; transformation or translation of a Source form, including but\n&nbsp; &nbsp; &nbsp; not limited to compiled object code, generated documentation,\n&nbsp; &nbsp; &nbsp; and conversions to other media types.\n\n&nbsp; &nbsp; &nbsp; "Work" shall mean the work of authorship, whether in Source or\n&nbsp; &nbsp; &nbsp; Object form, made available under the License, as indicated by a\n&nbsp; &nbsp; &nbsp; copyright notice that is included in or attached to the work\n&nbsp; &nbsp; &nbsp; (an example is provided in the Appendix below).\n\n&nbsp; &nbsp; &nbsp; "Derivative Works" shall mean any work, whether in Source or Object\n&nbsp; &nbsp; &nbsp; form, that is based on (or derived from) the Work and for which the\n&nbsp; &nbsp; &nbsp; editorial revisions, annotations, elaborations, or other modifications\n&nbsp; &nbsp; &nbsp; represent, as a whole, an original work of authorship. For the purposes\n&nbsp; &nbsp; &nbsp; of this License, Derivative Works shall not include works that remain\n&nbsp; &nbsp; &nbsp; separable from, or merely link (or bind by name) to the interfaces of,\n&nbsp; &nbsp; &nbsp; the Work and Derivative Works thereof.\n\n&nbsp; &nbsp; &nbsp; "Contribution" shall mean any work of authorship, including\n&nbsp; &nbsp; &nbsp; the original version of the Work and any modifications or additions\n&nbsp; &nbsp; &nbsp; to that Work or Derivative Works thereof, that is intentionally\n&nbsp; &nbsp; &nbsp; submitted to Licensor for inclusion in the Work by the copyright owner\n&nbsp; &nbsp; &nbsp; or by an individual or Legal Entity authorized to submit on behalf of\n&nbsp; &nbsp; &nbsp; the copyright owner. For the purposes of this definition, "submitted"\n&nbsp; &nbsp; &nbsp; means any form of electronic, verbal, or written communication sent\n&nbsp; &nbsp; &nbsp; to the Licensor or its representatives, including but not limited to\n&nbsp; &nbsp; &nbsp; communication on electronic mailing lists, source code control systems,\n&nbsp; &nbsp; &nbsp; and issue tracking systems that are managed by, or on behalf of, the\n&nbsp; &nbsp; &nbsp; Licensor for the purpose of discussing and improving the Work, but\n&nbsp; &nbsp; &nbsp; excluding communication that is conspicuously marked or otherwise\n&nbsp; &nbsp; &nbsp; designated in writing by the copyright owner as "Not a Contribution."\n\n&nbsp; &nbsp; &nbsp; "Contributor" shall mean Licensor and any individual or Legal Entity\n&nbsp; &nbsp; &nbsp; on behalf of whom a Contribution has been received by Licensor and\n&nbsp; &nbsp; &nbsp; subsequently incorporated within the Work.\n\n&nbsp; &nbsp;2. Grant of Copyright License. Subject to the terms and conditions of\n&nbsp; &nbsp; &nbsp; this License, each Contributor hereby grants to You a perpetual,\n&nbsp; &nbsp; &nbsp; worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n&nbsp; &nbsp; &nbsp; copyright license to reproduce, prepare Derivative Works of,\n&nbsp; &nbsp; &nbsp; publicly display, publicly perform, sublicense, and distribute the\n&nbsp; &nbsp; &nbsp; Work and such Derivative Works in Source or Object form.\n\n&nbsp; &nbsp;3. Grant of Patent License. Subject to the terms and conditions of\n&nbsp; &nbsp; &nbsp; this License, each Contributor hereby grants to You a perpetual,\n&nbsp; &nbsp; &nbsp; worldwide, non-exclusive, no-charge, royalty-free, irrevocable\n&nbsp; &nbsp; &nbsp; (except as stated in this section) patent license to make, have made,\n&nbsp; &nbsp; &nbsp; use, offer to sell, sell, import, and otherwise transfer the Work,\n&nbsp; &nbsp; &nbsp; where such license applies only to those patent claims licensable\n&nbsp; &nbsp; &nbsp; by such Contributor that are necessarily infringed by their\n&nbsp; &nbsp; &nbsp; Con…</div><div><br /></div><div>最常用的是 generate() 與 chat() 函式.&nbsp;</div><div><br /></div><div>generate() 的功能為單次文字接龍, 給它一段提示詞就會直接回應答案. generate() 不適合用來做需要記住來回歷史紀錄的聊天機器人, 但非常適合用來做自動化任務, 文章寫作, 程式碼補全, 或資料格式化等單次搞定的工作. 必要的傳入參數為用來指定模型的 model 與提示詞 prompt, 提示詞若傳入空字串時會觸發檢查模型是否載入, 若尚未載入就會強迫載入運行, 有 ollama run 的效果 :</div><div><br /></div><div>當久未呼叫 API (預設 5 分鐘), Ollama 會自動將模型從記憶體中卸載, 這時呼叫 ollama.ps() 會傳回一個 model 屬性值為空串列的物件 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">ollama.ps()</span></b>&nbsp;&nbsp;</div><div>ProcessResponse(models=[])</div></div><div><br /></div><div>然後呼叫 ollma.generate() 並傳入模型名稱與空提示詞 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=ollama.generate(model='gemma4:e4b', prompt='')</span></b>&nbsp;&nbsp;</div><div><br /></div><div>這時再次呼叫 ollama.ps() 就會顯示目前記憶體中已載入指定之模型 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">ollama.ps()</span></b>&nbsp; &nbsp;</div><div>ProcessResponse(models=[Model(model='gemma4:e4b', name='gemma4:e4b', digest='c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb', expires_at=datetime.datetime(2026, 5, 30, 22, 47, 1, 112747, tzinfo=TzInfo(+08:00)), size=10579079040, size_vram=10579079040, details=ModelDetails(parent_model='', format='gguf', family='gemma4', families=['gemma4'], parameter_size='8.0B', quantization_level='Q4_K_M'), context_length=4096)])</div></div><div><br /></div><div>檢視回應物件會發現&nbsp;done_reason 是 'load' (載入模型) 而非 'stop' (回應結束), 這是因為提示詞空字串, 模型無法做出回應, 只是將模型載入記憶體而已 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply</span></b>&nbsp;&nbsp;</div><div>GenerateResponse(model='gemma4:e4b', created_at='2026-05-30T14:42:01.1127475Z', done=True, <span style="background-color: #fcff01;">done_reason='load'</span>, total_duration=None, load_duration=None, prompt_eval_count=None, prompt_eval_duration=None, eval_count=None, eval_duration=None, response='', thinking=None, context=None, logprobs=None, image=None, completed=None, total=None)</div></div><div><br /></div><div>當然, 如果提示詞不為空, 則會先載入模型, 然後做出回應.&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=ollama.generate(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model='gemma4:e4b',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; prompt='你是誰?'</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )&nbsp;</span></b>&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply</span></b>&nbsp;&nbsp;</div></div><div>GenerateResponse(model='gemma4:e4b', created_at='2026-05-30T14:53:45.7277926Z', done=True, done_reason='stop', total_duration=5928727000, load_duration=252832000, prompt_eval_count=19, prompt_eval_duration=260057900, eval_count=386, eval_duration=5196020600, <span style="background-color: #fcff01;">response</span>='我是 **Gemma 4**。\n\n我是一個由 Google DeepMind 開發的大型語言模型 (Large Language Model)。\n\n我的設計目的是理解和生成人類語言，我可以回答問題、撰寫文本、總結資訊，並協助您進行各種知識性的任務。\n\n簡而言之，我是一個用來與您互動、提供資訊和協助您完成任務的 AI 模型。', thinking=None, context=[2, 105, 9731, 107, 98, 107, 106, 107, 105, 2364, 107, 95841, 240560, 236881, 106, 107, 105, 4368, 107, 100, 45518, 107, 120474, 12364, 236787, 108, 236770, 236761, 138, 1018, 137938, 506, 2430, 236789, 236751, 5192, 532, 7609, 53121, 669, 2430, 19565, 623, 95841, 240560, 7462, 568, 236797, 240622, 127880, 704, 155268, 78546, 837, 55544, 531, 623, 15938, 659, 611, 7462, 107, 236778, 236761, 138, 1018, 102752, 506, 19080, 236786, 36809, 53121, 564, 1921, 8932, 3894, 531, 1041, 6697, 9834, 6366, 236761, 107, 140, 236829, 139, 1567, 236787, 147224, 236743, 236812, 236761, 107, 140, 236829, 139, 96089, 236787, 108388, 684, 6475, 22267, 65153, 236761, 107, 140, 236829, 139, 46797, 236787, 25093, 22160, 9483, 568, 2182, 236792, 769, 107, 140, 236829, 139, 2328, 236787, 7607, 18710, 2028, 236761, 107, 236800, 236761, 138, 1018, 40414, 7157, 9834, 3298, 1131, 8555, 568, 63190, 653, 164557, 236764, 10264, 496, 12643, 236764, 54651, 15737, 1473, 1018, 107, 140, 236829, 139, 236775, 236777, 1006, 147224, 236743, 236812, 1781, 3921, 26911, 237026, 147224, 236743, 236812, 236924, 107, 140, 236829, 139, 236775, 236777, 1006, 496, 2455, 5192, 2028, 1781, 3921, 26911, 90432, 56762, 146569, 26609, 568, 236824, 241747, 127880, 166012, 236759, 237022, 110652, 236781, 122720, 185960, 236762, 3238, 35843, 236781, 122720, 45511, 107, 140, 236829, 139, 236775, 165684, 684, 6475, 22267, 65153, 1781, 3921, 26911, 237852, 6475, 22267, 65153, 65706, 238623, 568, 236824, 241747, 570, 236916, 236756, 6475, 22267, 65153, 54181, 584, 237448, 45511, 107, 140, 236829, 139, 236769, 43983, 840, 1535, 5428, 236787, 127297, 506, 1932, 236772, 38357, 4135, 2907, 3921, 26911, 90432, 237857, 238076, 240551, 101462, 26609, 568, 236777, 1006, 614, 1932, 236772, 6078, 2028, 769, 107, 236812, 236761, 138, 1018, 64477, 506, 1626, 3890, 53121, 70535, 1239, 3298, 1131, 496, 40137, 236764, 54651, 236764, 532, 13611, 3072, 528, 8555, 236761, 108, 16907, 25864, 236772, 135778, 236786, 7166, 29357, 1473, 41152, 506, 15737, 563, 40564, 532, 1982, 9603, 108, 236810, 236761, 138, 1018, 17667, 8555, 14503, 32955, 99382, 101, 44889, 5213, 236823, 12367, 236743, 236812, 1018, 236924, 108, 237169, 90432, 237852, 6475, 22267, 65153, 65706, 238623, 29854, 237731, 146569, 26609, 568, 31534, 22160, 9483, 45511, 108, 21480, 34611, 159605, 29666, 237206, 25352, 126592, 146569, 236900, 183868, 49695, 18053, 236951, 242761, 240564, 57489, 236951, 240018, 238331, 89998, 236900, 238953, 181143, 238602, 43682, 77722, 77880, 39146, 152807, 236924, 108, 239309, 63229, 237437, 236900, 237169, 90432, 237105, 237967, 238693, 238602, 206182, 236951, 12680, 89998, 237206, 181143, 238602, 19843, 152807, 236918, 12498, 228546, 236924], logprobs=None, image=None, completed=None, total=None)</div><div><br /></div><div>可見回應是放在 response 屬性裡 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(reply.response)</span></b>&nbsp; &nbsp; # 或 reply['response'] 亦可</div><div>我是 **Gemma 4**。</div><div><br /></div><div>我是一個由 Google DeepMind 開發的大型語言模型 (Large Language Model)。</div><div><br /></div><div>我的設計目的是理解和生成人類語言，我可以回答問題、撰寫文本、總結資訊，並協助您進行各種知識性的任務。</div><div><br /></div><div>簡而言之，我是一個用來與您互動、提供資訊和協助您完成任務的 AI 模型。</div></div><div><br /></div><div>串流用法是傳入 stream=True, 這會讓回應像打字機那樣一個字一個字出現, 例如 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">stream=ollama.generate(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model='gemma4:e4b',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; prompt='請用一句話形容春天?',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; stream=True</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">type(stream)</span></b>&nbsp;</div><div>&lt;class '<span style="background-color: #fcff01;">generator</span>'&gt;</div><div><br /></div><div>可見串流的回應是一個生成器, 可用迴圈將內容一個字一個字取出來 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">for chunk in stream:</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; print(chunk['response'], end='', flush=True)</span></b></div><div>&nbsp; &nbsp;&nbsp;</div><div><span style="color: #2b00fe;">由於「形容」是一個主觀的過程，我提供幾個不同風格的選項，讓您可以選擇最符合您心情的句子。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">---</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 🌷 意境詩歌型（強調氛圍、柔美）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">&gt; **春天是萬物甦醒時，最輕盈的暖意。**</span></div><div><span style="color: #2b00fe;">&gt; （或：春天是帶著希望，甦醒在萬物深處的溫柔。）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### ✨ 生動描寫型（強調活力、色彩）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">&gt; **春天是色彩灑滿大地，重新燃起生命躍動的季節。**</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 🕊️ 意象比喻型（強調轉變、希望）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">&gt; **春天，是大地從沉睡中醒來，換上了一身嫩綠夢衣的時刻。**</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 💖 簡潔有力型（最直白、直接）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">&gt; **春天，是充滿希望和生機的季節。**</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">***</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**💡 建議：** 如果想要最能概括春天的「氣質」，我會推薦：「**春天是萬物甦醒時，最輕盈的暖意。**」</span></div></div><div><br /></div><div>在多輪連續對話應用 (例如聊天機器人) 中會使用 ollama.chat() 而非 ollama.generate(), 它包含對話歷史紀錄的訊息串列, 只需要在 Python 中維護一個串列來記錄歷史對話, 並將每次的新對話附加進去, 模型就能擁有上下文記憶了.&nbsp;</div><div><br /></div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=ollama.chat(</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; model='gemma4:e4b',</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; messages=[</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "你是美食專家, 請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請列出台灣最知名的十大美食"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ]</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; )</span></b></div><div><br /></div><div>回應內容放在 message 鍵的 content 鍵裡 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(reply['message']['content'])</span></b>&nbsp;&nbsp;</div><div><span style="color: #2b00fe;">👋 嗨，各位美食愛好者！您可算是問對人了。身為一個從小吃到大、跟台灣巷弄小吃一路走來的美食專家，我必須跟您說，台灣這片土地，簡直就是「吃」的藝術品！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">不過，要列出「十大」實在是天勾地鉤，因為台灣的美食幾乎沒有盡頭！但如果硬要選十個**最具代表性、最能代表台灣味**，又讓外地朋友一聽就能聯想到台灣的招牌，那這十個絕對沒錯！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">我會把這些美食用最在地、最興奮的台灣口吻，為您好好介紹一下，請準備好您的胃，跟你的味蕾一起接受一次「爆炸性的旅行」吧！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">---</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">## 🍜✨ 台灣傳說！必吃十大招牌美食 🥢🇹🇼</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">（這裡我會混合包含主食、點心、飲料，讓您體驗台灣餐飲的廣度！）</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 【主食與心靈慰藉系】</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🥇 1. 滷肉飯 (Braised Pork Rice)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：它就是台灣人的「精神食糧」。上層那香氣四溢、滷得軟爛、化不開的五花肉，搭配台灣獨有的醬香，配上一碗熱騰騰的米飯，幸福感直接拉滿。它簡潔到不可思議，卻有辦法讓您感到最踏實的滿足。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 找一家肉燥油底不會太鹹的店，才是真高手！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🥈 2. 台灣牛肉麵 (Taiwanese Beef Noodle Soup)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：台灣牛肉麵那燉湯的底子，厲害到一個層次。從湯頭的醇厚、牛肉的Q彈到麵體的吸汁能力，它已經超越了「碗麵」，更像是一道具有文化底蘊的「療癒藝術品」。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 記得一定要點乾辣椒和醋，讓那層豐富的層次感被釋放出來！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🥉 3. 小籠包 (Xiao Long Bao)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：這已經不是一個「點心」了，它是一道國際化的「品牌代表」。那薄如紙片、咬下去一口湯汁爆開的瞬間，搭配肉香與薑絲的點睛之筆，是每個外地遊客都無法免疫的誘惑。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 重點是！吃之前一定要用熱油燙一下，讓那表皮酥脆度提升一個檔次！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">---</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 【街邊小吃與罪惡誘惑系】</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🏅 4. 蚵仔煎 (Oyster Omelet)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：台灣街頭小吃的靈魂！軟嫩的蛋皮、Q彈到爆的蚵仔、甜鹹交織的醬汁，幾張食材完美地結合在一起。它的味道是那種「簡單，但又極其完美」的口感。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 建議搭配一份清爽的酸梅汁，解膩又開胃，CP值超級高！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🎖️ 5. 臭豆腐 (Stinky Tofu)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：它就是一個「挑戰味蕾」的極品！雖然氣味超強，但一旦放進嘴裡，那外酥內嫩、吸飽了肉汁的口感，配上一床酸甜的泡菜，會讓你產生一種「咦？其實超好吃耶！」的驚喜感。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 這道美食永遠不會讓您覺得無聊，它總能帶來一場味覺的騷動！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**⭐ 6. 雞排 (Taiwanese Chicken Cutlet)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：體積巨大、外皮酥炸到金黃酥脆，內裡肉質卻保持著夠汁的彈性。它完美詮釋了台灣小吃那種「讓人吃完超飽，但又一點也不覺得罪惡」的魔法。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 選擇帶有蒜香粉或泰式粉料的店家，更能提升風味層次。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">---</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 【甜點、飲料與特殊風味系】</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**💎 7. 珍珠奶茶 (Bubble Tea)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：雖然它不是傳統的台式小吃，但它已經是台灣在全球最「出名」的文化輸出之一。那Q彈黏牙的波霸珍珠，搭配奶茶特有的香氣，給人一種隨時都能享受的甜蜜能量。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 建議搭配幾塊小零食一起享用，讓甜度和鹹味互相平衡。</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🧁 8. 鳳梨酥 (Pineapple Cake)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：這不只是一種糕點，更是一種台灣的伴手禮代表。那酥皮的鬆化感、搭配一顆晶瑩剔透、酸甜適中的鳳梨果肉，完美代表了「甜而不膩，有層次感」的台灣風味。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 購買時，可以多留意一下果肉和奶油的比例，有時會發現風味大升級！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**🍎 9. 芒果冰 (Mango Dessert/Mango Shaved Ice)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：尤其在芒果季節，這幾乎是台灣夏天的代名詞。酸甜濃郁的芒果香氣，搭配冰涼的刨冰口感，是給人視覺和味覺雙重降溫的藝術品。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 盡量選擇當季、果肉新鮮度爆高的店家，那香氣才能夠「炸」到您心房！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">**✨ 10. 飯糰／早點心 (Rice Balls / Traditional Breakfast)**</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**為什麼經典？**：這代表了台灣人那種「對生活最樸實、最熱情」的早晨。不論是配上鹹肉鬆的飯糰，還是店家現烤的麵包，它總是以最踏實的碳水化合物，讓您的一天從最在地的地方開始啟動。</span></div><div><span style="color: #2b00fe;">*&nbsp; &nbsp;**專家心得：** 台灣的早點心，是承載了時間與人文的味道，請一定要配上一杯熱咖啡或豆漿！</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">---</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">### 💡 總結給您的專家提醒：</span></div><div><span style="color: #2b00fe;"><br /></span></div><div><span style="color: #2b00fe;">各位，這十樣美食只是冰山一角！真正代表台灣的，是那**「巷弄深處，攤販呼喚聲」**裡散發出來的熱情和獨特性。下次您來台灣，除了吃這十樣，記得也要隨時留意那些不起眼、卻散發著濃郁誘惑力的街邊小攤，那才是最最真實的「台灣胃」味！🍽️❤️</span></div></div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-43601757192929649102026-05-30T09:48:51.666+08:002026-05-30T17:09:30.981+08:00Ollama 學習筆記 : REST API 用法 (二)<div>在前一篇測試中, 我們在 PS 視窗裡使用 curl.exe 程式向 Ollama 建立的 REST API 伺服器提出請求, 本篇則是要改用 Python 程式來存取 REST API 端點 (使用 requests 套件).&nbsp;</div><div><div><br /></div><div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div></div><div><br /></div><div><div>茲將常用之 Ollama REST API 端點鈔錄如下 :</div><div><br /></div><div><br /></div></div> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;POST 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/generate</td> <td style="padding: 3px;">&nbsp;文字生成 (用於單次提示詞的輸入與基本文字接龍)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/chat</td> <td style="padding: 3px;">&nbsp;多輪對話 (用於需要記錄上下文, 用 user/assistant/system 區分角色)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/embeddings</td> <td style="padding: 3px;">&nbsp;使用 Embedding 模型將文字轉成可用於 RAG 語意搜尋的嵌入向量。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/pull</td> <td style="padding: 3px;">&nbsp;從 Ollama 官方模型庫下載指定的 LLM 模型到本機。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/push</td> <td style="padding: 3px;">&nbsp;將模型推入 Ollama 官方模型庫 (需登入 &amp; 輸入金鑰)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/show</td> <td style="padding: 3px;">&nbsp;顯示模型的詳細資訊。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/create</td> <td style="padding: 3px;">&nbsp;透過傳入 Modelfile 的內容從本機檔案建立或自訂一個全新的模型。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/copy</td> <td style="padding: 3px;">&nbsp;在本地將一個現有的模型複製並重新命名為另一個名稱。</td> </tr> </tbody> </table> </div><div><br /></div><div> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;GET 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/tags</td> <td style="padding: 3px;">&nbsp;列出本地所有模型。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/ps</td> <td style="padding: 3px;">&nbsp;查看目前有哪些模型正載入在記憶體 (DRAM/VRAM)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/version</td> <td style="padding: 3px;">&nbsp;取得 Ollama 的版本資訊。</td> </tr> </tbody> </table> </div><div><br /></div><div> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;DELETE 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/delete</td> <td style="padding: 3px;">&nbsp;刪除本地指定之模型 (使用 name 鍵指定)。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div><span style="background-color: #fce5cd; color: #990000; font-weight: 700;">2. 使用 requests&nbsp; 提出請求 :</span></div><div><br /></div><div>requests 套件提供 get(), post(), 與 delete() 等相對於 GET, POST, 與 DELETE 方法的函式, 首先來測試 GET 方法, 例如呼叫 /a[i/tags 端點來取得函式清單. 由於 get() 無法指定模型, 因此需先手動載入模型, 開啟一個 CMD 或 PS 視窗, 輸入下列指令載入執行 gemma4:e4b 模型 :&nbsp;</div><div><br /></div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">ollama run gemma4:e4b</span></b>&nbsp;&nbsp;</div><div><br /></div><div>然後進入 Python 執行環境匯入 requests 套件, 呼叫 requests.get() 函式向 /api/tags 端點提出 GET 請求, requests 套件會將 Ollama 回應的 HTTP 封包打包成一個 Response 物件傳回 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import requests</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/tags'</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.get(url)</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply</span></b>&nbsp; &nbsp;&nbsp;</div><div>&lt;Response [200]&gt;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">type(reply)</span></b>&nbsp;</div><div>&lt;class 'requests.models.<span style="background-color: #fcff01;">Response</span>'&gt;&nbsp;&nbsp;</div></div><div><br /></div><div>可呼叫 Response 物件的 json() 方法轉成字典, 為了整齊顯示字典結構, 可用 pprint.print() 來輸出 :&nbsp;</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">from pprint import pprint</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp; &nbsp;</div><div>{'models': [{'details': {'families': ['gemma4'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'gemma4',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '8.0B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">gemma4:e4b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-28T22:36:46.3502486+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'gemma4:e4b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 9608350718},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['mllama'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'mllama',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '10.7B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': '6f2f9757ae97e8a3f8ea33d6adb2b11d93d9a35bef277cd2c0b1b5af8e8d0b1e',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">llama3.2-vision:11b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-25T11:30:36.0322613+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'llama3.2-vision:11b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 7816589186},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['phi3'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'phi3',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '14.7B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'ac896e5b8b34a1f4efa7b14d7520725140d5512484457fab45d2a4ea14c69dba',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">phi4:14b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-24T17:23:24.1907029+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'phi4:14b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 9053116391},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['deepseek2'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'deepseek2',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '15.7B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'a6f5c73087ad25fc8666929492449eb0dc694326e4ca5b2313fef75b66645583',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">dagbs/deepseek-coder-v2-lite-instruct:q4_k_m</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-24T00:17:21.5900783+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'dagbs/deepseek-coder-v2-lite-instruct:q4_k_m',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 10364417401},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['deepseek2'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'deepseek2',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '15.7B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': '6171206208d0529a47806ebcf8ed37a88fe322859e269396dd16fdd98a56a102',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">mannix/deepseek-coder-v2-lite-instruct:q4_k_m</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-23T21:15:32.5340567+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'mannix/deepseek-coder-v2-lite-instruct:q4_k_m',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 10364432240},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['qwen2'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'qwen2',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '14.8B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'c333b7232bdb521236694ffbb5f5a6b11cc45d98e9142c73123b670fca400b09',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">deepseek-r1:14b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-23T16:43:35.5527992+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'deepseek-r1:14b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 8988112209},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['qwen3'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'qwen3',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '14.8B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'bdbd181c33f2ed1b31c972991882db3cf4d192569092138a7d29e973cd9debe8',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">qwen3:14b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-22T00:30:36.3395998+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'qwen3:14b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 9276198565},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; {'details': {'families': ['gemma4'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'gemma4',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '8.0B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">gemma4:latest</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'modified_at': '2026-05-20T11:45:08.2471048+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'gemma4:latest',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 9608350718}]}</div></div><div><br /></div><div>呼叫 /api/ps 端點會傳回目前載入記憶體之模型資訊 (相當於下 ollama ps 指令) :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/ps'</span></b>&nbsp; &nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.get(url)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp;&nbsp;</div><div>{'models': [{'context_length': 4096,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'details': {'families': ['gemma4'],</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'family': 'gemma4',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'format': 'gguf',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parameter_size': '8.0B',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'parent_model': '',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'quantization_level': 'Q4_K_M'},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'digest': 'c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'expires_at': '2026-05-30T08:51:38.2129288+08:00',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'model': '<span style="background-color: #fcff01;">gemma4:e4b</span>',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'name': 'gemma4:e4b',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size': 10579079040,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'size_vram': 10579079040}]}&nbsp;&nbsp;</div></div><div><br /></div><div><div><br /></div><div>可見呼叫 /api/tags 端點會列出目前 Ollama 已下載的全部模型.&nbsp;</div></div><div><br /></div><div>但若呼叫距離前一次存取模型超過逾時時間 (預設 5 分鐘), Ollama 會自動將模型會從記憶體中清除, 這時呼叫 /api/ps 端點會得到 models 鍵為空串列的回應 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp;&nbsp;</div><div>{'models': []}</div></div><div><br /></div><div>這時只要用 POST 方法呼叫 /api/generate 或 /api/chat, Ollama 便會觸發 Ollama 自動將指定之模型載入記憶體 (毋須手動在命令列用 ollama run 指令載入).&nbsp;</div><div><br /></div><div>注意, 呼叫 get() 無法指定模型, 必須呼叫 post() 才能指定模型, 因為呼叫 post() 時可傳入一個 json 參數, 用字典來打包要傳給 Ollama 端點之酬載資訊 (payload), 其中的 model 鍵便是用來指定要載入之模型, 例如下面是呼叫文字接龍端點 /api/generate 的酬載資訊 :</div><div><br /></div><div><div>payload={</div><div>&nbsp; &nbsp; "model": "gemma4:e4b",</div><div>&nbsp; &nbsp; "prompt": "請簡介量子糾纏",</div><div>&nbsp; &nbsp; "stream": False</div><div>&nbsp; &nbsp; }</div></div><div><br /></div><div>例如 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/generate'</span></b>&nbsp; &nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">payload={</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "model": "gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "prompt": "你是誰",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "stream": False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.post(url, json=payload)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp;&nbsp;</div></div><div><div>{'context': [2,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;105,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;9731,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>... (略) ...</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;237169,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;240975,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;239967,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;236918,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;238463,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;237536],</div><div>&nbsp;'created_at': '2026-05-30T01:28:46.9601674Z',</div><div>&nbsp;'done': True,</div><div>&nbsp;'done_reason': 'stop',</div><div>&nbsp;'eval_count': 443,</div><div>&nbsp;'eval_duration': 5984802600,</div><div>&nbsp;'load_duration': 229697200,</div><div>&nbsp;'model': 'gemma4:e4b',</div><div>&nbsp;'prompt_eval_count': 18,</div><div>&nbsp;'prompt_eval_duration': 235963400,</div><div>&nbsp;'<span style="background-color: #fcff01;">response</span>': '我叫 **Gemma 4**。\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'我是一個大型語言模型（Large Language Model, LLM），由 Google DeepMind 開發。\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'我的設計目的是用來理解和生成人類語言，我可以協助您回答問題、撰寫不同類型的文本、總結資訊，或進行創意性的對話。\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'您有什麼問題需要我幫忙的呢？',</div><div>&nbsp;'total_duration': 6698511600}</div></div><div><br /></div><div>可見對於單一問答 (接龍) 呼叫, Response 物件中包含一長串的 context 訊息, 而模型實際的回應則是放在 resposne 鍵裡 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(result['response'])</span></b>&nbsp; &nbsp;</div><div>我叫 **Gemma 4**。</div><div><br /></div><div>我是一個大型語言模型（Large Language Model, LLM），由 Google DeepMind 開發。</div><div><br /></div><div>我的設計目的是用來理解和生成人類語言，我可以協助您回答問題、撰寫不同類型的文本、總結資訊，或進行創意性的對話。</div><div><br /></div><div>您有什麼問題需要我幫忙的呢？</div></div><div><br /></div><div>如果是呼叫對話端點 /api/chat, 則可在 messages 鍵中放入包含 role 與 content 鍵之對話字典, role 可以是 system/user/assistant, 格式與 OpenAI API 類似 :&nbsp;</div><div><br /></div><div><div>payload={</div><div>&nbsp; &nbsp; "model": "gemma4:e4b",</div><div>&nbsp; &nbsp; "messages": [</div><div>&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請用 100 個字簡介量子糾纏"}</div><div>&nbsp; &nbsp; &nbsp; &nbsp; ],</div><div>&nbsp; &nbsp; "stream": False</div><div>&nbsp; &nbsp; }</div></div><div><br /></div><div>例如 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/chat'</span></b>&nbsp; &nbsp;&nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">payload={</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "model": "gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "messages": [</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請用 100 個字簡介量子糾纏"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "stream": False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.post(url, json=payload)</span></b>&nbsp;&nbsp;</div></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp; &nbsp;</div><div>{'created_at': '2026-05-30T01:39:36.5316751Z',</div><div>&nbsp;'done': True,</div><div>&nbsp;'done_reason': 'stop',</div><div>&nbsp;'eval_count': 595,</div><div>&nbsp;'eval_duration': 8100674200,</div><div>&nbsp;'load_duration': 4374553000,</div><div>&nbsp;'<span style="background-color: #fcff01;">message</span>': {'<span style="background-color: #fcff01;">content</span>': '量子糾纏是一種奇特的量子現象，描述兩個或多個粒子之間存在的緊密關聯。無論這些粒子相距多遠，牠們的量子狀態都會形成一個不可分割的整體。\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '當我們測量其中一個粒子的某個性質（例如自旋）時，另一個粒子的狀態會瞬間、同步地確定，彷彿有無形的超距聯繫一般。愛因斯坦曾戲稱為「幽靈般的超距作用」。\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '這種現象是量子計算和量子密碼學的基礎技術，是未來尖端科技不可或缺的核心概念。它挑戰了我們對物理世界空間和資訊傳輸的傳統認知。',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'role': 'assistant',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'thinking': "Here's a thinking process to ensure the response "</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'meets the criteria (Taiwanese Mandarin, Traditional '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Chinese, approximately 100 characters, and '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'scientifically accurate description of quantum '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'entanglement):\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'1.&nbsp; **Analyze the Request:**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Topic: Quantum Entanglement (量子糾纏).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Length: Approximately 100 characters.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Style: Use Taiwanese terminology and '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Traditional Chinese (台灣用語與繁體中文).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'2.&nbsp; **Identify Key Concepts (Quantum '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Entanglement):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Definition: A unique connection between two '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'or more quantum particles.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Feature: Their fates are linked, regardless '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'of the distance separating them.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Measurement Effect: Measuring a property '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'(like spin) of one instantly tells you the property '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'of the other.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Paradox/Implication: Einstein called it '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'"spooky action at a distance" (幽靈般的超距作用).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Application: Crucial for quantum computing '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'and quantum cryptography.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'3.&nbsp; **Drafting the Content (Focusing on the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'"Taiwanese Tone" and Flow):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Goal:* Make it accessible yet precise.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Initial Draft Check (Mental Outline):* '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Start with the definition, explain the link, mention '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'the distance independence, and end with the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'application/impact.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'4.&nbsp; **Refining and Polishing (Character Count and '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Tone):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Self-Correction/Refinement:* Ensure the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'language feels natural for a Taiwanese audience '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'(using common Taiwanese scientific phrasing, but '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'remaining formal).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*(Structuring the 100 characters):*\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; &nbsp; *&nbsp; &nbsp;Introduction (What it is).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; &nbsp; *&nbsp; &nbsp;Mechanism (The spooky link).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; &nbsp; *&nbsp; &nbsp;Conclusion (Why it matters).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'5.&nbsp; **Final Polish (Review against constraints):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Taiwanese/Traditional Chinese: Yes.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Length: Close to 100 characters (The target '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'is an *approximation*).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Accuracy: Yes, captures the core idea.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'6.&nbsp; **Final Output Generation.** (This leads to the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'provided good answer.)'},</div><div>&nbsp;'model': 'gemma4:e4b',</div><div>&nbsp;'prompt_eval_count': 40,</div><div>&nbsp;'prompt_eval_duration': 35168300,</div><div>&nbsp;'total_duration': 12994831300}</div></div><div><br /></div><div>模型的實際回應是放在 message 鍵的 content 鍵裡 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(result['message']['content'])</span></b>&nbsp; &nbsp;</div><div>量子糾纏是一種奇特的量子現象，描述兩個或多個粒子之間存在的緊密關聯。無論這些粒子相距多遠，牠們的量子狀態都會形成一個不可分割的整體。</div><div><br /></div><div>當我們測量其中一個粒子的某個性質（例如自旋）時，另一個粒子的狀態會瞬間、同步地確定，彷彿有無形的超距聯繫一般。愛因斯坦曾戲稱為「幽靈般的超距作用」。</div><div><br /></div><div>這種現象是量子計算和量子密碼學的基礎技術，是未來尖端科技不可或缺的核心概念。它挑戰了我們對物理世界空間和資訊傳輸的傳統認知。</div></div><div><br /></div><div>或者連續呼叫字典的 get() 方法亦可 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">print(result.get('message').get('content'))</span></b>&nbsp; &nbsp;</div><div>量子糾纏是一種奇特的量子現象，描述兩個或多個粒子之間存在的緊密關聯。無論這些粒子相距多遠，牠們的量子狀態都會形成一個不可分割的整體。</div><div><br /></div><div>當我們測量其中一個粒子的某個性質（例如自旋）時，另一個粒子的狀態會瞬間、同步地確定，彷彿有無形的超距聯繫一般。愛因斯坦曾戲稱為「幽靈般的超距作用」。</div><div><br /></div><div>這種現象是量子計算和量子密碼學的基礎技術，是未來尖端科技不可或缺的核心概念。它挑戰了我們對物理世界空間和資訊傳輸的傳統認知。</div></div><div><br /></div><div><br /></div><div><div><b><span style="background-color: #fce5cd; color: #990000;">3. 核心參數 (外層) 與進階參數 (內層) :</span></b>&nbsp;&nbsp;</div></div><div><br /></div><div>Ollama REST API 對話端點 /api/chat 與文字接龍端點 /api/generate 的參數結構分為外層的核心參數與內層的進階參數 (options) 兩種, Ollama 原生端點規定最外層只能放控制 API 行為的核心參數, 其餘微調參數必須全部打包塞進一個叫 "options" 的子字典裡; OpenAI 則是把所有控制 AI 腦袋的參數 (如隨機性與記憶力等) 全部平鋪在最外層.&nbsp;</div><div><br /></div><div>核心參數說明如下表 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 109.011px;">&nbsp;核心參數名稱</th> <th style="padding: 3px; width: 552.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;model</td> <td style="padding: 3px;">&nbsp;必要。指定要執行的模型名稱（字串），例如 "gemma4:e4b"。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;messages</td> <td style="padding: 3px;">&nbsp;必要。對話紀錄串列，內含多個指定 role (system/user/assistant) 與 content 鍵之字典。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;stream</td> <td style="padding: 3px;">&nbsp;選用。是否以串流回應 (預設 True)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;format</td> <td style="padding: 3px;">&nbsp;選用。指定回傳格式 (目前僅支援 "json" : 強迫模型輸出結構化的 JSON 字串, 預設 null)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;keep_alive</td> <td style="padding: 3px;">&nbsp;選用。設定模型在留在記憶體中的時間 (預設為 "5m"，設為 0 表示立刻釋放)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;options</td> <td style="padding: 3px;">&nbsp;選用。進階模型微調參數字典 (例如 Temperature 或 num_ctx 等)。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>內層的 options 的微調參數 (鍵) 說明如下表 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 109.011px;">&nbsp;內層參數名稱</th> <th style="padding: 3px; width: 552.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;temperature</td> <td style="padding: 3px;">&nbsp;隨機性 : 值越高回答越有創意但易離題; 值越低回答越精準與固定 (預設 0.8)</td> </tr> <tr> <td style="padding: 3px;">&nbsp;num_ctx</td> <td style="padding: 3px;">&nbsp;含輸入與輸出之上下文視窗大小 (預設為 2048 個 token)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;num_predict</td> <td style="padding: 3px;">&nbsp;單次回答的最大 Token 數 (預設 -1 表示無限制)</td> </tr> <tr> <td style="padding: 3px;">&nbsp;top_p</td> <td style="padding: 3px;">&nbsp;核心採樣機率, 限制模型只能從累積機率前 P% 的高機率詞彙中挑選字 (預設 0.9)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;top_k</td> <td style="padding: 3px;">&nbsp;最高機率詞彙數。限制模型選字時只考慮機率最高的前 K 個詞 (預設 40)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;repeat_penalty</td> <td style="padding: 3px;">&nbsp;重複字詞懲罰係數。避開已經講過的詞來防止重覆同一句話 (預設 1.1)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;seed</td> <td style="padding: 3px;">&nbsp;隨機數種子 (整數)。強迫模型吐出一模一樣的答案 (預設 -1, 完全隨機)。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>format 參數預設為 null, 模型會以自由文本 (Free Text) 模式回應. 如果 Python 程式需要串接資料庫, 或是進行網頁 RAG 數據結構化時, 可以利用&nbsp;"format": "json" 參數設定, 要求模型以符合 JSON 格式的字串回應 (可直接轉成 Python 字典). 注意, 啟用 JSON 輸出必須同時在 prompt 或 messages 提示詞中同時明確告訴模型要輸出 JSON, 否則可能會讓此設定破功, 例如 :&nbsp;</div><div><br /></div><div><div>payload={</div><div>&nbsp; &nbsp; "model": "gemma4:e4b",</div><div>&nbsp; &nbsp; "messages": [</div><div>&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</div><div>&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請列出全球人均 GDP 前十大國家, 包含 Country 與 GDP 兩欄位, 且<span style="background-color: #fcff01;">以 JSON 格式回覆</span>"}</div><div>&nbsp; &nbsp; &nbsp; &nbsp; ],</div><div>&nbsp; &nbsp; <span style="background-color: #fcff01;">"format": "json"</span>,</div><div>&nbsp; &nbsp; "stream": False</div><div>&nbsp; &nbsp; }</div></div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/chat'</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">payload={</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "model": "gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "messages": [</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "system", "content": "請一律用台灣用語與繁體中文回答"},</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; {"role": "user", "content": "請列出全球人均 GDP 前十大國家, 包含 Country 與 GDP 兩欄位, 且以 JSON 格式回覆"}</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; &nbsp; &nbsp; ],</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "format": "json",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "stream": False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.post(url, json=payload)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;</div></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">type(result)</span></b>&nbsp; &nbsp;</div><div>&lt;class '<span style="background-color: #fcff01;">dict'</span>&gt;&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)&nbsp;</span></b> &nbsp;</div><div>{'created_at': '2026-05-30T07:48:28.8572311Z',</div><div>&nbsp;'done': True,</div><div>&nbsp;'done_reason': 'stop',</div><div>&nbsp;'eval_count': 512,</div><div>&nbsp;'eval_duration': 6949736800,</div><div>&nbsp;'load_duration': 4340405700,</div><div>&nbsp;'<span style="background-color: #fcff01;">message</span>': {'<span style="background-color: #fcff01;">content</span>': '{\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; "note": "由於「人均 '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 'GDP」是高度依賴數據來源、計算時間點（例如是預估值還是實際值）以及使用哪個國際組織（如 IMF, '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '世界銀行）數據而有巨大變動的數據。此外，名單的具體數字和排名也可能因年度資料更新而改變。此處提供的數據為一個範例的參考格式，您應使用最新的權威資料庫（如世界銀行或 '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 'IMF）來獲取最精準且最新的全球排名。由於我無法即時存取即時的國際金融數據，且無法確定您需要的「哪個年份」的數據，我會提供一個結構化的範例，並用一個說明性的年份數據作為範例。",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; "<span style="background-color: #fcff01;">top_10_countries</span>": [\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "路易斯利亞 (Luxembourg)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 120000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "新加坡 (Singapore)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 105000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "挪威 (Norway)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 95000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "瑞士 (Switzerland)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 90000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "阿聯酋 (United Arab Emirates)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 85000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "葡萄牙 (Portugal)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 78000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "冰島 (Iceland)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 75000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "愛爾蘭 (Ireland)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 72000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "韓國 (South Korea)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 65000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "Country": "美國 (United States)",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; &nbsp; "GDP (PPP, USD)": 60000\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; &nbsp; }\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '&nbsp; ]\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; '}',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'role': 'assistant',</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'thinking': "Here's a thinking process to construct the desired "</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'JSON response:\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'1.&nbsp; **Understand the Goal:** The user wants a list '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'of the top 10 countries by *per capita GDP* (人均 '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'GDP).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'2.&nbsp; **Understand the Format:** The output *must* be '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'in JSON format, with two specific fields: `Country` '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'and `GDP`.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'3.&nbsp; **Identify Constraints/Caveats (Crucial '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Step):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;The request requires real-time, global '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'economic data (top 10 per capita GDP).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;My knowledge cutoff is January 2025, and I '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'do not have access to a live, definitive, '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'up-to-the-minute economic database (like World Bank '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'or IMF).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;Therefore, the data I provide must be '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'*representative* of recent trends but might not be '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'the absolute current ranking. I must use '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'placeholder/estimated values and acknowledge the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'nature of the data.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'4.&nbsp; **Determine the Data Structure (JSON):** The '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'best structure will be an array of objects.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; ```json\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; [\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; {"Country": "...", "GDP": "..." },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; {"Country": "...", "GDP": "..." },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; // ... up to 10\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; ]\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; ```\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'5.&nbsp; **Gather Representative Data (Top 10 Per Capita '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'GDP):** *Self-Correction/Knowledge Retrieval:* High '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'per capita GDP countries usually include Luxembourg, '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Singapore, Qatar, Ireland, Norway, etc. I will use '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'estimated/representative figures for the JSON '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'structure. (Note: I must ensure the GDP figures are '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'plausible approximations, usually given in USD.)\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'6.&nbsp; **Final JSON Construction (Iterative '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Drafting):** (Populate the structure with the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'data.)\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'7.&nbsp; **Review and Refine (Adherence to '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'Tone/Format):**\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Language:* Use Taiwanese Mandarin (台灣用語, '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'繁體中文).\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Tone:* Formal, professional.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Format:* Strict JSON.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; *&nbsp; &nbsp;*Disclaimer:* It is essential to include a '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'disclaimer that economic figures change constantly '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'and the data provided is based on the knowledge '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'cutoff date.\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'8.&nbsp; **Final Output Generation.** (This leads to the '</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'provided JSON and explanatory text.)'},</div><div>&nbsp;'model': 'gemma4:e4b',</div><div>&nbsp;'prompt_eval_count': 61,</div><div>&nbsp;'prompt_eval_duration': 27626100,</div><div>&nbsp;'total_duration': 18852258300}</div></div><div><br /></div><div>可見模型成功傳回含有 JSON 格式的資料, 但它擅自將 GDP 欄位名稱改為&nbsp;GDP (PPP, USD), 如果要避免此情形, 提示詞需要加以限制, 例如改為 :</div><div><br /></div><div><i><span style="color: #2b00fe;">請列出全球人均 GDP 前十大國家, 包含 Country 與 GDP 兩欄位, 且以 JSON 格式回覆. 注意, 欄位名稱請忠實使用 Country 與 GDP, 不可自行更改</span></i></div><div><br /></div><div>上面的回應結果顯示, GDP 資料放在 result["messages"]["content"], 其值是包含 JSON 回應的字串, 先取出此回應字串 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">json_string=result['message']['content']</span></b>&nbsp; &nbsp;</div><div><br /></div><div>然後匯入內建的 json 模組, 呼叫其 loads() 函式將此 JSON 字串轉成 Python 字典 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import json</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">clean_data=json.loads(json_string)</span></b>&nbsp; &nbsp; &nbsp;</div></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">clean_data</span></b>&nbsp;&nbsp;</div><div>{'note': '由於「人均 GDP」是高度依賴數據來源、計算時間點（例如是預估值還是實際值）以及使用哪個國際組織（如 IMF, 世界銀行）數據而有巨大變動的數據。此外，名單的具體數字和排名也可能因年度資料更新而改變。此處提供的數據為一個範例的參考格式，您應使用最新的權威資料庫（如世界銀行或 IMF）來獲取最精準且最新的全球排名。由於我無法即時存取即時的國際金融數據，且無法確定您需要的「哪個年份」的數據，我會提供一個結構化的範例，並用一個說明性的年份數據作為範例。', '<span style="background-color: #fcff01;">top_10_countries</span>': [{'Country': '路易斯利亞 (Luxembourg)', 'GDP (PPP, USD)': 120000}, {'Country': '新加坡 (Singapore)', 'GDP (PPP, USD)': 105000}, {'Country': '挪威 (Norway)', 'GDP (PPP, USD)': 95000}, {'Country': '瑞士 (Switzerland)', 'GDP (PPP, USD)': 90000}, {'Country': '阿聯酋 (United Arab Emirates)', 'GDP (PPP, USD)': 85000}, {'Country': '葡萄牙 (Portugal)', 'GDP (PPP, USD)': 78000}, {'Country': '冰島 (Iceland)', 'GDP (PPP, USD)': 75000}, {'Country': '愛爾蘭 (Ireland)', 'GDP (PPP, USD)': 72000}, {'Country': '韓國 (South Korea)', 'GDP (PPP, USD)': 65000}, {'Country': '美國 (United States)', 'GDP (PPP, USD)': 60000}]}</div></div><div><br /></div><div>人均 GDP 前十大國家資料放在&nbsp;top_10_countries 鍵裡 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">countries_list=clean_data['top_10_countries']</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">countries_list</span></b>&nbsp; &nbsp;</div><div>[{'Country': '路易斯利亞 (Luxembourg)', 'GDP (PPP, USD)': 120000}, {'Country': '新加坡 (Singapore)', 'GDP (PPP, USD)': 105000}, {'Country': '挪威 (Norway)', 'GDP (PPP, USD)': 95000}, {'Country': '瑞士 (Switzerland)', 'GDP (PPP, USD)': 90000}, {'Country': '阿聯酋 (United Arab Emirates)', 'GDP (PPP, USD)': 85000}, {'Country': '葡萄牙 (Portugal)', 'GDP (PPP, USD)': 78000}, {'Country': '冰島 (Iceland)', 'GDP (PPP, USD)': 75000}, {'Country': '愛爾蘭 (Ireland)', 'GDP (PPP, USD)': 72000}, {'Country': '韓國 (South Korea)', 'GDP (PPP, USD)': 65000}, {'Country': '美國 (United States)', 'GDP (PPP, USD)': 60000}]</div></div><div><br /></div><div>可以用 Pandas 將其轉成容易處理的 DataFrame :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">import pandas as pd</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">df=pd.DataFrame(countries_list)</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">df</span></b>&nbsp; &nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; Country&nbsp; GDP (PPP, USD)</div><div>0&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 路易斯利亞 (Luxembourg)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 120000</div><div>1&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;新加坡 (Singapore)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 105000</div><div>2&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;挪威 (Norway)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;95000</div><div>3&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 瑞士 (Switzerland)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;90000</div><div>4&nbsp; 阿聯酋 (United Arab Emirates)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;85000</div><div>5&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 葡萄牙 (Portugal)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;78000</div><div>6&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 冰島 (Iceland)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;75000</div><div>7&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;愛爾蘭 (Ireland)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;72000</div><div>8&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 韓國 (South Korea)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;65000</div><div>9&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 美國 (United States)&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;60000</div></div><div><br /></div><div>注意, 由於未聯網搜尋, 資料的正確性可能受限.&nbsp;</div><div><br /></div><div>對於文字接龍的單次生成請求呼叫 /api/generate 端點時也可以用 json 參數限制輸出 JSON 格式資料, 下面我們改用 /api/generate 端點, 並加強提示詞關於欄位的限制重新測試 :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">url='http://localhost:11434/api/generate'&nbsp;</span></b>&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">payload={&nbsp;</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "model": "gemma4:e4b",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "prompt": "請列出全球人均 GDP 前十大國家, 包含 Country 與 GDP 兩欄位, 且以 JSON 格式回覆. 注意, <span style="background-color: #fcff01;">欄位名稱請忠實使用 Country 與 GDP, 不可自行更改</span>",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "format": "json",</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; "stream": False</span></b></div><div><b><span style="color: #2b00fe;">&nbsp; &nbsp; }</span></b></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">reply=requests.post(url, json=payload)</span></b>&nbsp;&nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">result=reply.json()</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">pprint(result)</span></b>&nbsp; &nbsp;</div><div>{'context': [2,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;105,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;9731,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;98,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;106,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;105,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;2364,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div></div></div><div>... (略) ...</div><div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;236783,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;138,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;236842,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;107,</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;236783],</div><div>&nbsp;'created_at': '2026-05-30T08:30:01.0055623Z',</div><div>&nbsp;'done': True,</div><div>&nbsp;'done_reason': 'stop',</div><div>&nbsp;'eval_count': 329,</div><div>&nbsp;'eval_duration': 4434360500,</div><div>&nbsp;'load_duration': 4382299000,</div><div>&nbsp;'model': 'gemma4:e4b',</div><div>&nbsp;'prompt_eval_count': 64,</div><div>&nbsp;'prompt_eval_duration': 40775600,</div><div>&nbsp;'<span style="background-color: #fcff01;">response</span>': '{\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; "global_per_capita_gdp_top_10": [\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Singapore",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 80,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Luxembourg",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 140,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Qatar",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 85,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Ireland",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 90,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Norway",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 80,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Switzerland",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 95,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "United States",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 75,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Australia",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 65,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Iceland",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 60,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; },\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; {\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "Country": "Denmark",\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; &nbsp; "GDP": "約 55,000 USD"\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; &nbsp; }\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'&nbsp; ]\n'</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;'}',</div><div>&nbsp;'total_duration': 9343526600}</div></div><div><br /></div><div>可見這回有遵照約束, 使用指定的 Country 與 GDP 做欄位名稱了. 其次, 與 /api/chat 不同的是, /api/generate 將回應放在字典的 response 鍵裡, 取出後用 json.loads() 將其轉成 Python 字典 :</div><div><br /></div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">json_string=result['response']</span></b>&nbsp;&nbsp;</div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">clean_data=json.loads(json_string)</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">clean_data</span></b>&nbsp; &nbsp;</div><div>{'<span style="background-color: #fcff01;">global_per_capita_gdp_top_10</span>': [{'Country': 'Singapore', 'GDP': '約 80,000 USD'}, {'Country': 'Luxembourg', 'GDP': '約 140,000 USD'}, {'Country': 'Qatar', 'GDP': '約 85,000 USD'}, {'Country': 'Ireland', 'GDP': '約 90,000 USD'}, {'Country': 'Norway', 'GDP': '約 80,000 USD'}, {'Country': 'Switzerland', 'GDP': '約 95,000 USD'}, {'Country': 'United States', 'GDP': '約 75,000 USD'}, {'Country': 'Australia', 'GDP': '約 65,000 USD'}, {'Country': 'Iceland', 'GDP': '約 60,000 USD'}, {'Country': 'Denmark', 'GDP': '約 55,000 USD'}]}</div></div><div><br /></div><div>人均 GDP 前十大國家資料放在 global_per_capita_gdp_top_10 鍵裡 :&nbsp;</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">countries_list=clean_data['global_per_capita_gdp_top_10']</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">countries_list</span></b>&nbsp;&nbsp;</div><div>[{'Country': 'Singapore', 'GDP': '約 80,000 USD'}, {'Country': 'Luxembourg', 'GDP': '約 140,000 USD'}, {'Country': 'Qatar', 'GDP': '約 85,000 USD'}, {'Country': 'Ireland', 'GDP': '約 90,000 USD'}, {'Country': 'Norway', 'GDP': '約 80,000 USD'}, {'Country': 'Switzerland', 'GDP': '約 95,000 USD'}, {'Country': 'United States', 'GDP': '約 75,000 USD'}, {'Country': 'Australia', 'GDP': '約 65,000 USD'}, {'Country': 'Iceland', 'GDP': '約 60,000 USD'}, {'Country': 'Denmark', 'GDP': '約 55,000 USD'}]</div></div><div><br /></div><div>轉成 DataFrame :</div><div><br /></div><div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">df=pd.DataFrame(countries_list)</span></b>&nbsp;&nbsp;</div><div>&gt;&gt;&gt; <b><span style="color: #2b00fe;">df</span></b>&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp;Country&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; GDP</div><div>0&nbsp; &nbsp; &nbsp; Singapore&nbsp; &nbsp;約 80,000 USD</div><div>1&nbsp; &nbsp; &nbsp;Luxembourg&nbsp; 約 140,000 USD</div><div>2&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; Qatar&nbsp; &nbsp;約 85,000 USD</div><div>3&nbsp; &nbsp; &nbsp; &nbsp; Ireland&nbsp; &nbsp;約 90,000 USD</div><div>4&nbsp; &nbsp; &nbsp; &nbsp; &nbsp;Norway&nbsp; &nbsp;約 80,000 USD</div><div>5&nbsp; &nbsp; Switzerland&nbsp; &nbsp;約 95,000 USD</div><div>6&nbsp; United States&nbsp; &nbsp;約 75,000 USD</div><div>7&nbsp; &nbsp; &nbsp; Australia&nbsp; &nbsp;約 65,000 USD</div><div>8&nbsp; &nbsp; &nbsp; &nbsp; Iceland&nbsp; &nbsp;約 60,000 USD</div><div>9&nbsp; &nbsp; &nbsp; &nbsp; Denmark&nbsp; &nbsp;約 55,000 USD</div></div><div><br /></div><div>怎麼排名跟上面呼叫 /api/chat 的不同啊.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-76874349843647630262026-05-29T00:15:27.344+08:002026-05-29T18:36:47.359+08:00Ollama 學習筆記 : REST API 用法 (一)<div>本篇旨在測試 Ollama REST API 的第一種用法 : 使用 curl.exe 提出請求.&nbsp;</div><div><br /></div><div><div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div></div><div><br /></div><div>當執行 Ollama 應用程式後, 它會自動啟動一個 REST API 伺服器, 網址為 :</div><div><div><br /></div><div># <a href="http://localhost:11434" target="_blank">http://localhost:11434</a>&nbsp;</div><div><br /></div><div>把這網址貼到瀏覽器, 若顯示 "Ollama is running" 表示 Ollama 伺服器已啟動 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLVGiMxB_D0EOoKAYWPWKBufv7Z6VzKCeK7h3ViZGedKx1udUUIyERCSiU1E9hADg8UbLN5JMnrU88H082Emu6tPZMd2Fz9YjwCqpPur41_25jnLKS6VpwR-sE5mp3E7rt-Mu2U19ru2_Hzjd16RBxg-3e4slk_an9ciDvv9Ug6KXEBAhRQoclBQEUR3-A/s571/ollama-is-running.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="199" data-original-width="571" height="112" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLVGiMxB_D0EOoKAYWPWKBufv7Z6VzKCeK7h3ViZGedKx1udUUIyERCSiU1E9hADg8UbLN5JMnrU88H082Emu6tPZMd2Fz9YjwCqpPur41_25jnLKS6VpwR-sE5mp3E7rt-Mu2U19ru2_Hzjd16RBxg-3e4slk_an9ciDvv9Ug6KXEBAhRQoclBQEUR3-A/s320/ollama-is-running.jpg" width="320" /></a></div><br /><div>以下測試必須在 Ollama 伺服器執行中才會成功.&nbsp;</div><div><br /></div><div>此 REST API 伺服器的大部分端點都要用 POST 方法提出請求 (主要是叫模型做事或改變系統狀態), 因為這些操作需要傳遞複雜的參數 (以 JSON 格式打包在 Body 中); 少部分則可以用 GET 方法 (只查詢看資料不改變任何東西的唯讀操作, 例如查詢狀態或列舉清單).&nbsp;&nbsp;</div><div><br /></div><div>以下是常用的 POST 端點 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;POST 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/generate</td> <td style="padding: 3px;">&nbsp;文字生成 (用於單次提示詞的輸入與基本文字接龍)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/chat</td> <td style="padding: 3px;">&nbsp;多輪對話 (用於需要記錄上下文, 用 user/assistant/system 區分角色)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/embeddings</td> <td style="padding: 3px;">&nbsp;使用 Embedding 模型將文字轉成可用於 RAG 語意搜尋的嵌入向量。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/pull</td> <td style="padding: 3px;">&nbsp;從 Ollama 官方模型庫下載指定的 LLM 模型到本機。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/push</td> <td style="padding: 3px;">&nbsp;將模型推入 Ollama 官方模型庫 (需登入 &amp; 輸入金鑰)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/show</td> <td style="padding: 3px;">&nbsp;顯示模型的詳細資訊。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/create</td> <td style="padding: 3px;">&nbsp;透過傳入 Modelfile 的內容從本機檔案建立或自訂一個全新的模型。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/copy</td> <td style="padding: 3px;">&nbsp;在本地將一個現有的模型複製並重新命名為另一個名稱。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>常用的 GET 端點 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;GET 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/tags</td> <td style="padding: 3px;">&nbsp;列出本地所有模型。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/ps</td> <td style="padding: 3px;">&nbsp;查看目前有哪些模型正載入在記憶體 (DRAM/VRAM)。</td> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/version</td> <td style="padding: 3px;">&nbsp;取得 Ollama 的版本資訊。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>刪除模型則會用到 DELETE 方法 :</div><div><br /></div><div><br /> <table border="1" bordercolor="#888" cellspacing="0" style="background-color: white; border-collapse: collapse; border-color: rgb(136, 136, 136); border-width: 1px; color: #534741; font-family: Arial, Verdana, sans-serif; font-size: 13.3333px; margin: 0px;"> <tbody> <tr> <th style="padding: 3px; width: 119.011px;">&nbsp;DELETE 操作端點</th> <th style="padding: 3px; width: 472.011px;">&nbsp;說明</th> </tr> <tr> <td style="padding: 3px;">&nbsp;/api/delete</td> <td style="padding: 3px;">&nbsp;刪除本地指定之模型 (使用 name 鍵指定)。</td> </tr> </tbody> </table> </div><div><br /></div><div><br /></div><div>向 REST API 伺服器提出請求可以在命令列用 curl 指令或在 Python 程式中使用 requests 套件的 post(), get(), 或 delete() 函式.&nbsp;</div></div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 使用 cURL 提出請求 :</span></b></div></div><div><br /></div><div>在 CMD 或 PS 視窗用 curl.exe 指令呼叫 REST API 端點, 例如 :</div><div><br /></div><div>curl.exe http://localhost:11434/api/tags</div><div><br /></div><div>此指令會列出本地所有模型 :&nbsp;</div><div><br /></div><div><div>PS C:\Users\USER&gt;<b><span style="color: #2b00fe;"> curl.exe http://localhost:11434/api/tags</span></b>&nbsp; &nbsp;&nbsp;</div><div>{"models":[{"name":"llama3.2-vision:11b","model":"llama3.2-vision:11b","modified_at":"2026-05-25T11:30:36.0322613+08:00","size":7816589186,"digest":"6f2f9757ae97e8a3f8ea33d6adb2b11d93d9a35bef277cd2c0b1b5af8e8d0b1e","details":{"parent_model":"","format":"gguf","family":"mllama","families":["mllama"],"parameter_size":"10.7B","quantization_level":"Q4_K_M"}},{"name":"phi4:14b","model":"phi4:14b","modified_at":"2026-05-24T17:23:24.1907029+08:00","size":9053116391,"digest":"ac896e5b8b34a1f4efa7b14d7520725140d5512484457fab45d2a4ea14c69dba","details":{"parent_model":"","format":"gguf","family":"phi3","families":["phi3"],"parameter_size":"14.7B","quantization_level":"Q4_K_M"}},{"name":"dagbs/deepseek-coder-v2-lite-instruct:q4_k_m","model":"dagbs/deepseek-coder-v2-lite-instruct:q4_k_m","modified_at":"2026-05-24T00:17:21.5900783+08:00","size":10364417401,"digest":"a6f5c73087ad25fc8666929492449eb0dc694326e4ca5b2313fef75b66645583","details":{"parent_model":"","format":"gguf","family":"deepseek2","families":["deepseek2"],"parameter_size":"15.7B","quantization_level":"Q4_K_M"}},{"name":"mannix/deepseek-coder-v2-lite-instruct:q4_k_m","model":"mannix/deepseek-coder-v2-lite-instruct:q4_k_m","modified_at":"2026-05-23T21:15:32.5340567+08:00","size":10364432240,"digest":"6171206208d0529a47806ebcf8ed37a88fe322859e269396dd16fdd98a56a102","details":{"parent_model":"","format":"gguf","family":"deepseek2","families":["deepseek2"],"parameter_size":"15.7B","quantization_level":"Q4_K_M"}},{"name":"deepseek-r1:14b","model":"deepseek-r1:14b","modified_at":"2026-05-23T16:43:35.5527992+08:00","size":8988112209,"digest":"c333b7232bdb521236694ffbb5f5a6b11cc45d98e9142c73123b670fca400b09","details":{"parent_model":"","format":"gguf","family":"qwen2","families":["qwen2"],"parameter_size":"14.8B","quantization_level":"Q4_K_M"}},{"name":"qwen3:14b","model":"qwen3:14b","modified_at":"2026-05-22T00:30:36.3395998+08:00","size":9276198565,"digest":"bdbd181c33f2ed1b31c972991882db3cf4d192569092138a7d29e973cd9debe8","details":{"parent_model":"","format":"gguf","family":"qwen3","families":["qwen3"],"parameter_size":"14.8B","quantization_level":"Q4_K_M"}},{"name":"gemma4:latest","model":"gemma4:latest","modified_at":"2026-05-20T11:45:08.2471048+08:00","size":9608350718,"digest":"c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb","details":{"parent_model":"","format":"gguf","family":"gemma4","families":["gemma4"],"parameter_size":"8.0B","quantization_level":"Q4_K_M"}}]}</div></div><div><br /></div><div>可見全部模型會放在鍵為 models 的串列中, 以 JSON 格式傳回. 注意, 一定要用 curl.exe 不要用 curl,&nbsp;前者呼叫的是 Linux 生態系那個真正的 cURL 工具; 而單純的 curl 只是 PowerShell 幫內建指令 Invoke-WebRequest 取的一個別名, 其傳回值是一個包含模型資訊的複雜 JSON 物件, 長度太長會被自動截掉, 無法顯示全部模型清單.&nbsp;</div><div><br /></div><div><div>接下來呼叫 /api/ps 查詢目前記憶體中是否有載入任何模型 :&nbsp;</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/ps</span></b>&nbsp;&nbsp;</div><div>{"models":<span style="background-color: #fcff01;">[]</span>}</div></div><div><br /></div><div>models 鍵之值為空串列, 表示記憶體中未載入模型.&nbsp;</div><div><br /></div></div><div>開啟一個新的 PS 視窗, 用 ollama run 載入執行一個模型, 例如 Gemma 4 :&nbsp;</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">ollama run gemma4:e4b</span></b>&nbsp; &nbsp;</div><div>&gt;&gt;&gt;&nbsp;</div></div><div><br /></div><div>然後回原 PS 視窗再次呼叫 /api/ps, 這時就會顯示已載入&nbsp;gemma4:e4b 模型 :&nbsp;</div><div><br /></div><div><div>PS C:\Users\USER&gt;<b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/ps</span></b>&nbsp; &nbsp;</div><div>{"models":[{"name":"<span style="background-color: #fcff01;">gemma4:e4b</span>","model":"gemma4:e4b","size":10579079040,"digest":"c6eb396dbd5992bbe3f5cdb947e8bbc0ee413d7c17e2beaae69f5d569cf982eb","details":{"parent_model":"","format":"gguf","family":"gemma4","families":["gemma4"],"parameter_size":"8.0B","quantization_level":"Q4_K_M"},"expires_at":"2026-05-28T22:41:52.1342692+08:00","size_vram":10579079040,"context_length":4096}]}</div></div><div><br /></div><div>呼叫 /api/version 會傳回 Ollama 的版本 :</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/version</span></b>&nbsp; &nbsp;</div><div>{"version":"0.24.0"}</div></div><div><br /></div><div>如果要呼叫 POST 方法的 API 端點, 必須在 -d 參數後面帶入要放入 Body 中的資料 (-d 是 -data 的縮寫), 例如呼叫 /api/show 會顯示指定模型之詳細資訊. 注意, 在 Windows PowerShell 中使用 curl.exe 傳送 JSON 資料時, 外層須用單引號 ', 內層的雙引號必須加上反斜線 \" 進行轉義, 例如 :</div><div><br /></div><div>curl.exe http://localhost:11434/api/show -d '{\"name\": \"llama3.2-vision:11b\"}'&nbsp;&nbsp;</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/show -d '{\"name\": \"gemma4:e4b\"}'</span></b>&nbsp; &nbsp;</div><div>{"license":"&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; Apache License\n&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;Version 2.0, January 2004\n&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; http://www.apache.org/licenses/\n\n&nbsp; &nbsp;TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION\n\n&nbsp; &nbsp;1. Definitions.\n\n&nbsp; &nbsp; &nbsp; \"License\" shall mean the terms and conditions for use, reproduction,\n&nbsp; &nbsp; &nbsp; and distribution as defined by Sections 1 through 9 of this document.\n\n&nbsp; &nbsp; &nbsp; \"Licensor\" shall mean the copyright owner or entity authorized by\n&nbsp; &nbsp; &nbsp; the copyright owner that is granting the License.\n\n&nbsp; &nbsp; &nbsp; \"Legal Entity\" shall mean the union of the acting entity and all\n&nbsp; &nbsp; &nbsp; other entities that control, are controlled by, or are under common\n&nbsp; &nbsp; &nbsp; control with that entity. For the purposes of this definition,\n&nbsp; &nbsp; &nbsp; \"control\" means (i) the power, direct or indirect, to cause the\n ... 非常長 ...</div></div><div>... (略) ...</div><div>3072]},{"name":"v.blk.9.ln1.weight","type":"F32","shape":[768]},{"name":"v.blk.9.ln2.weight","type":"F32","shape":[768]},{"name":"v.patch_embd.weight","type":"F16","shape":[16,16,3,768]},{"name":"v.position_embd.weight","type":"F32","shape":[768,10240,2]}],"capabilities":["completion","vision","audio","tools","thinking"],"modified_at":"2026-05-28T22:36:46.3502486+08:00","requires":"0.20.0"}</div><div><br /></div><div>比較常用的 API 端點是 /api/generate (文字生成) 與 /api/chat (對話), 例如呼叫 /api/generate 指令如下 (注意參數內容的雙引號要用 \ 轉義) :&nbsp;</div><div><br /></div><div><div>curl.exe http://localhost:11434/api/generate -d '{</div><div>&nbsp; \"model\": \"gemma4:e4b\",</div><div>&nbsp; \"prompt\": \"你是誰?\",</div><div>&nbsp; \"stream\": false</div><div>}'</div></div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/generate -d '{</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"model\": \"gemma4:e4b\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"prompt\": \"你是誰?\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"stream\": false</span></b></div><div>&gt;&gt; <b><span style="color: #2b00fe;">}'</span></b>&nbsp;&nbsp;</div><div>{"model":"gemma4:e4b","created_at":"2026-05-28T15:05:57.5062972Z","response":"我叫 **Gemma 4**。\n\n我是一個由 Google DeepMind 開發的大型語言模型 (Large Language Model)。\n\n我的作用是處理和生成人類語言，我可以幫助您回答問題、寫作文本、翻譯語言，或是進行創意寫作。\n\n請問您需要我幫您做些什麼呢？","done":true,"done_reason":"stop","<span style="background-color: #fcff01;">context</span>":[2,105,9731,107,98,107,106,107,105,2364,107,95841,240560,236881,106,107,105,4368,107,100,45518,107,120474,12364,236787,108,236770,236761,138,1018,115863,506,16499,53121,669,2430,4733,623,95841,240560,7462,568,236797,240622,127880,704,155268,78546,837,55544,531,623,15938,659,611,7462,107,236778,236761,138,1018,102752,17354,46320,53121,564,1202,531,2847,496,63510,532,11459,3890,2721,580,1041,19080,11808,236761,107,140,236829,139,1567,236787,147224,236743,236812,236761,107,140,236829,139,96089,236787,6475,22267,65153,236761,107,140,236829,139,46797,236787,25093,22160,9483,568,2182,236792,769,107,140,236829,139,2328,236787,7607,18710,2028,236761,107,236800,236761,138,1018,102752,22160,53121,669,2744,5192,691,43899,8555,568,504,164557,236764,840,506,7609,3904,563,4077,8555,779,834,506,3072,1921,577,528,8555,236761,107,236812,236761,138,1018,88293,506,14503,568,495,8555,1473,1018,107,140,236829,139,6302,684,29354,506,1463,236787,26911,237026,147224,236743,236812,568,236777,1006,147224,236743,236812,769,107,140,236829,139,3112,506,9059,236786,39822,236787,26911,90432,56762,146569,26609,568,236777,1006,496,2455,5192,2028,769,107,140,236829,139,3112,506,19788,236787,108965,6475,22267,65153,65706,238623,568,165684,684,6475,22267,65153,769,107,140,236829,139,236769,43983,840,11045,1473,99082,9779,1292,568,236744,236761,236759,1126,9795,3890,4137,236764,5712,1816,769,107,236810,236761,138,1018,30852,532,10867,688,568,25864,236772,135778,236786,87228,1473,1018,41152,506,3072,563,54651,236764,11459,236764,532,5467,19246,506,11172,1651,98469,531,506,5221,19080,236761,108,236825,236761,138,1018,17667,16887,32955,99382,568,2094,9025,531,506,3847,8555,3072,2907,101,237169,239138,5213,236823,12367,236743,236812,1018,236924,108,237169,90432,237852,6475,22267,65153,65706,238623,29854,237731,146569,26609,568,31534,22160,9483,45511,108,21480,24654,237026,93521,237206,25352,126592,146569,236900,183868,116904,238602,49695,18053,236951,240564,237284,57489,236951,205963,146569,236900,67375,43682,215960,240564,237284,236924,108,130557,238602,10042,237169,240975,238602,237893,237709,26549,238463,237536],"total_duration":9540299900,"load_duration":4356764800,"prompt_eval_count":19,"prompt_eval_duration":28490800,"eval_count":366,"eval_duration":4822382400}</div></div><div><br /></div><div>context 鍵裡面放的一大堆整數是 "提示詞 + 模型的回答" 的 Token ID, 是用來記錄對話歷史的記憶卡, 用來延續對話, 下一次發送請求時會把這串 context 原封不動放在 Payload 的 "context" 欄位裡傳回去, Ollama 就能立刻讀取這串 ID, 知道剛才聊了什麼, 達到記憶上下文的效果.&nbsp;</div><div><br /></div><div>/api/chat 端點用來回應對話, 它與 /api/generate 最大的不同點是不需要自己管理那串整數的 context (Token ID), 而是直接傳送一組聊天紀錄列表 (messages 串列) 給它, 用 user/system/assistant 三種不同角色來區分誰說的, 格式與 OpenAI API 相同.&nbsp;</div><div><br /></div><div>下面是起始對話 :&nbsp;</div><div><br /></div><div><div>curl.exe http://localhost:11434/api/chat -d '{</div><div>&nbsp; \"model\": \"gemma4:e4b\",&nbsp;</div><div>&nbsp; \"messages\": [{\"role\":&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"user\",&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"content\": \"我有兩隻貓咪, 咪咪與萬萬。\"}],&nbsp;</div><div>&nbsp; \"stream\": false</div><div>}'</div></div><div><br /></div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/chat -d '{</span></b></div><div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"model\": \"gemma4:e4b\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"messages\": [{\"role\":&nbsp;</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">&nbsp;\"user\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">&nbsp;\"content\": \"我有兩隻貓咪, 咪咪與萬萬。\"}],</span></b></div><div>&gt;&gt;&nbsp; <b><span style="color: #2b00fe;">&nbsp;\"stream\": false</span></b></div><div>&gt;&gt; <b><span style="color: #2b00fe;">}'</span></b></div><div>{"model":"gemma4:e4b","created_at":"2026-05-28T15:56:58.0837637Z","message":{"role":"assistant","content":"哇，兩隻貓咪！🐈🐈 🐱\n\n咪咪和萬萬，名字很可愛耶！牠們的個性怎麼樣呢？是活潑好動的類型，還是比較溫柔黏人的？\n\n歡迎多向我分享關於牠們的故事，我很喜歡聽貓咪的故事！🐈❤️","thinking":"Thinking Process:\n\n1.&nbsp; **Analyze the user's input:** The user provided a statement: \"我有兩隻貓咪, 咪咪與萬萬。\" (I have two cats, Mimi and Wanwan.)\n2.&nbsp; **Determine the goal:** The user is sharing personal information. The appropriate response should be warm, engaging, and encourage further interaction, showing interest in the topic.\n3.&nbsp; **Identify key elements for the response:**\n&nbsp; &nbsp; *&nbsp; &nbsp;Acknowledge the information (two cats).\n&nbsp; &nbsp; *&nbsp; &nbsp;Use warm, positive language.\n&nbsp; &nbsp; *&nbsp; &nbsp;Ask open-ended questions related to cats (e.g., their personalities, breeds, feeding habits, or recent activities).\n\n4.&nbsp; **Drafting the response (in Chinese, matching the input language):**\n\n&nbsp; &nbsp; *&nbsp; &nbsp;*Greeting/Acknowledgement:* Oh, you have two cats! (哇，有兩隻貓咪！)\n&nbsp; &nbsp; *&nbsp; &nbsp;*Express interest:* That must be fun/lovely. (一定很可愛/很高興！)\n&nbsp; &nbsp; *&nbsp; &nbsp;*Engage (Question 1 - Personality):* What are they like? Are they playful? (牠們個性怎麼樣？是活潑的嗎？)\n&nbsp; &nbsp; *&nbsp; &nbsp;*Engage (Question 2 - Interaction/Detail):* Do they get along well? (牠們相處融洽嗎？)\n\n5.&nbsp; **Review and refine (Tone check):** The tone is friendly, affectionate, and curious. (This is a good response.)"},"done":true,"done_reason":"stop","total_duration":11303125100,"load_duration":5890298300,"prompt_eval_count":29,"prompt_eval_duration":30451400,"eval_count":383,"eval_duration":5043021600}</div></div><div><br /></div><div>接下來起始第二輪對話, 把上面的第一輪對話內容放入 messages 串列中 (提問角色為 user, 回應角色為 assistant), 另外加入第二輪提問 (role=user) :</div><div><br /></div><div>我家的兩隻貓咪叫甚麼名字?&nbsp;</div><div><br /></div><div><div>curl.exe http://localhost:11434/api/chat -d '{</div><div>&nbsp; \"model\": \"gemma4:e4b\",&nbsp;</div><div>&nbsp; \"messages\": [{\"role\":&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"user\",&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"content\": \"我有兩隻貓咪, 咪咪與萬萬。\"},&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;{\"role\": \"assistant\",&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"content\": \"哇，兩隻貓咪！🐈🐈 🐱\n\n咪咪和萬萬，名字很可愛耶！牠們的個性怎麼樣呢？是活潑好動的類型，還是比較溫柔黏人的？\n\n歡迎多向我分享關於牠們的故事，我很喜歡聽貓咪的故事！🐈❤️\"},&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;{\"role\": \"user\",&nbsp;</div><div>&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; \"content\": \"我家的兩隻貓咪叫甚麼名字?\"}],&nbsp;</div><div>&nbsp; \"stream\": false}'</div></div><div><br /></div><div><br /></div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">curl.exe http://localhost:11434/api/chat -d '{</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"model\": \"gemma4:e4b\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp;<b><span style="color: #2b00fe;">\"messages\": [{\"role\":</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;<b><span style="color: #2b00fe;">\"user\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;<b><span style="color: #2b00fe;">\"content\": \"我有兩隻貓咪, 咪咪與萬萬。\"},</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">{\"role\": \"assistant\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">&nbsp;\"content\": \"哇，兩隻貓咪！🐈🐈 🐱\n\n咪咪和萬萬，名字很可愛耶！牠們的個性怎麼樣呢？是活潑好動的類型，還是比較溫柔黏人的？\n\n歡迎多向我分享關於牠們的故事，我很喜歡聽貓咪的故事！🐈❤️\"},</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; <b><span style="color: #2b00fe;">{\"role\": \"user\",</span></b></div><div>&gt;&gt;&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp;<b><span style="color: #2b00fe;">\"content\": \"我家的兩隻貓咪叫甚麼名字?\"}],</span></b></div><div>&gt;&gt;&nbsp; <b><span style="color: #2b00fe;">&nbsp;\"stream\": false}'</span></b></div><div>{"model":"gemma4:e4b","created_at":"2026-05-28T16:03:10.0080872Z","message":{"role":"assistant","content":"您在最一開始就告訴我了呀！🐱🐱\n\n<span style="background-color: #fcff01;">牠們叫做：\n\n**咪咪 (Mimi)** 和 **萬萬 (Wanwan)**</span> 💖\n\n這麼可愛的名字！😊","thinking":"Thinking Process:\n\n1.&nbsp; **Analyze the user's input:** The user asked: \"我家的兩隻貓咪叫甚麼名字?\" (What are the names of my two cats?).\n2.&nbsp; **Review the previous context:** The user stated earlier: \"我有兩隻貓咪, 咪咪與萬萬。\" (I have two cats, Mimi and Wanwan).\n3.&nbsp; **Identify the contradiction/question:** The user is asking for the names, but they already provided them in the very first turn of the conversation.\n4.&nbsp; **Determine the appropriate response:**\n&nbsp; &nbsp; *&nbsp; &nbsp;Politely point out that the names were mentioned before.\n&nbsp; &nbsp; *&nbsp; &nbsp;Restate the names to confirm the information.\n&nbsp; &nbsp; *&nbsp; &nbsp;Maintain a friendly, conversational, and engaging tone.\n5.&nbsp; **Draft the response (Internal Monologue Check):**\n&nbsp; &nbsp; *&nbsp; &nbsp;*Option A (Too blunt):* You already told me. They are Mimi and Wanwan.\n&nbsp; &nbsp; *&nbsp; &nbsp;*Option B (Ideal - Friendly and Confirming):* 您剛才告訴我了！牠們叫做咪咪和萬萬。\n6.&nbsp; **Final Polish (Chinese):** (Drafting the final empathetic and helpful response.)"},"done":true,"done_reason":"stop","total_duration":8877803600,"load_duration":4535185600,"prompt_eval_count":115,"prompt_eval_duration":57920400,"eval_count":305,"eval_duration":4005207200}</div></div><div><br /></div><div>模型回應了正確答案.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-13629012447336991172026-05-27T19:27:10.183+08:002026-05-27T19:29:21.542+08:00博客來買書一本 : 養成你的AI龍蝦管家<div>最近把 Ollama + 本地模型架起來後, 準備往 AI 代理人 OpenClaw 進發, 上博客來買了阿亮老師的這本養龍蝦好書 :</div><div><br /></div><div>#&nbsp;<a href="https://www.books.com.tw/products/0011049125" target="_blank">養成你的AI龍蝦管家！OpenClaw × NemoClaw × Google Antigravity × Claude Code：打造能查行程、收郵件、遠端控制電腦的AI代理人</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjVDmp8k2lGdN9tPOdQcT39QOXo9XFJDv9abUXafDKTEWfu1E7WCJJhWgc6vwKonYj0HhCi9ByCQ2BVkuHW_EJb63g4-YRGqbE_4O_H_RYkQJYBqFBuRyEn2tYL_yLYvukdpiiPE4K5ioq9mNVR2HP-SaGtKbGbajtRguAXEIOi60PDDKOXU9cQsNpUdNhU/s2085/%E5%8D%9A%E5%AE%A2%E4%BE%86%E8%B2%B7%E6%9B%B8%E4%B8%80%E6%9C%AC-%E9%A4%8A%E6%88%90%E4%BD%A0%E7%9A%84AI%E9%BE%8D%E8%9D%A6%E7%AE%A1%E5%AE%B6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1017" data-original-width="2085" height="156" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjVDmp8k2lGdN9tPOdQcT39QOXo9XFJDv9abUXafDKTEWfu1E7WCJJhWgc6vwKonYj0HhCi9ByCQ2BVkuHW_EJb63g4-YRGqbE_4O_H_RYkQJYBqFBuRyEn2tYL_yLYvukdpiiPE4K5ioq9mNVR2HP-SaGtKbGbajtRguAXEIOi60PDDKOXU9cQsNpUdNhU/s320/%E5%8D%9A%E5%AE%A2%E4%BE%86%E8%B2%B7%E6%9B%B8%E4%B8%80%E6%9C%AC-%E9%A4%8A%E6%88%90%E4%BD%A0%E7%9A%84AI%E9%BE%8D%E8%9D%A6%E7%AE%A1%E5%AE%B6.jpg" width="320" /></a></div><br /><div><br /></div><div>打 79 折後為 592 元, 抵用 OpenPoint 324 點, 實付 268 元, 小七取付.&nbsp;</div><div><br /></div><div>龍蝦好書還有旗標這本, 有需要再買 :</div><div><br /></div><div>#&nbsp;<a href="https://www.books.com.tw/products/0011049678" target="_blank">全民養龍蝦！OpenClaw 超入門</a></div><div><br /></div><div>嘉璽老師的大作因市圖有電子書可借, 就省下來了 (主要是省書架空間考量).&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-31860422076207121572026-05-26T23:17:46.219+08:002026-05-26T23:48:27.704+08:00Ollama 學習筆記 : 安裝網頁介面 Page Assist<div><div><div>在前面的測試中都是使用 Ollama 原生介面或 CLI 介面與本地模型對話, 事實上 Ollama 是一個服務軟體, 它可對多種窗口介面發送服務, 例如 CLI, 特定程式或網頁等, 本篇測試則是要在 Chrome 瀏覽器上安裝擴充元件 Page Assist 以便能用網頁介面與本地模型對話.</div><div><br /></div><div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div></div></div><div><br /></div><div><br /></div><div><b><span style="background-color: #fce5cd; color: #990000;">1. 安裝擴充元件 :</span></b></div><div><br /></div><div>首先來安裝 Chrome 擴充元件 Page Assist, 按 Chrome 右上角的三個點按鈕, 點選 "擴充功能", 再點選 "前往 Chrome 線上應用程式商店" :</div><div><br /></div><div><br /></div></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEijXEC3Cr7Ri3690Gqi1zGQq-QOF5b8dyv8MRtYZcLAospTfneg0LJ-NA4O7X9l9HIrNYLS8EtcgGmY9ri8yvCIekVJ7kbVicdU-BmdfzuOpdW8eHXtFZNd8M5PLcqtMfAUbiOe6tFwUci0FCgAnAC9y-uGoFaEg_1gQfRSKiScNMQYuHo-LnHUC2L2KTgl/s1406/ollama-page-assist-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1406" data-original-width="1355" height="320" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEijXEC3Cr7Ri3690Gqi1zGQq-QOF5b8dyv8MRtYZcLAospTfneg0LJ-NA4O7X9l9HIrNYLS8EtcgGmY9ri8yvCIekVJ7kbVicdU-BmdfzuOpdW8eHXtFZNd8M5PLcqtMfAUbiOe6tFwUci0FCgAnAC9y-uGoFaEg_1gQfRSKiScNMQYuHo-LnHUC2L2KTgl/s320/ollama-page-assist-1.jpg" width="308" /></a></div><br /><div><br /></div><div>在右上角的搜尋框輸入 "Page Asist", 點選搜尋到的項目 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhnNSDRowlujkLJuuCJ7YAvIjegEASO6KyACv-vj7STgCq-UcW_8Q52QdHCeaFo664Lbql-lU1OghJeSzWnJVRiH3i1nyUm8Cy_IISElB1AaNQWE4_Kwa0kuEB4JmF60MMh0iLtlI7pRwo1KQ-6zRNQZACW-wzi4oGCAYNMqSkHv5PWGX_V00es8thJtySV/s2045/ollama-page-assist-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1134" data-original-width="2045" height="177" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhnNSDRowlujkLJuuCJ7YAvIjegEASO6KyACv-vj7STgCq-UcW_8Q52QdHCeaFo664Lbql-lU1OghJeSzWnJVRiH3i1nyUm8Cy_IISElB1AaNQWE4_Kwa0kuEB4JmF60MMh0iLtlI7pRwo1KQ-6zRNQZACW-wzi4oGCAYNMqSkHv5PWGX_V00es8thJtySV/s320/ollama-page-assist-2.jpg" width="320" /></a></div><br /><div><br /></div><div><br /></div><div>按右上角的 "加入到 Chrome" 鈕 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEibdVoPVw3aXhr2csxQNTJxkMMRh1lubL5AZJfrjlbIVrciMzDZ4WsT-VKK-lxk0Vg6jlpQZb4wPWQZ8SrOlcC7rDwt5vwwNlUzFPdq06BwTyOoR6dBYt5Fgexhf0CDSsEWwJ7w6Gkt9025GoNMNVBHBYxQnEt0bU-X5jI9_CI9JYv1UqOF7vuIR7OPp8iy/s2343/ollama-page-assist-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1174" data-original-width="2343" height="160" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEibdVoPVw3aXhr2csxQNTJxkMMRh1lubL5AZJfrjlbIVrciMzDZ4WsT-VKK-lxk0Vg6jlpQZb4wPWQZ8SrOlcC7rDwt5vwwNlUzFPdq06BwTyOoR6dBYt5Fgexhf0CDSsEWwJ7w6Gkt9025GoNMNVBHBYxQnEt0bU-X5jI9_CI9JYv1UqOF7vuIR7OPp8iy/s320/ollama-page-assist-3.jpg" width="320" /></a></div><br /><div><br /></div><div>按 "新增擴充功能" 鈕 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZMPD11bYX6ozSRvOVoy2gdn3VQEFZOAtNS4rlfrYXlCE0nZ4ENZ6YnUmh1ub4jc3MMHjqaoDx_rPMOp5U9ulOAppwR6_7SOsBfjMMCGtle6xevskPFtLYKuWMH2oItSFD0J4-XrKWI6yezjAdjnDeK2OQ9x_3ExSCpZ4wCA0qahLHPVRqIasvwMf5qElS/s963/ollama-page-assist-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="471" data-original-width="963" height="157" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhZMPD11bYX6ozSRvOVoy2gdn3VQEFZOAtNS4rlfrYXlCE0nZ4ENZ6YnUmh1ub4jc3MMHjqaoDx_rPMOp5U9ulOAppwR6_7SOsBfjMMCGtle6xevskPFtLYKuWMH2oItSFD0J4-XrKWI6yezjAdjnDeK2OQ9x_3ExSCpZ4wCA0qahLHPVRqIasvwMf5qElS/s320/ollama-page-assist-4.jpg" width="320" /></a></div><br /><div><br /></div><div>按上方 Chrome 擴充功能鈕會出現所有擴充功能項目, 按 Page Asist 那列右邊的圖釘將其固定在上方工具列上 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgqOOVKTXL-rMOObaUxU8153tb_INd-zo_E4JmznYnlqB1SmfIiHWjE_bEbXboJ2TUm7oq2pHxyLMARzVGnCPYrEoZIojih2o1loyT5LsWwK_Mlt6dTDe_uiTBaJaQ_mrnwjMgcEBKghzPzqi_scbqvWgYM5VLFCeLYCNJTcTDeDMppNEFOCV08LpZ2-nZ5/s1503/ollama-page-assist-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1123" data-original-width="1503" height="239" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgqOOVKTXL-rMOObaUxU8153tb_INd-zo_E4JmznYnlqB1SmfIiHWjE_bEbXboJ2TUm7oq2pHxyLMARzVGnCPYrEoZIojih2o1loyT5LsWwK_Mlt6dTDe_uiTBaJaQ_mrnwjMgcEBKghzPzqi_scbqvWgYM5VLFCeLYCNJTcTDeDMppNEFOCV08LpZ2-nZ5/s320/ollama-page-assist-5.jpg" width="320" /></a></div><br /><div><br /></div><div><br /></div><div>這樣以後 Page Assist 按鈕就會一直出現在上方工具列, 直接按就會開啟 Page Assist 頁籤了 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg5ALNUwGdACKh4yL_rDXTbGioV1vOigfQT9gRWqWzppfbr_oyV7WtuntGEcAiTK1toJDPTREM0KMOa0IfltPBN77HXVEo_rlFfQbF5HlhO-EKbVoQs-NVx5eeT1U5AxQA89e_vl7ar2FvfVMHEqvL0ssIP1XmJaUzV1daUR85Gw56m4asxS4bLCHOaivlN/s1501/ollama-page-assist-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="825" data-original-width="1501" height="176" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg5ALNUwGdACKh4yL_rDXTbGioV1vOigfQT9gRWqWzppfbr_oyV7WtuntGEcAiTK1toJDPTREM0KMOa0IfltPBN77HXVEo_rlFfQbF5HlhO-EKbVoQs-NVx5eeT1U5AxQA89e_vl7ar2FvfVMHEqvL0ssIP1XmJaUzV1daUR85Gw56m4asxS4bLCHOaivlN/s320/ollama-page-assist-6.jpg" width="320" /></a></div><div>點模型框會出現已下載的模型選單供選擇 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgHEBsU1nnSZWC9_qb8inBGA-eAu1L5Ch96kxSDrYEduRTOPzOcOWEzV_C7ndr5HLqjWYxrOgSFTrh4or3FztPCt-jBNI54qUWGQxpgle7VEnOrYjLMMz-v49AUa7KpKbF-3BebsClGcl_LUxIDeBxzz3l_AbFNrNcIgHl0NhIHWP_zPI9orGKXvVPpNcsw/s1392/ollama-page-assist-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="499" data-original-width="1392" height="115" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgHEBsU1nnSZWC9_qb8inBGA-eAu1L5Ch96kxSDrYEduRTOPzOcOWEzV_C7ndr5HLqjWYxrOgSFTrh4or3FztPCt-jBNI54qUWGQxpgle7VEnOrYjLMMz-v49AUa7KpKbF-3BebsClGcl_LUxIDeBxzz3l_AbFNrNcIgHl0NhIHWP_zPI9orGKXvVPpNcsw/s320/ollama-page-assist-1.jpg" width="320" /></a></div><br /><div><br /></div><div>不過在對話前先來作操作介面設定.&nbsp;</div><div><br /></div><div><br /></div><div><div><b><span style="background-color: #fce5cd; color: #990000;">2. 操作介面設定 :</span></b></div></div><div><br /></div><div>按右上角的齒輪按鈕進入設定頁面 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi8G7t4-lFjINzOHgPg_FQHx3tUFWk0JNM0SP9aBx2K7Mzhr-FgoWlgI_bOC2si1cXJs1h4gDt8gloN1Nh4MIKFPQ9EcGtwyYMkw8MN9rUchwL0u3xlAnWjyr5NOuquCWY-dtbKpTgjGj39U2lxhIjQ0dsA1UeKWVt9KpDQxJRv-Z5ZG-NXef4ONF6Ip4K3/s410/ollama-page-assist-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="284" data-original-width="410" height="222" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi8G7t4-lFjINzOHgPg_FQHx3tUFWk0JNM0SP9aBx2K7Mzhr-FgoWlgI_bOC2si1cXJs1h4gDt8gloN1Nh4MIKFPQ9EcGtwyYMkw8MN9rUchwL0u3xlAnWjyr5NOuquCWY-dtbKpTgjGj39U2lxhIjQ0dsA1UeKWVt9KpDQxJRv-Z5ZG-NXef4ONF6Ip4K3/s320/ollama-page-assist-2.jpg" width="320" /></a></div><br /><div>Page Assist 預設是英文介面, 可以在 Language 選項選擇繁體中文 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjmoaoDC6CWjb9wWC4YYA_IMcDh-309KR3Mq3FP5zZvdKuUjpIq9VFu6AJN5uW_lNw0v6rXrLZXsqNTfxxDr0iNYDX5gJms214pdMVe3qzmWPBfXcM45helpeihQC7KhRSFNuOi1-Tz_ayMTm45RWQdF1rr-yy4tuKEgBRolo8GJiXjumh6lBYz4_byfqsD/s1134/ollama-page-assist-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="639" data-original-width="1134" height="180" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjmoaoDC6CWjb9wWC4YYA_IMcDh-309KR3Mq3FP5zZvdKuUjpIq9VFu6AJN5uW_lNw0v6rXrLZXsqNTfxxDr0iNYDX5gJms214pdMVe3qzmWPBfXcM45helpeihQC7KhRSFNuOi1-Tz_ayMTm45RWQdF1rr-yy4tuKEgBRolo8GJiXjumh6lBYz4_byfqsD/s320/ollama-page-assist-3.jpg" width="320" /></a></div><br /><div><br /></div><div><br /></div><div>在一般設定中常用的項目是更改 "OCR 預設語言" (預設為 English) 與 "更改主題" (預設為亮色) :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg6rFVI9_oRu2cYFM-5csLhEUVwQLvfiC11ezHx8Ofg7_ybuSP10XqENtZFCiHUio_FjbkwZ74kCL7TdlzCP4oaMc80Vj02tXqZtHQstIOPVWRZEoRaL_7jcVY6FmwaTnNSBGyIOSd-yuGmbRr7VtHyCKjgHLngQ-Vk_hOt5H13izuSEGCjPW7zRX3wOBrW/s1118/ollama-page-assist-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="531" data-original-width="1118" height="152" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg6rFVI9_oRu2cYFM-5csLhEUVwQLvfiC11ezHx8Ofg7_ybuSP10XqENtZFCiHUio_FjbkwZ74kCL7TdlzCP4oaMc80Vj02tXqZtHQstIOPVWRZEoRaL_7jcVY6FmwaTnNSBGyIOSd-yuGmbRr7VtHyCKjgHLngQ-Vk_hOt5H13izuSEGCjPW7zRX3wOBrW/s320/ollama-page-assist-4.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEivfX_-MLiy-0IXRYrIOVOYYZVObRLcYB20BZsO_Vl_XDob6cUYjTzRAHhVxuZJSE9j-qbH6kGACAbtOyqzMCxd_uDdd40ZsAYdvMD5BFnTRW_QuNHWYHWZrxWqDAa6erh35hj54UeoPeI7D7c9d0DfiCdQOIUHHBt5hdPF6BYsaHANfkoXcfILhapbrJKY/s1127/ollama-page-assist-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="484" data-original-width="1127" height="137" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEivfX_-MLiy-0IXRYrIOVOYYZVObRLcYB20BZsO_Vl_XDob6cUYjTzRAHhVxuZJSE9j-qbH6kGACAbtOyqzMCxd_uDdd40ZsAYdvMD5BFnTRW_QuNHWYHWZrxWqDAa6erh35hj54UeoPeI7D7c9d0DfiCdQOIUHHBt5hdPF6BYsaHANfkoXcfILhapbrJKY/s320/ollama-page-assist-5.jpg" width="320" /></a></div><div><br /></div><br /><div>網頁搜尋設定可以選擇模型使用的搜尋引擎 (預設 DuckDuckGo) 與更改搜尋結果筆數等 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhip3ki9Yf5LhwfbaHY62ve0etjFiFKK8D0y_wKhZ3bx36ZiQnPhb9syez1dP0e93W77MNEezf_r73lNgzJuvChauMhBRAvyILK8XlYVSFN2ebikuDjqlb8WwawjEdNiIkm7fWqE2-fviwTaJGn0fGwKHMIt271sY1EJ9XVclPSdQU3d9nqvpj0QyYm1hhU/s1111/ollama-page-assist-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="471" data-original-width="1111" height="136" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhip3ki9Yf5LhwfbaHY62ve0etjFiFKK8D0y_wKhZ3bx36ZiQnPhb9syez1dP0e93W77MNEezf_r73lNgzJuvChauMhBRAvyILK8XlYVSFN2ebikuDjqlb8WwawjEdNiIkm7fWqE2-fviwTaJGn0fGwKHMIt271sY1EJ9XVclPSdQU3d9nqvpj0QyYm1hhU/s320/ollama-page-assist-6.jpg" width="320" /></a></div><br /><div><br /></div><div>語音轉文字部分可以設定語言 (預設 English) 等 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhXW1IO7pczijrIOp36x9ZB9k4N3s7Xlt2AkLTGnB-KrqKBQIPGd9zZhb5iBz6WGPe-fTYXkGCnhmAIHEPUYkVqd5nJPd2nymTe7RXt1U-w8a2sDmuloqAJdnRDCvwDk3rTBFA1Hw2tlUG_XrX5bUsCLCnBY5lUUO4hiOmrhTDjIQ1jBJfQ_gBSui55mhR1/s1112/ollama-page-assist-7.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="684" data-original-width="1112" height="197" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhXW1IO7pczijrIOp36x9ZB9k4N3s7Xlt2AkLTGnB-KrqKBQIPGd9zZhb5iBz6WGPe-fTYXkGCnhmAIHEPUYkVqd5nJPd2nymTe7RXt1U-w8a2sDmuloqAJdnRDCvwDk3rTBFA1Hw2tlUG_XrX5bUsCLCnBY5lUUO4hiOmrhTDjIQ1jBJfQ_gBSui55mhR1/s320/ollama-page-assist-7.jpg" width="320" /></a></div><div><br /></div><br /><div>系統設定部分則可以更改頁面字型大小, 匯入/匯出聊天紀錄等 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjvSMemaod_DMlVlSv2HwMGS94MaaE5gNtolyO13orY8GHJ27vbKwAGv33ag2TTv-Bk57YPyx1ftg7Q0JEF3wvsB7DcpPEuA86kPE4VtwUcVOs9HI0fEhxWcPw9RXGBZVIYJyilQpA8Gl1pej9B5UvW0VBOemrC3HWLmqsanQeYAK7L2za9yIvOmG7oPNwS/s1114/ollama-page-assist-8.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="560" data-original-width="1114" height="161" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjvSMemaod_DMlVlSv2HwMGS94MaaE5gNtolyO13orY8GHJ27vbKwAGv33ag2TTv-Bk57YPyx1ftg7Q0JEF3wvsB7DcpPEuA86kPE4VtwUcVOs9HI0fEhxWcPw9RXGBZVIYJyilQpA8Gl1pej9B5UvW0VBOemrC3HWLmqsanQeYAK7L2za9yIvOmG7oPNwS/s320/ollama-page-assist-8.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div><div><b><span style="background-color: #fce5cd; color: #990000;">3. 與模型對話 :</span></b></div><div><br /></div></div><div>在前一篇測試中我們使用 Ollama 原生介面與 Gemma 4 多模態模型對話, 上傳圖片檔要求描述或辨識圖片內容, 此處改用 Page Assist 的網頁介面來與模型交談.&nbsp;</div><div><br /></div><div>我借用程序猿技術部落格關於 Mermaid 介紹的心智圖, 截圖存檔 (jpg) 後上傳給 gemma4 模型, 要求它讀取圖片中的訊息並轉成 Markdown 格式 :</div><div><br /></div><div>#&nbsp;<a href="https://programmur.substack.com/p/chatgpt-mermaid-3-minute-diagram-productivity" target="_blank">3 分鐘打造資訊圖表 - ChatGPT + Mermaid 輕鬆提升生產力</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjjaBmr01eKlNqYDOWhEEPSzNigf042T1zP0oaaf8KF33nTmejhv-GPpEoIT2mpAsCi3WpyKq91N-Q9xGw6Z5lxASvEaVIgmCDwm2StzFMyz2MbEpLwcO5lsOHg2nns6ir1Af_dlczcTx2Kk93-VqEpmEjVoPetpyG5KWj5DqkOv4Pmqss-8ldjUQZPbl2O/s1076/what-is-mermaid.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="559" data-original-width="1076" height="166" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjjaBmr01eKlNqYDOWhEEPSzNigf042T1zP0oaaf8KF33nTmejhv-GPpEoIT2mpAsCi3WpyKq91N-Q9xGw6Z5lxASvEaVIgmCDwm2StzFMyz2MbEpLwcO5lsOHg2nns6ir1Af_dlczcTx2Kk93-VqEpmEjVoPetpyG5KWj5DqkOv4Pmqss-8ldjUQZPbl2O/s320/what-is-mermaid.jpg" width="320" /></a></div><br /><div><br /></div><div>選擇 gemma4 模型後輸入提示詞 :&nbsp;</div><div><br /></div><div><i><span style="color: #2b00fe;">請將這張心智圖內容轉成 Markdown 格式</span></i></div><div><br /></div><div>點圖片按鈕選取要上傳的心智圖後提交 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEinX-FwFr-NTXn_-IMy96_4ILU5pwoT6_NcmXbLiEOE3mXvSfkvVrnCU9Wta5-Nw_hX5SeYIJGw2PxvrlmHKJOb4gDkZATtEprt6LG0wtIcfKB0TmefrBS99BEpRcqwYg8yPcE6vKgVc4nUaIJTqP_ITTTfosr-Pj-q7D2ESEDnd3PQM4IFzLYisMXlsQ66/s1188/ollama-page-assist-9.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="454" data-original-width="1188" height="122" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEinX-FwFr-NTXn_-IMy96_4ILU5pwoT6_NcmXbLiEOE3mXvSfkvVrnCU9Wta5-Nw_hX5SeYIJGw2PxvrlmHKJOb4gDkZATtEprt6LG0wtIcfKB0TmefrBS99BEpRcqwYg8yPcE6vKgVc4nUaIJTqP_ITTTfosr-Pj-q7D2ESEDnd3PQM4IFzLYisMXlsQ66/s320/ollama-page-assist-9.jpg" width="320" /></a></div><br /><div>結果如下 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhPB8IObOl21sIuXZgcvEtbAGrqKyU5FLqy4aecQde_zizWxmU6eUUFEvAczg3co9K0Mc_-BGZfziLUeoVOJXW3frfb9B0B3urZ9ni48ewOMmRt77t2vx0FKuJV7R1etYgV7nwdqZdyjFqU3DlHos3WTxCLAAVCF0-TMNAOepxp3LocKemKbEvAvcDP9n4L/s884/ollama-page-assist-10.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="577" data-original-width="884" height="209" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhPB8IObOl21sIuXZgcvEtbAGrqKyU5FLqy4aecQde_zizWxmU6eUUFEvAczg3co9K0Mc_-BGZfziLUeoVOJXW3frfb9B0B3urZ9ni48ewOMmRt77t2vx0FKuJV7R1etYgV7nwdqZdyjFqU3DlHos3WTxCLAAVCF0-TMNAOepxp3LocKemKbEvAvcDP9n4L/s320/ollama-page-assist-10.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjnPkqYLovLzPu9vHRUXKHZHCjfD24gKfgYltk3t0g7bDBo9opthFiBnTXBAlOjHe5E79RZxQhWfjtwLe8GZVtaCSp0fEH5fALehb5yYy5g77oLLPHqeu7HQIzjX5z3G0fk88zZ2QZ1xUfC__e41pEnq0gTEamMTDlyDi-boCOtG0CSmCmlMbQa4W8PwOAn/s908/ollama-page-assist-11.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="609" data-original-width="908" height="215" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjnPkqYLovLzPu9vHRUXKHZHCjfD24gKfgYltk3t0g7bDBo9opthFiBnTXBAlOjHe5E79RZxQhWfjtwLe8GZVtaCSp0fEH5fALehb5yYy5g77oLLPHqeu7HQIzjX5z3G0fk88zZ2QZ1xUfC__e41pEnq0gTEamMTDlyDi-boCOtG0CSmCmlMbQa4W8PwOAn/s320/ollama-page-assist-11.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiJhmrOi3UU5I_irJiIEyM8KkrhOeBSdhQQZ8kFoyWgsxwT0f0eJaoIKf6eIzO-aEnmF24oqzubtNgD0gWSFo_Jwlpe0w7BXTinTEEycMYm0w5QpsWuvxiNTotZxQwUaTYSzzkRmATnaZzFHWSkm8cgAdFPbz_WfXnP3YCqtwIPuHdYuW6WSyWQ3WmXzYAL/s918/ollama-page-assist-12.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="758" data-original-width="918" height="264" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiJhmrOi3UU5I_irJiIEyM8KkrhOeBSdhQQZ8kFoyWgsxwT0f0eJaoIKf6eIzO-aEnmF24oqzubtNgD0gWSFo_Jwlpe0w7BXTinTEEycMYm0w5QpsWuvxiNTotZxQwUaTYSzzkRmATnaZzFHWSkm8cgAdFPbz_WfXnP3YCqtwIPuHdYuW6WSyWQ3WmXzYAL/s320/ollama-page-assist-12.jpg" width="320" /></a></div><br /><div><br /></div><div>可見多模態對話還是網頁介面方便, 終於有 ChatGPT 的 Fu 了.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-87738317185906582412026-05-25T23:38:11.646+08:002026-05-26T00:30:06.095+08:00Ollama 學習筆記 : Gemma 4 模型測試 (gemma4:e4b)<div>在前面的測試中, 安裝 Ollama 後第一個下載執行的模型是 gemma4, 但只是做了簡單的對話而已, 本篇旨在對此模型做進一步測試.</div><div><br /></div><div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a></div></div><div><br /></div><div>以下摘要整理 Gemini 關於 gemma4 的描述 :</div><div><br /></div><div><div>Gemma 4 是 Google 最新釋出的全新世代旗艦級開源大語言模型 (由 Gemini 3 技術下放打造), 採用 Apache 2.0 授權, 允許完全自由商用. 此模型最大的突破在於把原本屬於雲端旗艦的 AI 模型 (例如 Gemini 1.5/2.0 Pro) 的原生全模態 (文字/圖片/音訊) 與類似 DeepSeek-R1 的內建思維鏈推理完美融合, 並且在架構上首次引進了混合專家 (MoE) 與高效的有效參數 (Effective Parameters) 技術, 把小模型的智商推向了全新巔峰.&nbsp;</div></div><div><br /></div><div>Gemma 4 具有如下特點 :&nbsp;</div><div><ul style="text-align: left;"><li>會思考的開源大腦 (Native Reasoning) :<br />Gemma 4 全系列都支援 thinking 模式, 在回答複雜的數學, 科學或軟體工程代碼時, 它會先進行內部邏輯推演 (Chain-of-Thought), 想通了才動手. 這讓它在軟體工程解題測試上的成功率飆升, 抓 Bug 和重構 Python 程式碼的實力直逼一線雲端大模型.&nbsp;</li><li>原生全模態感知 (能看能聽) :<br />底層原生支援全模態感知, 不僅能精準做圖片 OCR, 解讀複雜商業圖表, 偵測網頁 UI 元素, 甚至支援音訊原生輸入 (其音訊編碼器比上一代縮小了 77%).&nbsp;</li><li>強悍的 Agent 與工具呼叫能力 :<br />經過極其嚴格的 Function Calling 優化後, 在像 Claude Code, OpenClaw 或是 Android Studio 新版的本地 AI 代理工作流中, 它輸出 JSON 的穩定度極高, 不會輕易導致 Agent 介面崩潰, 是目前最適合拿來當 Local Coding Agent 後台的小模型.&nbsp;</li><li>快如閃電的多詞元預測 (MTP) 技術 :&nbsp;<br />Gemma 4 內建了專用的草稿模型進行推測解碼 (Speculative Decoding), 這讓它的吐字速度非常驚人, 即便是開啟 Thinking 模式依然能維持極度順暢的生成體驗.&nbsp;</li><li>恐怖的 128K 超長上下文記憶 :<br />此模型支援高達 128,000 個 Token 的上下文, 即使是把一整套複雜的 Python 專案架構完全塞給它, 也能完全離線地保持脈絡記憶.</li></ul></div><div>不過 Gemma 4 也有一些缺點 :</div><div><ul style="text-align: left;"><li>靜態記憶體開銷高於預期 :<br />Gemma 4 採用了先進的每層嵌入 (PLE)技術來壓縮核心運算, 雖然它的有效活躍參數很低, 但因為它包含了龐大的靜態嵌入表與龐大的 128K KV 快取空間, 它在剛載入顯卡時的 VRAM 佔用率會比傳統純文字 8B 模型稍微吃滿一點點.&nbsp;</li><li>複雜思考會拉長首字反應時間 (Time-to-First-Token) :<br />因為模型內建了推理機制, 當提出一個非常困難的程式邏輯題時, 它會在背景面壁思索一小段時間, 這時候 GPU 雖然在瘋狂運算, 但畫面上可能要等個 1~2 秒才會開始吐字, 可能會讓人誤以為它卡住了.&nbsp;</li></ul></div><div>總評 :&nbsp;Gemma 4 是全面進化的全能全模態大師, 我們可以將網頁設計截圖丟給它叫它直接生成 HTML/CSS 碼, 也可以作 Claude Code 後台, 全自動為 Python 專案生成程式碼與除錯, 還能丟錄音檔或語音叫它離線做精準摘要與理解.&nbsp;</div><div><br /></div><div>Gemma 4 的系列模型參考 :</div><div><br /></div><div>#&nbsp;<a href="https://ollama.com/library/gemma4" target="_blank">https://ollama.com/library/gemma4</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiOCmGI9etF68SRLkG6zcze6Y6iVObfFFmZYCo9mA5oVfvt3Gp1aZlO5PyD1uLjaiyPe80WKz9ukHdVH4USSTiY2YHLmRHfSXpae6XzPDG5Am4RsUc5sKZ7f-fIyA94wye5cY4emJxH9eBhUknbFZF1TbJUQep4hCB_YZLv4Jzwo9k3awrW4gghLpelp0YF/s1736/ollama-gemma4-e4b-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1195" data-original-width="1736" height="220" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiOCmGI9etF68SRLkG6zcze6Y6iVObfFFmZYCo9mA5oVfvt3Gp1aZlO5PyD1uLjaiyPe80WKz9ukHdVH4USSTiY2YHLmRHfSXpae6XzPDG5Am4RsUc5sKZ7f-fIyA94wye5cY4emJxH9eBhUknbFZF1TbJUQep4hCB_YZLv4Jzwo9k3awrW4gghLpelp0YF/s320/ollama-gemma4-e4b-1.jpg" width="320" /></a></div><br /><div><br /></div><div>之前我用 Ollama 原生介面下載的應該是 8B 參數, 經 Q4_K_M 量化後大小為 9.6 GB 的 e4b 版, 可用 ollama show 指令來確認 :</div><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">ollama show gemma4</span></b>&nbsp; &nbsp;</div><div>&nbsp; Model</div><div>&nbsp; &nbsp; architecture&nbsp; &nbsp; &nbsp; &nbsp; gemma4</div><div>&nbsp; &nbsp; parameters&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 8.0B</div><div>&nbsp; &nbsp; context length&nbsp; &nbsp; &nbsp; 131072</div><div>&nbsp; &nbsp; embedding length&nbsp; &nbsp; 2560</div><div>&nbsp; &nbsp; quantization&nbsp; &nbsp; &nbsp; &nbsp; Q4_K_M</div><div>&nbsp; &nbsp; requires&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 0.20.0</div><div><br /></div><div>&nbsp; Capabilities</div><div>&nbsp; &nbsp; completion</div><div>&nbsp; &nbsp; vision</div><div>&nbsp; &nbsp; audio</div><div>&nbsp; &nbsp; tools</div><div>&nbsp; &nbsp; thinking</div><div><br /></div><div>&nbsp; Parameters</div><div>&nbsp; &nbsp; temperature&nbsp; &nbsp; 1</div><div>&nbsp; &nbsp; top_k&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 64</div><div>&nbsp; &nbsp; top_p&nbsp; &nbsp; &nbsp; &nbsp; &nbsp; 0.95</div><div><br /></div><div>&nbsp; License</div><div>&nbsp; &nbsp; Apache License</div><div>&nbsp; &nbsp; Version 2.0, January 2004</div><div>&nbsp; &nbsp; ...</div></div><div><br /></div><div>如果要在 CLI 介面下載執行, 指令如下 :</div><div><br /></div><div><b><span style="color: #2b00fe;">ollama run gemma4:e4b</span></b>&nbsp;&nbsp;</div><div><br /></div><div>以下多模態測試使用 Ollama 原生介面, 首先測試車牌辨識, 選擇模型 gemma4, 輸入 "請辨識這張車牌號碼", 按輸入框中的 + 鈕, 上傳車牌圖片後按提交鈕 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiYV2uXKrQpjqlFUIB74sr_MbT0RS0SYk7qYb9-NU2TDBr7nrP5cXoxgthol5Tkchgfny0134dlc4SwB5hu_av3gGW9lT8oFQiAdRC8cgmHhpk4c3vgTD-hmXP6dVWqsCn-Ndgtdw3fChlqRPxGgvqKAVnjX9aLR6x1pZV1nykdihkheX6cs4dOuqFzxL8g/s640/NBX-3388.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="359" data-original-width="640" height="180" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiYV2uXKrQpjqlFUIB74sr_MbT0RS0SYk7qYb9-NU2TDBr7nrP5cXoxgthol5Tkchgfny0134dlc4SwB5hu_av3gGW9lT8oFQiAdRC8cgmHhpk4c3vgTD-hmXP6dVWqsCn-Ndgtdw3fChlqRPxGgvqKAVnjX9aLR6x1pZV1nykdihkheX6cs4dOuqFzxL8g/s320/NBX-3388.jpg" width="320" /></a></div><br /><div><br /></div><div>經過快速的推理思考後, 直接回覆正確之車牌號碼 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLMTOWXZ14cKDmtoLpRDpW1_BDUrSelib_MNYRtFoThTUAnn-C3NIMYv6kO_zd8SUrvfPP4e9l9t6pVRTmCtTgm1NJLGhvJguemeyqdA9W_NP10OUBmVoAB5aeaznSyvsO1CYPiBLNYozu1jxXdQKHw5uvy_aDxV0szC2JueIXuZBEZtdGdLxJ-f1O5r8W/s1048/ollama-gemma4-e4b-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="761" data-original-width="1048" height="232" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiLMTOWXZ14cKDmtoLpRDpW1_BDUrSelib_MNYRtFoThTUAnn-C3NIMYv6kO_zd8SUrvfPP4e9l9t6pVRTmCtTgm1NJLGhvJguemeyqdA9W_NP10OUBmVoAB5aeaznSyvsO1CYPiBLNYozu1jxXdQKHw5uvy_aDxV0szC2JueIXuZBEZtdGdLxJ-f1O5r8W/s320/ollama-gemma4-e4b-2.jpg" width="320" /></a></div><br /><div><br /></div><div>下面是從 TVBS 新聞網站下載的貨車圖片, 上傳給 gemma4 看看能否辨識車牌 :</div><div><br /></div><div>#&nbsp;<a href="https://news.tvbs.com.tw/fun/810520" target="_blank">https://news.tvbs.com.tw/fun/810520</a></div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjlTzwvXvwRskPR6Fml9R-_9PiRh2dGRRnQcYAFDdkxkI_0LgDW_fWuhXX1xheYwmgVcDjBm1gLaVgauV5jJw1RQFtFrnKyAmKYiAuBdWMq-IrJ5czM1vbR_0KcWb71Eae-ua7prv0PWnq5MYQvFa11GkLxz8DW3zd62YgLulI8Awvpl1wSTSYGcjgwAB2i/s779/ollama-gemma4-4eb-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="543" data-original-width="779" height="223" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjlTzwvXvwRskPR6Fml9R-_9PiRh2dGRRnQcYAFDdkxkI_0LgDW_fWuhXX1xheYwmgVcDjBm1gLaVgauV5jJw1RQFtFrnKyAmKYiAuBdWMq-IrJ5czM1vbR_0KcWb71Eae-ua7prv0PWnq5MYQvFa11GkLxz8DW3zd62YgLulI8Awvpl1wSTSYGcjgwAB2i/s320/ollama-gemma4-4eb-3.jpg" width="320" /></a></div><div><br /></div><br /><div>果然不同凡響, 正確辨識出&nbsp; 9487-XD.&nbsp;</div><div><br /></div><div>接下來請它描述有名的 Lenna 圖 :&nbsp;</div><div><br /></div><div>"請描述這張圖片, 此圖非常有名, 你知道是甚麼圖嗎?"</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh_TEt0qAbaE4meT8vhpw55_IlfOROn7kcu9JEIPhVGUupU0COg0jvgzckw5b1Y6GuC0bXLADdonHZpH1-0EoX7A5tSRS-qpwzvSXcZNDhclsI8jyMbFTbrUOSMoVliXAi-p6MqGlo6fu6Ya_fhX3GN7-oqGttS59rtEdXLUy09COVnj27lp6F4k7xnSi20/s770/ollama-gemma4-4eb-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="574" data-original-width="770" height="239" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEh_TEt0qAbaE4meT8vhpw55_IlfOROn7kcu9JEIPhVGUupU0COg0jvgzckw5b1Y6GuC0bXLADdonHZpH1-0EoX7A5tSRS-qpwzvSXcZNDhclsI8jyMbFTbrUOSMoVliXAi-p6MqGlo6fu6Ya_fhX3GN7-oqGttS59rtEdXLUy09COVnj27lp6F4k7xnSi20/s320/ollama-gemma4-4eb-4.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiRgqfNuplnfBkCwtlMQJCCUdn2SSVuYimIeRLYc1HpmWi1f8bHaPF3_fZqvOrYPs-kWYk3UaTa0zeLQkDbr_pXGXaMFdQOnYxMUZDUchrn4zfpi8xgIP_FRiJOZTIBUmxFLsX12C-4TsgjetIItfGsnCOXWQSI2lOLGkq2mXRFtHTsV3A_F94u9u_jqpjd/s764/ollama-gemma4-4eb-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="665" data-original-width="764" height="279" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiRgqfNuplnfBkCwtlMQJCCUdn2SSVuYimIeRLYc1HpmWi1f8bHaPF3_fZqvOrYPs-kWYk3UaTa0zeLQkDbr_pXGXaMFdQOnYxMUZDUchrn4zfpi8xgIP_FRiJOZTIBUmxFLsX12C-4TsgjetIItfGsnCOXWQSI2lOLGkq2mXRFtHTsV3A_F94u9u_jqpjd/s320/ollama-gemma4-4eb-5.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhdFsZydHdIUTmlvvaLMp3qU2J1aLHL1rkSsm2eTzoKltFl0XY2C5aHpn4jH0XMRnG78solAwRLqeO_1G_7ufAt0fibozRhaBuqcmKrbKNHJ_DDzS4_K4Wu1OduaVvZVKsLfsOliGV_yax9RhRI2jb_5H262iNEnUnOSXVphZM_TGUznPsXoaX4yMAOLqD_/s768/ollama-gemma4-4eb-6.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="652" data-original-width="768" height="272" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhdFsZydHdIUTmlvvaLMp3qU2J1aLHL1rkSsm2eTzoKltFl0XY2C5aHpn4jH0XMRnG78solAwRLqeO_1G_7ufAt0fibozRhaBuqcmKrbKNHJ_DDzS4_K4Wu1OduaVvZVKsLfsOliGV_yax9RhRI2jb_5H262iNEnUnOSXVphZM_TGUznPsXoaX4yMAOLqD_/s320/ollama-gemma4-4eb-6.jpg" width="320" /></a></div><br /><div><br /></div><div>描述得很詳盡, 但它不知道這是有名的 Lenna 圖, 經我提示 "<span style="color: #2b00fe;"><i>是影像處理書籍上經常出現的模特兒 Lenna 的照片</i></span>" 後才恍然大悟 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEii2fVK1-oikD_8OwdsMagMcaqp5mijBBvcPuT9oxIu9dPSVtQQZlkBHLTiT5hz-M4VjYqmZDHhFK6PHw7VtKXM-Pn6OIJaU78VzcBIYjP0NqyZRbqHGl9hyixXWtLl8wSKusgWNBWi8xrAjtZBGowVxJHi_Z7u28HMVKFqAF1RrQOYgNeIPFCL3uVi48BJ/s767/ollama-gemma4-4eb-7.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="664" data-original-width="767" height="277" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEii2fVK1-oikD_8OwdsMagMcaqp5mijBBvcPuT9oxIu9dPSVtQQZlkBHLTiT5hz-M4VjYqmZDHhFK6PHw7VtKXM-Pn6OIJaU78VzcBIYjP0NqyZRbqHGl9hyixXWtLl8wSKusgWNBWi8xrAjtZBGowVxJHi_Z7u28HMVKFqAF1RrQOYgNeIPFCL3uVi48BJ/s320/ollama-gemma4-4eb-7.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjwlVkX7zuD_gyQHv-emNdjY47nx1m9Z0r77IhMktpHA3trI6Yu2VX3mLLwPN0zOPWw96PYtgjwdadN3KPZvYCN1qohGbKrm-ar3v92c70h1CWaW1SJhZYVb5AgbB2Xa4hJlsBoh5Li9LExf8VwQkuj-s8ezk5QDz7r6VAqm6XptLiV74WW6I0v139s0nC7/s760/ollama-gemma4-4eb-8.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="644" data-original-width="760" height="271" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjwlVkX7zuD_gyQHv-emNdjY47nx1m9Z0r77IhMktpHA3trI6Yu2VX3mLLwPN0zOPWw96PYtgjwdadN3KPZvYCN1qohGbKrm-ar3v92c70h1CWaW1SJhZYVb5AgbB2Xa4hJlsBoh5Li9LExf8VwQkuj-s8ezk5QDz7r6VAqm6XptLiV74WW6I0v139s0nC7/s320/ollama-gemma4-4eb-8.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg8wnMNUhJ1HpIGmAA8bUVc-5MEtDq3xXgtqJz6suhe-ym5EwWnQAhDHhyV7Jxj_ypKyD89-A166TdF_J33-y2CZ4WOx45kEiztTjSGvEBz5pccFWFWNj_9IJIPEdV2judO8RhObkuedOpg58gv-XAnh_si1f74nhabmW0-2-8NmRDOEO9hAwXWbHbe4i4X/s776/ollama-gemma4-4eb-9.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="662" data-original-width="776" height="273" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEg8wnMNUhJ1HpIGmAA8bUVc-5MEtDq3xXgtqJz6suhe-ym5EwWnQAhDHhyV7Jxj_ypKyD89-A166TdF_J33-y2CZ4WOx45kEiztTjSGvEBz5pccFWFWNj_9IJIPEdV2judO8RhObkuedOpg58gv-XAnh_si1f74nhabmW0-2-8NmRDOEO9hAwXWbHbe4i4X/s320/ollama-gemma4-4eb-9.jpg" width="320" /></a></div><br /><div><br /></div><span>Gemma 4 回應內容真的蠻有深度的.&nbsp;</span><div><br /></div><div>接下來上傳一張貓咪圖片 "<span style="color: #2b00fe;"><i>請描述這張圖片內容</i></span>" :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgjcEU6i4nWo0M5sQTm6NKb96_sdJlEoD40VD4YKXPjyxVk9B3gNVPEH_6G8k5BZIdxOHOE_ntDuJtEi2r30w-mJLOzSoICdLDRPwpzZxC0BsyGYlWa4an8Y0OA7JCj5cvW1vabd3JT1U4CT_if5xUdEdWyCUYS8cblKktSrtfGZpcPYz3HptWtoIRpvUyy/s320/cdo.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="181" data-original-width="320" height="181" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgjcEU6i4nWo0M5sQTm6NKb96_sdJlEoD40VD4YKXPjyxVk9B3gNVPEH_6G8k5BZIdxOHOE_ntDuJtEi2r30w-mJLOzSoICdLDRPwpzZxC0BsyGYlWa4an8Y0OA7JCj5cvW1vabd3JT1U4CT_if5xUdEdWyCUYS8cblKktSrtfGZpcPYz3HptWtoIRpvUyy/s320/cdo.jpg" width="320" /></a></div><br /><div><br /></div><div>結果成功地辨識出三隻小貓, 這比 llama3.2-vision 模型厲害 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi0xbBza-3ra1FncfoFQse8RHbih2Z2-Qr56TbDfjwodfVzs_wTG71F-JFNoRvzfuCRvBusFT_jYtG6sZoEtzo46uzYkP-Ky9QugZ_jWnvQiOSeq8XmJjmBHUlBaUm8GpOD2tVPZqYS3PYOVpOs4pNMLHBpPwQone3QXfqx3JlgTubrVfDDSEvl-AWZUNuK/s774/ollama-gemma4-4eb-10.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="657" data-original-width="774" height="272" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi0xbBza-3ra1FncfoFQse8RHbih2Z2-Qr56TbDfjwodfVzs_wTG71F-JFNoRvzfuCRvBusFT_jYtG6sZoEtzo46uzYkP-Ky9QugZ_jWnvQiOSeq8XmJjmBHUlBaUm8GpOD2tVPZqYS3PYOVpOs4pNMLHBpPwQone3QXfqx3JlgTubrVfDDSEvl-AWZUNuK/s320/ollama-gemma4-4eb-10.jpg" width="320" /></a></div><div><br /></div><div><br /></div><div>接下來考驗它能否看懂 K 線圖, 我上傳一張用 kbar 模組繪製的 K 線圖 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhfw3BH0pgR14WR5Zcpi4dn_db1HipU6EkDUyQn6_A2SWF9hip7gI8A9eZT1dAhkWrnq0CbsJLeNxKRHtdU-yG1q6uo0ffUdLe1mE8K0hBeL6iMv5HFan48gg7HP0jC25D4LRcgdgWZzPgBF8dX9DO4uu0OUTDdFn1_zJw6EKjoasIrqnt6VVVeUArUtSuk/s816/kbar-0050.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="647" data-original-width="816" height="254" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhfw3BH0pgR14WR5Zcpi4dn_db1HipU6EkDUyQn6_A2SWF9hip7gI8A9eZT1dAhkWrnq0CbsJLeNxKRHtdU-yG1q6uo0ffUdLe1mE8K0hBeL6iMv5HFan48gg7HP0jC25D4LRcgdgWZzPgBF8dX9DO4uu0OUTDdFn1_zJw6EKjoasIrqnt6VVVeUArUtSuk/s320/kbar-0050.jpg" width="320" /></a></div><br /><div><br /></div><div>結果如下 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhTg1p7jErWdrYx-w3qQmMHZh3ZZ9eAM4-m2OlWA55EgNBNlqWtkvIumPOj5_rtj0qYsDvXHR3QbWi86kcwzKocAjn0fp2pau_kCBSfuTdaw9QPjShss667a0RRBxm-k0csd1AHyQHtnyYQe4sliUprJw5GZFzbBJbgpEPf0m5u4DyLTJY0EqlV7J2lluU6/s763/ollama-gemma4-4eb-11.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="555" data-original-width="763" height="233" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhTg1p7jErWdrYx-w3qQmMHZh3ZZ9eAM4-m2OlWA55EgNBNlqWtkvIumPOj5_rtj0qYsDvXHR3QbWi86kcwzKocAjn0fp2pau_kCBSfuTdaw9QPjShss667a0RRBxm-k0csd1AHyQHtnyYQe4sliUprJw5GZFzbBJbgpEPf0m5u4DyLTJY0EqlV7J2lluU6/s320/ollama-gemma4-4eb-11.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgtmmzPVTJhA32mngxT5yTFksUa4LT5-nKSRdXGrQf8IxXldiXVIF9_JoqvU1bJVlApMgZ4VXN1nlHP_QRk3-zNRaWSUh2rSev93lBAhb8VnkcMAbftK94xDHAPQ_7HIVnFGNIqZRJT7xPPZLFDAyabAi-Rfol1EM-zZGns0c0YKpztyC0Pjv3ec8dAAKfR/s758/ollama-gemma4-4eb-12.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="660" data-original-width="758" height="279" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgtmmzPVTJhA32mngxT5yTFksUa4LT5-nKSRdXGrQf8IxXldiXVIF9_JoqvU1bJVlApMgZ4VXN1nlHP_QRk3-zNRaWSUh2rSev93lBAhb8VnkcMAbftK94xDHAPQ_7HIVnFGNIqZRJT7xPPZLFDAyabAi-Rfol1EM-zZGns0c0YKpztyC0Pjv3ec8dAAKfR/s320/ollama-gemma4-4eb-12.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjaRkW0cPvncsH7bno2SJuZGXx6AFq_E8OLr6FBMJt7uomiyszzlLFNqs0A_FIkOHCFqXdAPfMcgXOIefHvmnS1JE8dZnTOGhUr_Kfyd7IIHPY2u-6ZfutNUkSI3OuRmSfhWpCBf0JuOFaAIX7cUpdttQAYaoqIqpnGEZ6SLgye4wrZljPaypNDo1ntY5iK/s768/ollama-gemma4-4eb-13.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="605" data-original-width="768" height="252" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjaRkW0cPvncsH7bno2SJuZGXx6AFq_E8OLr6FBMJt7uomiyszzlLFNqs0A_FIkOHCFqXdAPfMcgXOIefHvmnS1JE8dZnTOGhUr_Kfyd7IIHPY2u-6ZfutNUkSI3OuRmSfhWpCBf0JuOFaAIX7cUpdttQAYaoqIqpnGEZ6SLgye4wrZljPaypNDo1ntY5iK/s320/ollama-gemma4-4eb-13.jpg" width="320" /></a></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjXFh24h68I933pGzD1k3jhw3af6tCkOk9VQHI_Y8pPpFiyHNwN6N9F9GsMsqGjJ28Atwt5y4p5Ab3D5OilAs9UoCGUazKMIimZT8EGyRjElveH8dtXK_r8NgfhuL6mmzRkxxvwE2whoQC8sxR3CewZ9ex8lVkHN9ZD8_ttxIzZEM-E_-GgiMk1UaOKhMyH/s761/ollama-gemma4-4eb-14.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="759" data-original-width="761" height="319" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjXFh24h68I933pGzD1k3jhw3af6tCkOk9VQHI_Y8pPpFiyHNwN6N9F9GsMsqGjJ28Atwt5y4p5Ab3D5OilAs9UoCGUazKMIimZT8EGyRjElveH8dtXK_r8NgfhuL6mmzRkxxvwE2whoQC8sxR3CewZ9ex8lVkHN9ZD8_ttxIzZEM-E_-GgiMk1UaOKhMyH/s320/ollama-gemma4-4eb-14.jpg" width="320" /></a></div><br /><div><br /></div><div>居然還能解盤, 真是厲害.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-31001940228133520772026-05-25T12:38:30.726+08:002026-05-25T15:10:02.676+08:00Ollama 學習筆記 : Llamma 模型測試 (llama3.2-vision:11b)<div>今天繼續用 Ollama 測試本地模型, 本篇旨在測試 Meta 釋出的第一款原生多模態大語言模型 llama3.2-vision, 也是開源 AI 發展史上的重要里程碑. 它最大的特色是讓 Llama 家族正式擺脫純文字的限制, 具備了看圖思考的能力.&nbsp;</div><div><br /></div><div>本系列全部測試文章索引參考 :</div><div><br /></div><div># <a href="https://yhhuang1966.blogspot.com/2026/05/ollama.html" target="_blank">Ollama 學習筆記索引</a>&nbsp;&nbsp;</div><div><br /></div><div>在 Ollama 生態系中預設執行的通常是 llama3.2-vision:11b (110億參數), 經過 4-bit 量化後檔案大小約為 7.8 GB, 此體積可完整塞進 RTX 3060 12GB 的 VRAM, 是個人桌機體驗本地端視覺 AI 的主力首選.&nbsp;</div><div><br /></div><div><div>llama3.2-vision 的核心強項如下 :&nbsp;</div><div><ul style="text-align: left;"><li>優秀的圖表與文檔理解 :<br />Llama 3.2-Vision 在解讀商業圖表 (如折線圖, 圓餅圖), 流程圖, 學術統計圖, 以及含有複雜排版的 PDF 文檔時表現極為強悍. 在官方基準測試中, 它在圖表和細節分析上甚至超越了部分同量級的商業閉源模型 (如 Claude 3 Haiku).</li><li>強大的光學字元辨識 (OCR) :<br />不論是印刷體還是清晰的手寫字, 它能非常精準地將圖片中的文字與表格數據抓取出來並轉換成 Markdown 格式.&nbsp;</li><li>影像定位與物件偵測 :&nbsp;<br />它不僅能看懂照片裡有什麼, 還能根據你的文字描述, 指出物件在圖片中的大概位置或方位.&nbsp;</li><li>完美契合 12GB VRAM 顯卡 :<br />7.8 GB 的模型大小, 載入後大約佔用 8~9 GB VRAM, 留下約 3GB 的緩衝空間來儲存聊天上下文, 這使得在本地端丟照片給它時, 生成和解析的速度非常流暢, 不需要耗時等待.&nbsp;</li></ul></div></div><div>但 也存在一些缺點 :</div><div><ul style="text-align: left;"><li>程式與高難度邏輯能力偏弱 :<br />由於核心技能都聚焦在視覺和基礎對話, 如果把它當作寫 Python 的大腦或拿去跑 Claude Code 這類高強度 Agent 任務, 它會因為邏輯不夠硬核或工具呼叫格式錯誤而卡死.</li><li>複雜圖表偶爾會有幻覺 :<br />雖然它很擅長看圖, 但在面對資訊量極度爆炸, 字體極小, 或是線條密密麻麻的複雜金融 K 線圖電路圖時, 它還是有一定機率會胡言亂語.&nbsp;</li><li>純文字智商相比同體積模型無優勢 :<br />如果純粹拿它來進行文字對話, 寫作或推理, 它的純文字能力基本上就是 Llama 3.1 8B 的水準, 在面對純文字高難度任務時表現會遜於同等體積的 Phi 4 (14B) 或 Qwen 3.5.&nbsp;</li></ul></div><div>總之,&nbsp;llama3.2-vision 擅長處理圖表資料, 但文字對話較弱且容易有幻覺. 另外,&nbsp;llama3.2-vision 的多模態只限於 "文字+圖片", 無法處理語音, 而且圖片需直接上傳, 給它圖片網址是無法處理的.&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjMFdJdXpOh1lPycTrqOkqdDJxiGtCv69oRQSRYh4ds9SNGFsa6WRihCtOFbbw-XaPp5KUCvdu26bIOD3sY8CA4Gd272WGxHjrrqRlfPvwPXGp_KuFpoN93wqgoJ5gyF4y7ixhP1TW3SbaH4UiEXrEBFXk2Bb0hmeBaZwyPrt42iqx9gUReMZqVvmnl6OWL/s1740/ollama-llama3.2-vision-11b-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1180" data-original-width="1740" height="217" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEjMFdJdXpOh1lPycTrqOkqdDJxiGtCv69oRQSRYh4ds9SNGFsa6WRihCtOFbbw-XaPp5KUCvdu26bIOD3sY8CA4Gd272WGxHjrrqRlfPvwPXGp_KuFpoN93wqgoJ5gyF4y7ixhP1TW3SbaH4UiEXrEBFXk2Bb0hmeBaZwyPrt42iqx9gUReMZqVvmnl6OWL/s320/ollama-llama3.2-vision-11b-1.jpg" width="320" /></a></div><div><br /></div><div><br /></div>下載執行此模型 :<br /><div><br /></div><div><div>PS C:\Users\USER&gt; <b><span style="color: #2b00fe;">ollama run llama3.2-vision:11b</span></b>&nbsp; &nbsp;&nbsp;</div><div>pulling manifest</div><div>pulling 9999d473417a: 100% ▕██████████████████████████████████████████▏ 7.8 GB</div><div>pulling 715415638c9c: 100% ▕██████████████████████████████████████████▏&nbsp; 269 B</div><div>pulling d3ed60b91757: 100% ▕██████████████████████████████████████████▏ 6.0 KB</div><div>pulling fcc5a6bec9da: 100% ▕██████████████████████████████████████████▏ 7.7 KB</div><div>pulling fefc914e46e6: 100% ▕██████████████████████████████████████████▏&nbsp; &nbsp;32 B</div><div>pulling 1506fb8a7284: 100% ▕██████████████████████████████████████████▏&nbsp; 564 B</div><div>verifying sha256 digest</div><div>writing manifest</div><div>success</div></div><div><br /></div><div>先來測試其圖片辨識能力, 我用谷歌搜尋圖片找到下面這張 NBX-3388&nbsp;車牌, 可以拿來測試看看&nbsp;llama3.2-vision 是否可辨識出號碼 :</div><div><br /></div><div>#&nbsp;<a href="https://auto.ltn.com.tw/news/13921" target="_blank">https://auto.ltn.com.tw/news/13921</a></div><div><br /></div><div>多模態測試不適合使用 CLI 介面, 應該使用 Ollama 原生介面, 選則模型 llama3.2-vision, 輸入 "請辨識這張車牌號碼", 按輸入框中的 + 鈕, 上傳或貼上車牌圖片後提交 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhkYHMZ5_6e6B-nRgc6q0r1D2N5a1VH7NHxOIt-t2FHyX61sjHIWG2JcSa8QfhxVv5taq2Cr7TiOb-G-TWl8DGYpRrCnJDSQobNro5n8U72S0_Re4ayw9oI0KlfUddAqo1bSgIJTw_eFVT1NYTJEfnc01C0xc4fVSsxg5eYukrwmtcOkdKlNYvcoHGmjKjA/s1574/ollama-llama3.2-vision-11b-3.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1011" data-original-width="1574" height="206" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhkYHMZ5_6e6B-nRgc6q0r1D2N5a1VH7NHxOIt-t2FHyX61sjHIWG2JcSa8QfhxVv5taq2Cr7TiOb-G-TWl8DGYpRrCnJDSQobNro5n8U72S0_Re4ayw9oI0KlfUddAqo1bSgIJTw_eFVT1NYTJEfnc01C0xc4fVSsxg5eYukrwmtcOkdKlNYvcoHGmjKjA/s320/ollama-llama3.2-vision-11b-3.jpg" width="320" /></a></div><br /><div><br /></div><div>結果如下, 正確辨識出號碼為 NBX-3388 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI4fizygdqXQ-5UFb9J6q3zigVZdMKk9cJ3bIGMsUuy_y9OTJsnvzsIy2QnN3OZSqHkVnT2luJbCjW4HTL1LkzK2b-fgPXZ015pL2cT9-v5vmZtpJVZA-4qPoQklMdVgnFo29SNgHpEgaSmTFGNF9ztqjHm3NHu3OJmMcTJnhDpzBL8fzdTyTZNTkYfjWj/s1567/ollama-llama3.2-vision-11b-4.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="1029" data-original-width="1567" height="210" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEhI4fizygdqXQ-5UFb9J6q3zigVZdMKk9cJ3bIGMsUuy_y9OTJsnvzsIy2QnN3OZSqHkVnT2luJbCjW4HTL1LkzK2b-fgPXZ015pL2cT9-v5vmZtpJVZA-4qPoQklMdVgnFo29SNgHpEgaSmTFGNF9ztqjHm3NHu3OJmMcTJnhDpzBL8fzdTyTZNTkYfjWj/s320/ollama-llama3.2-vision-11b-4.jpg" width="320" /></a></div><br /><div><br /></div><div>但它說這是巴西車牌顯然是幻想 + 胡扯 :&nbsp;</div><div><br /></div><div>"<i><span style="color: #2b00fe;">這張車牌號碼是來自巴西。巴西的車牌號碼由兩部分組成：區域代碼和序號。區域代碼是用來識別車牌所屬的州的三位數字，序號則是用來識別車牌的四位數字。</span></i></div><div><i><span style="color: #2b00fe;"><br /></span></i></div><div><i><span style="color: #2b00fe;">在這張車牌號碼中，區域代碼是「NBX」，序號是「3388」。因此，這張車牌號碼的完整代碼是「NBX-3388」。這張車牌號碼可能是屬於巴西的某個州，例如巴西利亞（Brasília）或聖保羅（São Paulo）等。</span></i>"</div><div><br /></div><div><br /></div><div>接下來我上傳一張 Lenna 的照片要求它描述, 參考 :</div><div><br /></div><div># <a href="https://zh.wikipedia.org/wiki/File:Lenna.jpg" target="_blank">https://zh.wikipedia.org/wiki/File:Lenna.jpg</a>&nbsp;&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgKQgAOhVAK8ChafR3vD7AocsHZCe_890DKow-uSWQKDkMZbWii0dBt89qE3loDDt6GH4clBb6ZCRT7Ewvme2LfpJw_Nwl9-zlknnplzi2vNYjxcxIAQl1Lp6ldFrXi3UQQEIRTpKKxoHl49oOllhRS3sv0Gfcg0TjOzK1uoVapIdUIxyenI8LOyQj0RSVV/s1071/ollama-llama3.2-vision-11b-5.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="994" data-original-width="1071" height="297" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgKQgAOhVAK8ChafR3vD7AocsHZCe_890DKow-uSWQKDkMZbWii0dBt89qE3loDDt6GH4clBb6ZCRT7Ewvme2LfpJw_Nwl9-zlknnplzi2vNYjxcxIAQl1Lp6ldFrXi3UQQEIRTpKKxoHl49oOllhRS3sv0Gfcg0TjOzK1uoVapIdUIxyenI8LOyQj0RSVV/s320/ollama-llama3.2-vision-11b-5.jpg" width="320" /></a></div><br /><div><br /></div><div>描述如下 :&nbsp;</div><div><br /></div><div>"<span style="color: #2b00fe;"><i>這照片是一個女性的頭像，她的頭戴著一頂白色的帽子，帽子上有兩個小的羽毛，帽子的頂部有一個小的圓形鏡子。照片中女性的頭部和肩膀被拍攝到了。女性的頭髮是黑色的，長得很長，從她的頭頂延伸到她的肩膀。她的臉部是白色的，臉部的輪廓清晰，眼睛的位置正確，鼻子和嘴巴的形狀也正確。她的頭部的輪廓是圓的，沒有任何凸凹。她的頭髮的顏色是黑色的，長得很長，從她的頭頂延伸到她的肩膀。她的臉部的輪廓是白色的，臉部的輪廓清晰，眼睛的位置正確，鼻子和嘴巴的形狀也正…</i></span>"</div><div><br /></div><div>形容人家很正呵呵.&nbsp;</div><div><br /></div><div>我在貓咪圖片網站隨機找了一張可愛的貓貓圖 :</div><div><br /></div><div>#&nbsp;<a href="https://thecatapi.com/" target="_blank">https://thecatapi.com/</a>&nbsp;&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgjcEU6i4nWo0M5sQTm6NKb96_sdJlEoD40VD4YKXPjyxVk9B3gNVPEH_6G8k5BZIdxOHOE_ntDuJtEi2r30w-mJLOzSoICdLDRPwpzZxC0BsyGYlWa4an8Y0OA7JCj5cvW1vabd3JT1U4CT_if5xUdEdWyCUYS8cblKktSrtfGZpcPYz3HptWtoIRpvUyy/s1127/cdo.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="636" data-original-width="1127" height="181" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgjcEU6i4nWo0M5sQTm6NKb96_sdJlEoD40VD4YKXPjyxVk9B3gNVPEH_6G8k5BZIdxOHOE_ntDuJtEi2r30w-mJLOzSoICdLDRPwpzZxC0BsyGYlWa4an8Y0OA7JCj5cvW1vabd3JT1U4CT_if5xUdEdWyCUYS8cblKktSrtfGZpcPYz3HptWtoIRpvUyy/s320/cdo.jpg" width="320" /></a></div><br /><div><br /></div><div><br /></div><div>先叫它描述一下看到甚麼 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgF3oJeopx49zdePeWEh8PAOM01JQMSBWYzrUw49AFmQpbdotRz2nm4gmex2mAPUiFzH8bFix3LTP7Q-X_kmH1ng5PlSL1yx0OlOD6iSRRZvaTUbSfqIxkoAS7KxHdOBn5YxL6ZdcP3UNWvJZ92YjP_AUSFPoym5XUH9Kd02WZYPqAkkQLxs1gQkuzc9QQP/s1025/ollama-llama3.2-vision-11b-6.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="980" data-original-width="1025" height="306" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgF3oJeopx49zdePeWEh8PAOM01JQMSBWYzrUw49AFmQpbdotRz2nm4gmex2mAPUiFzH8bFix3LTP7Q-X_kmH1ng5PlSL1yx0OlOD6iSRRZvaTUbSfqIxkoAS7KxHdOBn5YxL6ZdcP3UNWvJZ92YjP_AUSFPoym5XUH9Kd02WZYPqAkkQLxs1gQkuzc9QQP/s320/ollama-llama3.2-vision-11b-6.jpg" width="320" /></a></div><br /><div><br /></div><div>描述如下 :&nbsp;</div><div><br /></div><div>"<i><span style="color: #2b00fe;">這張圖片是一個小貓的照片。貓的毛色是白色的，毛髮長得很長，從他的頭頂延伸到他的肩膀。他的眼睛是黑色的，瞳孔很大，鼻子和嘴巴的形狀也正確。他的頭部的輪廓是圓的，沒有任何凸凹。他的耳朵是小的，很軟，很細，很圓。他的頭髮的顏色是黑色的，長得很長，從他的頭頂延伸到他的肩膀。他的眼睛是黑色的，瞳孔很大，鼻子和嘴巴的形狀也正…</span></i>"</div><div><br /></div><div>前半部是對的, 但後半部怎麼跟上面 Lenna 圖的描述很像啊? 也是很正.&nbsp;</div><div><br /></div><div>叫它數數看圖中有幾隻貓咪 :&nbsp;</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi58sdieLbXztzTjRGTs62wTD1NDwMwcaVLWXk4MRZRs4hLqAxJJy8AkMyW2h3ys_pJoqNji6XhtuHgS5cNtRTKtffz6KOOEiT-3uxd66OsFDGg0VY48fQGQ7J1393Vo3MDBa4_adZaAGm1iBPN9xj2Ly9tkCegNzS-bN_AQ8cIgxre2_Ya9HcWu_cBl-gp/s1024/ollama-llama3.2-vision-11b-7.jpg" imageanchor="1" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="744" data-original-width="1024" height="233" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEi58sdieLbXztzTjRGTs62wTD1NDwMwcaVLWXk4MRZRs4hLqAxJJy8AkMyW2h3ys_pJoqNji6XhtuHgS5cNtRTKtffz6KOOEiT-3uxd66OsFDGg0VY48fQGQ7J1393Vo3MDBa4_adZaAGm1iBPN9xj2Ly9tkCegNzS-bN_AQ8cIgxre2_Ya9HcWu_cBl-gp/s320/ollama-llama3.2-vision-11b-7.jpg" width="320" /></a></div><br /><div><br /></div><div>真是眼睛業障重, 明明是三隻貓, 但有可能是上面第一隻貓頭只照到下半部, 沒辨識出來所致.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0tag:blogger.com,1999:blog-4877487320781767952.post-8888032104017875512026-05-24T23:53:40.799+08:002026-05-26T19:52:32.004+08:002026 年第 20 周記事<div>週六早上有人來敲門, 兩個婦人要往山地門, 路過看到我家門口的波羅蜜結實纍纍, 問可否賣一顆給她們, 她們挑了一顆較小的, 我質疑可能尚未成熟, 但聞一下已散發香氣, 拿了柴刀砍下來, 她們掏出兩百元問說夠不夠, 我說我也不知目前行情, 但是 200 元似乎太多了, 於是只拿了 100 元, 臨去前還說下次經過要再來買, 我說要下午我爸才有在家.&nbsp;</div><div><br /></div><div>週六下午把剩下的一些酪梨都套了袋, 剩下約 20 個空袋子, 預估今年酪梨產量在 150~180 之間. 菜園那棵酪梨樹至今未開花, 都是六七年前已退休的同事鄭老大給我帶回鄉下試種的, 直到前年雞舍旁這棵樹才開始開花結果, 果實較岳父種植的要小一些, 與進口酪梨差不多大. 上週小舅請耕耘機來犁菜園, 說靠南邊那一側他不種菜了, 給我決定看看要種甚麼, 我看舅種一顆嫁接芭樂與酪梨好了.&nbsp;</div><div><br /></div><div>本周大部分時間都在已閒置兩年的 MSI 電競桌機上測試 Ollama, 終極目標是要用 Claude Code 魔改版把 API 呼叫導向到本地模型 (但也會訂閱 Claude Code 一段時間來做比較), 所以最近在忙著評估哪個模型最適合做後台模型.&nbsp;</div><div><br /></div><div>昨天上 momo 找書時突然想到兩年前花了 26000 元左右買的這台搭載 RTX3060 顯卡的桌機現在價格是多少? 不查還好, 一查快嚇死 :</div><div><br /></div><div><br /></div><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgcuvS_-HwyaDG6M8hxFNSLJAyVZqLk30M1VoSpPgUA-Uhol1rbCQQScRyOe5tYucrMzDPn5Mj7hFN6-rKqgFKzn2zAdJLFwB8wMniJFrCJu7PdZWia68yeUNkyZIGFLnyyQVIC0MwzarPAFeOrhtmrZHk50akXWKiomfqBeSsZMc-b0lvm7zundgBUu4s0/s1761/momo-msi-rtx2060-12gb-1.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="932" data-original-width="1761" height="169" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgcuvS_-HwyaDG6M8hxFNSLJAyVZqLk30M1VoSpPgUA-Uhol1rbCQQScRyOe5tYucrMzDPn5Mj7hFN6-rKqgFKzn2zAdJLFwB8wMniJFrCJu7PdZWia68yeUNkyZIGFLnyyQVIC0MwzarPAFeOrhtmrZHk50akXWKiomfqBeSsZMc-b0lvm7zundgBUu4s0/s320/momo-msi-rtx2060-12gb-1.jpg" width="320" /></a></div><br /><div class="separator" style="clear: both; text-align: center;"><a href="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgC_Ct39_YqFImq-oDJe5G-4irXi6LfdSB0tyFeBQlU-IJMhgRmh4VT-AaKZ6QXQpmrXywlcwKhXt9P7lbRG5yX5txNIrjY0GMaKqSf0Rhxs6IgJrPtI3zpL5fDy5tEkevkgAtZU_NNDs0Jv4622W8AetEajn9j5gr7FYd6Dhr9j_ZmfM5oJHEcUbuwBCHV/s1763/momo-msi-rtx2060-12gb-2.jpg" style="margin-left: 1em; margin-right: 1em;"><img border="0" data-original-height="775" data-original-width="1763" height="141" src="https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEgC_Ct39_YqFImq-oDJe5G-4irXi6LfdSB0tyFeBQlU-IJMhgRmh4VT-AaKZ6QXQpmrXywlcwKhXt9P7lbRG5yX5txNIrjY0GMaKqSf0Rhxs6IgJrPtI3zpL5fDy5tEkevkgAtZU_NNDs0Jv4622W8AetEajn9j5gr7FYd6Dhr9j_ZmfM5oJHEcUbuwBCHV/s320/momo-msi-rtx2060-12gb-2.jpg" width="320" /></a></div><br /><div><br /></div><div>天啊! 幾乎同樣配備, 現在都要 4~6 萬, 還好買得早.&nbsp;</div><div><br /></div>小狐狸事務所http://www.blogger.com/profile/09435160519044041137noreply@blogger.com0