[{"data":1,"prerenderedAt":964},["ShallowReactive",2],{"article-razer-core-x-egpu-hotplug-nixos":3,"articles-100":830,"articles-10":943},{"id":4,"title":5,"alt":6,"body":7,"description":54,"extension":814,"image":815,"meta":816,"navigation":83,"path":817,"postDate":818,"published":819,"seo":820,"stem":821,"tags":822,"updated":828,"url":828,"__hash__":829},"blog\u002Fblog\u002Frazer-core-x-egpu-hotplug-nixos.md","Getting a Razer Core X eGPU to work after login on NixOS, without timing the boot","Abstract image representing a broken connection",{"type":8,"value":9,"toc":804},"minimark",[10,14,17,22,38,41,45,48,173,179,187,191,194,199,206,226,239,355,365,394,398,408,423,453,459,466,610,615,625,629,640,651,654,723,730,765,771,775,778,800],[11,12,13],"p",{},"My eGPU never got detected if I was already logged into NixOS. To get it working, I had to time switching on the Razer Core X against switching on the laptop. Sometimes that worked first try. Sometimes it took a few reboots and more timing games, and I never found a pattern.",[11,15,16],{},"\"Not detected\" was the wrong diagnosis. The system saw the GPU, but the NVIDIA driver refused to use it, and after I fixed that, nothing in userspace could reach the card. The fixes were a kernel parameter and a device node. Now I can boot first and switch the enclosure on while logged in.",[18,19,21],"h2",{"id":20},"my-setup","My setup",[23,24,25,29,32,35],"ul",{},[26,27,28],"li",{},"Laptop: ThinkPad X1 Extreme Gen 1 (20MGS0EQ00), BIOS 1.47, with an internal GTX 1050 Ti Mobile and the Intel iGPU",[26,30,31],{},"eGPU: Razer Core X Chroma (Thunderbolt 3) with an RTX 3060 Ti",[26,33,34],{},"A Dell Thunderbolt 4 dock is also on the Thunderbolt chain",[26,36,37],{},"NixOS 26.11 (unstable-ish), kernel 6.18.63, NVIDIA proprietary driver 580.196.07, COSMIC on Wayland",[11,39,40],{},"The snippets are NixOS, but the ideas apply to any distro.",[18,42,44],{"id":43},"checking-each-layer-the-system-saw-the-gpu","Checking each layer: the system saw the GPU",[11,46,47],{},"When a hotplugged GPU doesn't appear, check four layers in order. I should have done this instead of rebooting repeatedly.",[49,50,55],"pre",{"className":51,"code":52,"language":53,"meta":54,"style":54},"language-bash shiki shiki-themes github-light github-dark","# 1. Did Thunderbolt authorize the enclosure?\nboltctl list\n\n# 2. Did the PCIe device enumerate? (mine was 09:00.0)\nlspci | grep -i nvidia\n\n# 3. Did the driver bind? Look for NVRM errors and bridge window failures\njournalctl -k | grep -E 'NVRM|bridge window'\n\n# 4. Does userspace have a device node, and does nvidia-smi see the card?\nls \u002Fdev\u002Fnvidia*\nnvidia-smi -L\n","bash","",[56,57,58,67,78,85,91,111,116,122,141,146,152,164],"code",{"__ignoreMap":54},[59,60,63],"span",{"class":61,"line":62},"line",1,[59,64,66],{"class":65},"sJ8bj","# 1. Did Thunderbolt authorize the enclosure?\n",[59,68,70,74],{"class":61,"line":69},2,[59,71,73],{"class":72},"sScJk","boltctl",[59,75,77],{"class":76},"sZZnC"," list\n",[59,79,81],{"class":61,"line":80},3,[59,82,84],{"emptyLinePlaceholder":83},true,"\n",[59,86,88],{"class":61,"line":87},4,[59,89,90],{"class":65},"# 2. Did the PCIe device enumerate? (mine was 09:00.0)\n",[59,92,94,97,101,104,108],{"class":61,"line":93},5,[59,95,96],{"class":72},"lspci",[59,98,100],{"class":99},"szBVR"," |",[59,102,103],{"class":72}," grep",[59,105,107],{"class":106},"sj4cs"," -i",[59,109,110],{"class":76}," nvidia\n",[59,112,114],{"class":61,"line":113},6,[59,115,84],{"emptyLinePlaceholder":83},[59,117,119],{"class":61,"line":118},7,[59,120,121],{"class":65},"# 3. Did the driver bind? Look for NVRM errors and bridge window failures\n",[59,123,125,128,131,133,135,138],{"class":61,"line":124},8,[59,126,127],{"class":72},"journalctl",[59,129,130],{"class":106}," -k",[59,132,100],{"class":99},[59,134,103],{"class":72},[59,136,137],{"class":106}," -E",[59,139,140],{"class":76}," 'NVRM|bridge window'\n",[59,142,144],{"class":61,"line":143},9,[59,145,84],{"emptyLinePlaceholder":83},[59,147,149],{"class":61,"line":148},10,[59,150,151],{"class":65},"# 4. Does userspace have a device node, and does nvidia-smi see the card?\n",[59,153,155,158,161],{"class":61,"line":154},11,[59,156,157],{"class":72},"ls",[59,159,160],{"class":76}," \u002Fdev\u002Fnvidia",[59,162,163],{"class":106},"*\n",[59,165,167,170],{"class":61,"line":166},12,[59,168,169],{"class":72},"nvidia-smi",[59,171,172],{"class":106}," -L\n",[11,174,175,176,178],{},"On my machine, steps 1 and 2 passed (bolt had authorized the enclosure, and ",[56,177,96],{}," listed the GPU). Step 3 failed:",[49,180,185],{"className":181,"code":183,"language":184,"meta":54},[182],"language-text","NVRM: This PCI I\u002FO region assigned to your NVIDIA device is invalid\nnvidia 0000:09:00.0: probe with driver nvidia failed with error -1\nbridge window [mem size 0x100000000]: can't assign; no space\n","text",[56,186,183],{"__ignoreMap":54},[18,188,190],{"id":189},"what-was-blocking-the-card","What was blocking the card",[11,192,193],{},"Checks 3 and 4 failed for different reasons. The first problem stops the driver from binding. The second hides the card from userspace once it does.",[195,196,198],"h3",{"id":197},"bug-1-hpmemsize4g-left-bar0-unassigned","Bug 1: hpmemsize=4G left BAR0 unassigned",[11,200,201,202,205],{},"My config had ",[56,203,204],{},"pci=hpmemsize=4G,hpiosize=0"," from an earlier attempt. I assumed a big GPU needs a big window.",[11,207,208,209,212,213,217,218,221,222,225],{},"It doesn't, because ",[56,210,211],{},"hpmemsize"," is the wrong window. It reserves the 32-bit ",[214,215,216],"strong",{},"non-prefetchable"," memory window for each hotplug bridge, and the kernel can't fit 4 gigabytes into 32-bit address space. It logged ",[56,219,220],{},"can't assign; no space"," and moved on. That left the GPU's 16 MB BAR0, which lives in the same window, without an address: the device's ",[56,223,224],{},"resource"," file in sysfs was all zeros, even though the 64-bit prefetchable BARs (256 MB and 32 MB) were fine.",[11,227,228,229,232,233,235,236,238],{},"The parameter for big GPU BARs is ",[56,230,231],{},"hpmmioprefsize",", which sizes the 64-bit prefetchable window. Use a small ",[56,234,211],{}," and a large ",[56,237,231],{},":",[49,240,244],{"className":241,"code":242,"language":243,"meta":54,"style":54},"language-nix shiki shiki-themes github-light github-dark","boot.kernelParams = lib.mkBefore [\n  \"pciehp.pciehp_force=1\"\n  \"pci=realloc\"\n  \"pcie_ports=native\"\n\n  # hpmemsize: 32-bit non-prefetchable window, keep it small.\n  # hpmmioprefsize: 64-bit prefetchable window for the GPU's big BARs.\n  \"pci=hpbussize=0x20,hpmemsize=64M,hpmmioprefsize=1G,hpiosize=0\"\n  \"iommu=pt\"\n];\nboot.extraModprobeConfig = lib.mkAfter ''\n  options pciehp pciehp_force=1\n'';\n","nix",[56,245,246,274,279,284,289,293,298,303,308,313,321,342,347],{"__ignoreMap":54},[59,247,248,252,255,258,262,265,267,270],{"class":61,"line":62},[59,249,251],{"class":250},"s4XuR","boot",[59,253,254],{"class":99},".",[59,256,257],{"class":250},"kernelParams",[59,259,261],{"class":260},"s7hpK"," =",[59,263,264],{"class":250}," lib",[59,266,254],{"class":99},[59,268,269],{"class":250},"mkBefore",[59,271,273],{"class":272},"sVt8B"," [\n",[59,275,276],{"class":61,"line":69},[59,277,278],{"class":76},"  \"pciehp.pciehp_force=1\"\n",[59,280,281],{"class":61,"line":80},[59,282,283],{"class":76},"  \"pci=realloc\"\n",[59,285,286],{"class":61,"line":87},[59,287,288],{"class":76},"  \"pcie_ports=native\"\n",[59,290,291],{"class":61,"line":93},[59,292,84],{"emptyLinePlaceholder":83},[59,294,295],{"class":61,"line":113},[59,296,297],{"class":65},"  # hpmemsize: 32-bit non-prefetchable window, keep it small.\n",[59,299,300],{"class":61,"line":118},[59,301,302],{"class":65},"  # hpmmioprefsize: 64-bit prefetchable window for the GPU's big BARs.\n",[59,304,305],{"class":61,"line":124},[59,306,307],{"class":76},"  \"pci=hpbussize=0x20,hpmemsize=64M,hpmmioprefsize=1G,hpiosize=0\"\n",[59,309,310],{"class":61,"line":143},[59,311,312],{"class":76},"  \"iommu=pt\"\n",[59,314,315,318],{"class":61,"line":148},[59,316,317],{"class":272},"]",[59,319,320],{"class":260},";\n",[59,322,323,325,327,330,332,334,336,339],{"class":61,"line":154},[59,324,251],{"class":250},[59,326,254],{"class":99},[59,328,329],{"class":250},"extraModprobeConfig",[59,331,261],{"class":260},[59,333,264],{"class":250},[59,335,254],{"class":99},[59,337,338],{"class":250},"mkAfter",[59,340,341],{"class":76}," ''\n",[59,343,344],{"class":61,"line":166},[59,345,346],{"class":76},"  options pciehp pciehp_force=1\n",[59,348,350,353],{"class":61,"line":349},13,[59,351,352],{"class":76},"''",[59,354,320],{"class":260},[11,356,357,358,361,362,254],{},"To check the result without reading dmesg, list the regions for your device (replace ",[56,359,360],{},"09:00.0"," with yours). Every region should have an address, not ",[56,363,364],{},"\u003Cunassigned>",[49,366,368],{"className":51,"code":367,"language":53,"meta":54,"style":54},"sudo lspci -vv -s 09:00.0 | grep Region\n",[56,369,370],{"__ignoreMap":54},[59,371,372,375,378,381,384,387,389,391],{"class":61,"line":62},[59,373,374],{"class":72},"sudo",[59,376,377],{"class":76}," lspci",[59,379,380],{"class":106}," -vv",[59,382,383],{"class":106}," -s",[59,385,386],{"class":76}," 09:00.0",[59,388,100],{"class":99},[59,390,103],{"class":72},[59,392,393],{"class":76}," Region\n",[195,395,397],{"id":396},"bug-2-nvidia-smi-ignored-the-card-because-devnvidia1-didnt-exist","Bug 2: nvidia-smi ignored the card because \u002Fdev\u002Fnvidia1 didn't exist",[11,399,400,401,404,405,407],{},"After the window fix, the driver bound and ",[56,402,403],{},"\u002Fproc\u002Fdriver\u002Fnvidia\u002Fgpus\u002F"," listed the 3060 Ti. But ",[56,406,169],{}," still showed only the internal 1050 Ti.",[11,409,410,411,414,415,418,419,422],{},"This took me longest. ",[56,412,413],{},"ls \u002Fdev\u002Fnvidia*"," showed ",[56,416,417],{},"nvidia0"," and nothing for the second card. The kernel module doesn't create device nodes itself. On a normal distro ",[56,420,421],{},"nvidia-modprobe"," creates them on demand, but it isn't available in my NixOS setup, so a GPU that appears after boot never gets a node. To confirm, I made one by hand:",[49,424,426],{"className":51,"code":425,"language":53,"meta":54,"style":54},"sudo mknod -m 666 \u002Fdev\u002Fnvidia1 c 195 1\n",[56,427,428],{"__ignoreMap":54},[59,429,430,432,435,438,441,444,447,450],{"class":61,"line":62},[59,431,374],{"class":72},[59,433,434],{"class":76}," mknod",[59,436,437],{"class":106}," -m",[59,439,440],{"class":106}," 666",[59,442,443],{"class":76}," \u002Fdev\u002Fnvidia1",[59,445,446],{"class":76}," c",[59,448,449],{"class":106}," 195",[59,451,452],{"class":106}," 1\n",[11,454,455,456,458],{},"Both GPUs showed up in ",[56,457,169],{}," immediately.",[11,460,461,462,465],{},"To automate that, a udev rule runs a script whenever the NVIDIA driver binds to a PCI device. The script reads each minor number from ",[56,463,464],{},"\u002Fproc\u002Fdriver\u002Fnvidia\u002Fgpus\u002F*\u002Finformation"," and creates any missing node:",[49,467,469],{"className":241,"code":468,"language":243,"meta":54,"style":54},"services.udev.extraRules = let\n  nvidiaNodes = pkgs.writeShellScript \"nvidia-create-nodes\" ''\n    for info in \u002Fproc\u002Fdriver\u002Fnvidia\u002Fgpus\u002F*\u002Finformation; do\n      [ -e \"$info\" ] || continue\n      minor=$(${pkgs.gawk}\u002Fbin\u002Fawk '\u002FDevice Minor\u002F {print $NF}' \"$info\")\n      [ -n \"$minor\" ] && [ ! -e \"\u002Fdev\u002Fnvidia$minor\" ] &&\n        ${pkgs.coreutils}\u002Fbin\u002Fmknod -m 666 \"\u002Fdev\u002Fnvidia$minor\" c 195 \"$minor\"\n    done\n    exit 0\n  '';\nin ''\n  ACTION==\"bind\", SUBSYSTEM==\"pci\", DRIVER==\"nvidia\", RUN+=\"${nvidiaNodes}\"\n'';\n",[56,470,471,491,511,516,521,543,548,565,570,575,582,589,604],{"__ignoreMap":54},[59,472,473,476,478,481,483,486,488],{"class":61,"line":62},[59,474,475],{"class":250},"services",[59,477,254],{"class":99},[59,479,480],{"class":250},"udev",[59,482,254],{"class":99},[59,484,485],{"class":250},"extraRules",[59,487,261],{"class":260},[59,489,490],{"class":260}," let\n",[59,492,493,496,498,501,503,506,509],{"class":61,"line":69},[59,494,495],{"class":250},"  nvidiaNodes",[59,497,261],{"class":260},[59,499,500],{"class":250}," pkgs",[59,502,254],{"class":99},[59,504,505],{"class":250},"writeShellScript",[59,507,508],{"class":76}," \"nvidia-create-nodes\"",[59,510,341],{"class":76},[59,512,513],{"class":61,"line":80},[59,514,515],{"class":76},"    for info in \u002Fproc\u002Fdriver\u002Fnvidia\u002Fgpus\u002F*\u002Finformation; do\n",[59,517,518],{"class":61,"line":87},[59,519,520],{"class":76},"      [ -e \"$info\" ] || continue\n",[59,522,523,526,529,532,534,537,540],{"class":61,"line":93},[59,524,525],{"class":76},"      minor=$(",[59,527,528],{"class":76},"${",[59,530,531],{"class":106},"pkgs",[59,533,254],{"class":99},[59,535,536],{"class":106},"gawk",[59,538,539],{"class":76},"}",[59,541,542],{"class":76},"\u002Fbin\u002Fawk '\u002FDevice Minor\u002F {print $NF}' \"$info\")\n",[59,544,545],{"class":61,"line":113},[59,546,547],{"class":76},"      [ -n \"$minor\" ] && [ ! -e \"\u002Fdev\u002Fnvidia$minor\" ] &&\n",[59,549,550,553,555,557,560,562],{"class":61,"line":118},[59,551,552],{"class":76},"        ${",[59,554,531],{"class":106},[59,556,254],{"class":99},[59,558,559],{"class":106},"coreutils",[59,561,539],{"class":76},[59,563,564],{"class":76},"\u002Fbin\u002Fmknod -m 666 \"\u002Fdev\u002Fnvidia$minor\" c 195 \"$minor\"\n",[59,566,567],{"class":61,"line":124},[59,568,569],{"class":76},"    done\n",[59,571,572],{"class":61,"line":143},[59,573,574],{"class":76},"    exit 0\n",[59,576,577,580],{"class":61,"line":148},[59,578,579],{"class":76},"  ''",[59,581,320],{"class":260},[59,583,584,587],{"class":61,"line":154},[59,585,586],{"class":260},"in",[59,588,341],{"class":76},[59,590,591,594,596,599,601],{"class":61,"line":166},[59,592,593],{"class":76},"  ACTION==\"bind\", SUBSYSTEM==\"pci\", DRIVER==\"nvidia\", RUN+=\"",[59,595,528],{"class":76},[59,597,598],{"class":106},"nvidiaNodes",[59,600,539],{"class":76},[59,602,603],{"class":76},"\"\n",[59,605,606,608],{"class":61,"line":349},[59,607,352],{"class":76},[59,609,320],{"class":260},[11,611,612,613,254],{},"After a rebuild and reboot, I cold booted, switched the enclosure on while logged in, and the eGPU appeared in ",[56,614,169],{},[11,616,617,618,621,622,254],{},"One caveat: I had created ",[56,619,620],{},"\u002Fdev\u002Fnvidia1"," by hand earlier and don't know whether I deleted it before testing. So I can't say the rule alone creates the node. It should, but I haven't proven it from a clean ",[56,623,624],{},"\u002Fdev",[18,626,628],{"id":627},"running-programs-on-the-egpu-with-a-wrapper-script","Running programs on the eGPU with a wrapper script",[11,630,631,632,635,636,639],{},"With both cards visible, I wanted to run one program on the eGPU. My ",[56,633,634],{},"egpu-prefer"," wrapper sets environment variables and then ",[56,637,638],{},"exec","s the command.",[11,641,642,643,646,647,650],{},"It failed at first. My session sets ",[56,644,645],{},"VK_LOADER_DRIVERS_SELECT=*intel*"," and ",[56,648,649],{},"DRI_PRIME=pci-0000_00_02_0",", which together hide the NVIDIA Vulkan driver. I never found where those variables come from, and they aren't in my dotfiles.",[11,652,653],{},"Native NVIDIA GLX offload also failed: it picked the 1050 Ti even when I gave it the eGPU's UUID. So the wrapper selects the card through Vulkan instead:",[49,655,657],{"className":51,"code":656,"language":53,"meta":54,"style":54},"# Override the session's restrictions\nexport VK_LOADER_DRIVERS_SELECT=\"*nvidia*\"\nunset DRI_PRIME\n\n# Pick the 3060 Ti by PCI vendor:device ID\nexport MESA_VK_DEVICE_SELECT=\"10de:2489\"\n\nexec \"$@\"\n",[56,658,659,664,678,686,690,695,707,711],{"__ignoreMap":54},[59,660,661],{"class":61,"line":62},[59,662,663],{"class":65},"# Override the session's restrictions\n",[59,665,666,669,672,675],{"class":61,"line":69},[59,667,668],{"class":99},"export",[59,670,671],{"class":272}," VK_LOADER_DRIVERS_SELECT",[59,673,674],{"class":99},"=",[59,676,677],{"class":76},"\"*nvidia*\"\n",[59,679,680,683],{"class":61,"line":80},[59,681,682],{"class":106},"unset",[59,684,685],{"class":76}," DRI_PRIME\n",[59,687,688],{"class":61,"line":87},[59,689,84],{"emptyLinePlaceholder":83},[59,691,692],{"class":61,"line":93},[59,693,694],{"class":65},"# Pick the 3060 Ti by PCI vendor:device ID\n",[59,696,697,699,702,704],{"class":61,"line":113},[59,698,668],{"class":99},[59,700,701],{"class":272}," MESA_VK_DEVICE_SELECT",[59,703,674],{"class":99},[59,705,706],{"class":76},"\"10de:2489\"\n",[59,708,709],{"class":61,"line":118},[59,710,84],{"emptyLinePlaceholder":83},[59,712,713,715,718,721],{"class":61,"line":124},[59,714,638],{"class":106},[59,716,717],{"class":76}," \"",[59,719,720],{"class":106},"$@",[59,722,603],{"class":76},[11,724,725,726,729],{},"This is a trimmed version. The full wrapper also sets ",[56,727,728],{},"VK_DRIVER_FILES"," to the NVIDIA ICD path and a few CUDA variables. To check the selection:",[49,731,733],{"className":51,"code":732,"language":53,"meta":54,"style":54},"MESA_VK_DEVICE_SELECT=10de:2489 VK_LOADER_DRIVERS_SELECT='*nvidia*' vulkaninfo --summary | grep deviceName\n",[56,734,735],{"__ignoreMap":54},[59,736,737,740,742,745,747,749,752,755,758,760,762],{"class":61,"line":62},[59,738,739],{"class":272},"MESA_VK_DEVICE_SELECT",[59,741,674],{"class":99},[59,743,744],{"class":76},"10de:2489",[59,746,671],{"class":272},[59,748,674],{"class":99},[59,750,751],{"class":76},"'*nvidia*'",[59,753,754],{"class":72}," vulkaninfo",[59,756,757],{"class":106}," --summary",[59,759,100],{"class":99},[59,761,103],{"class":72},[59,763,764],{"class":76}," deviceName\n",[11,766,767,768,770],{},"Vulkan lists the RTX 3060 Ti first, CUDA sees both GPUs, and rendering through ",[56,769,634],{}," works.",[18,772,774],{"id":773},"what-i-havent-tested","What I haven't tested",[11,776,777],{},"I'd rather list these than imply it all works:",[23,779,780,786,794],{},[26,781,782,785],{},[214,783,784],{},"Games."," Rendering through the wrapper works, but I haven't run a game on the eGPU yet.",[26,787,788,791,792,254],{},[214,789,790],{},"The udev rule alone",", without a hand-made node already in ",[56,793,624],{},[26,795,796,799],{},[214,797,798],{},"Safe unplug."," I haven't tried hot-unplugging while the driver is in use. I'd stop everything using the card first.",[801,802,803],"style",{},"html pre.shiki code .sJ8bj, html code.shiki .sJ8bj{--shiki-default:#6A737D;--shiki-dark:#6A737D}html pre.shiki code .sScJk, html code.shiki .sScJk{--shiki-default:#6F42C1;--shiki-dark:#B392F0}html pre.shiki code .sZZnC, html code.shiki .sZZnC{--shiki-default:#032F62;--shiki-dark:#9ECBFF}html pre.shiki code .szBVR, html code.shiki .szBVR{--shiki-default:#D73A49;--shiki-dark:#F97583}html pre.shiki code .sj4cs, html code.shiki .sj4cs{--shiki-default:#005CC5;--shiki-dark:#79B8FF}html .default .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .shiki span {color: var(--shiki-default);background: var(--shiki-default-bg);font-style: var(--shiki-default-font-style);font-weight: var(--shiki-default-font-weight);text-decoration: var(--shiki-default-text-decoration);}html .dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html.dark .shiki span {color: var(--shiki-dark);background: var(--shiki-dark-bg);font-style: var(--shiki-dark-font-style);font-weight: var(--shiki-dark-font-weight);text-decoration: var(--shiki-dark-text-decoration);}html pre.shiki code .s4XuR, html code.shiki .s4XuR{--shiki-default:#E36209;--shiki-dark:#FFAB70}html pre.shiki code .s7hpK, html code.shiki .s7hpK{--shiki-default:#B31D28;--shiki-default-font-style:italic;--shiki-dark:#FDAEB7;--shiki-dark-font-style:italic}html pre.shiki code .sVt8B, html code.shiki .sVt8B{--shiki-default:#24292E;--shiki-dark:#E1E4E8}",{"title":54,"searchDepth":69,"depth":69,"links":805},[806,807,808,812,813],{"id":20,"depth":69,"text":21},{"id":43,"depth":69,"text":44},{"id":189,"depth":69,"text":190,"children":809},[810,811],{"id":197,"depth":80,"text":198},{"id":396,"depth":80,"text":397},{"id":627,"depth":69,"text":628},{"id":773,"depth":69,"text":774},"md","\u002Fimages\u002Fegpu-nixos-laptop-enclosure.png",{},"\u002Fblog\u002Frazer-core-x-egpu-hotplug-nixos","2026-11-18",false,{"title":5,"description":54},"blog\u002Frazer-core-x-egpu-hotplug-nixos",[823,824,825,826,827],"nixos","egpu","thunderbolt","nvidia","linux",null,"a428IOS6dXRhzz_pTUJDyjM8tYE269SY2CttuILLWvI",[831,833,845,855,861,869,878,887,896,902,908,917,924,930,937],{"title":5,"description":54,"tags":832,"postDate":818,"path":817,"url":828},[823,824,825,826,827],{"title":834,"description":835,"tags":836,"postDate":843,"path":844,"url":828},"Stop stashing half-finished work: use Git worktrees and Worktrunk","Git worktrees keep separate branches checked out at once. Worktrunk automates per-worktree setup. The same isolation also works for parallel AI agents.",[837,838,839,840,841,842],"git","worktrees","worktrunk","developer productivity","command line","workflow","2026-08-16","\u002Fblog\u002Fworktrees-with-worktrunk",{"title":846,"description":847,"tags":848,"postDate":853,"path":854,"url":828},"Self-Hosting my Git forge with Forgejo","Why I moved away from GitHub and chose a lightweight, community-driven alternative",[849,850,851,852],"Homelab","Open source","Self-hosting","Workflow","2026-04-29","\u002Fblog\u002Fself-hosting-my-git-forge",{"title":856,"description":857,"tags":858,"postDate":859,"path":860,"url":828},"Curating my own reading and discovery feed with RSS","Taking concrete steps away from algorithms using FreshRSS, self-hosted tools, and a pipeline built for me.",[849,850,851,852],"2026-02-08","\u002Fblog\u002Freading-and-discovery-feed-curation",{"title":862,"description":863,"tags":864,"postDate":867,"path":868,"url":828},"Fixing my broken Docker Immich+Traefik setup","Photo indexing failed in my Immich+Traefik setup. I'd misconfigured my Docker networking. Here's how I figured it out and what I changed.",[865,866],"homelab","networking","2025-10-31","\u002Fblog\u002Ffixing-docker-immich-traefik-setup",{"title":870,"description":871,"tags":872,"postDate":876,"path":877,"url":828},"Quickly outline large JSON responses with a custom jq function","Use jq to automatically map and visualize the structure of large JSON responses, perfect for exploring undocumented endpoints.",[873,874,841,875,840],"jq","json","automation","2025-10-19","\u002Fblog\u002Fjq-json-outline",{"title":879,"description":880,"tags":881,"postDate":885,"path":886,"url":828},"Why I added RSS feeds to my blog (and how I did it in Nuxt 3)","How I built RSS and Atom feed support for my Nuxt 3 website to let readers follow my content without social media algorithms.",[882,883,884],"Nuxt","web development","tutorial","2025-08-31","\u002Fblog\u002Fsetting-up-rss-atom-feeds-nuxt3",{"title":888,"description":889,"tags":890,"postDate":894,"path":895,"url":828},"Revisiting my Stream Deck for music control","How I turned my neglected Stream Deck into a useful tool for controlling YouTube Music without breaking my workflow.",[891,892,875,893],"productivity","Developer Productivity","home-office","2025-06-07","\u002Fblog\u002Fstream-deck-youtube-music",{"title":897,"description":898,"tags":899,"postDate":900,"path":901,"url":828},"fzf + fabric: Smart selection for LLM patterns and models","Enhance your experience with fabric, the AI prompt tool, by using interactive selection for patterns and models.",[840],"2025-05-24","\u002Fblog\u002Fai-pattern-selection-in-fabric",{"title":903,"description":904,"tags":905,"postDate":906,"path":907,"url":828},"Effortless configuration syncing with GNU Stow and Git: a software engineer's workflow","Use GNU Stow and Git to keep config files in sync across Linux and macOS. Keep them in one repo with full history, and set up new machines in minutes.",[840],"2025-01-12","\u002Fblog\u002Feffortless-configuration-syncing-with-gnu-stow",{"title":909,"description":910,"tags":911,"postDate":915,"path":916,"url":828},"My 2023 travels: A visual journey","Looking back at the best pictures from my travel adventures in 2023",[912,913,914],"personal","travel","photography","2024-01-01","\u002Fblog\u002F2023-travels-visual-journey",{"title":918,"description":919,"tags":920,"postDate":922,"path":923,"url":828},"Re-living memories through proper media organization","Organizing media such as photos and videos can bring back cherished memories and help relive them. Here's my approach to organizing my digital media collection.",[912,921],"technology","2023-02-06","\u002Fblog\u002Freliving-memories-through-proper-media-organization",{"title":925,"description":926,"tags":927,"postDate":928,"path":929,"url":828},"Looking back: My 2022 in review","From new adventures and memories to lessons learned and personal growth, this year has been a year of discovery. Here's my 2022 in review.",[912,913],"2022-12-31","\u002Fblog\u002F2022-in-review",{"title":931,"description":932,"tags":933,"postDate":935,"path":936,"url":828},"Git and the command-line: Take your workflow to the next level","Some tips on how to improve your Git productivity using the command-line",[934,840],"programming","2022-12-18","\u002Fblog\u002Fgit-and-command-line-next-level-workflow",{"title":938,"description":939,"tags":940,"postDate":941,"path":942,"url":828},"Hiking Tsho-Rolpa lake trail in Nepal","Details of my hiking journey to the Tsho-Rolpa glacial lake in Nepal 🇳🇵, situated at 4,580m from sea level",[913],"2022-01-22","\u002Fblog\u002Fhiking-tsho-rolpa-lake-trail-in-nepal",[944,946,948,950,952,954,956,958,960,962],{"title":5,"description":54,"tags":945,"postDate":818,"path":817,"url":828},[823,824,825,826,827],{"title":834,"description":835,"tags":947,"postDate":843,"path":844,"url":828},[837,838,839,840,841,842],{"title":846,"description":847,"tags":949,"postDate":853,"path":854,"url":828},[849,850,851,852],{"title":856,"description":857,"tags":951,"postDate":859,"path":860,"url":828},[849,850,851,852],{"title":862,"description":863,"tags":953,"postDate":867,"path":868,"url":828},[865,866],{"title":870,"description":871,"tags":955,"postDate":876,"path":877,"url":828},[873,874,841,875,840],{"title":879,"description":880,"tags":957,"postDate":885,"path":886,"url":828},[882,883,884],{"title":888,"description":889,"tags":959,"postDate":894,"path":895,"url":828},[891,892,875,893],{"title":897,"description":898,"tags":961,"postDate":900,"path":901,"url":828},[840],{"title":903,"description":904,"tags":963,"postDate":906,"path":907,"url":828},[840],1795024279486]