You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
AnythingLLM turns documents, websites and other content into a chat-ready knowledge base with RAG, agents and multi-user workspaces. Connects to local runners (Ollama, LM Studio) as well as hosted providers.
Follows upstream BARE_METAL.md exactly:yarn setup, frontend/.env with VITE_API_BASE='/api', yarn build in frontend, cp -R frontend/dist server/public, npx prisma generate + npx prisma migrate deploy, then server and collector as separate processes.
Note: upstream explicitly does not support non-Docker deployments — BARE_METAL.md is a reference only and such issues are not answered. Called out as a warning in the JSON so nobody reports LXC problems upstream.
Two systemd units:anythingllm (server + UI on :3001) and anythingllm-collector (document processing).
Persistence:STORAGE_DIR=/opt/anythingllm_data/storage (LanceDB vectors, SQLite, documents) — outside the app dir. JWT_SECRET, SIG_KEY and SIG_SALT are generated at install; the update path backs up and restores all three .env files around CLEAN_INSTALL.
No models run here — point it at a separate Ollama container or a hosted API.
The curl Output is clear in this case. Its an Network recv issue. The File is downloadable and round about 36mb big
The original AnythingLLM source archive now downloads and deploys successfully. I also tested the exact archive URL from both my Mac and the Proxmox host on the same connection; the complete 35.2 MB archive downloaded in about two seconds in both cases.
On the second installation attempt, the script proceeded to yarn setup but failed while sharp 0.32.6 was trying to download:
I also noticed that the dev installer uses Node.js 22:
NODE_VERSION="22" NODE_MODULE="yarn" setup_nodejs
However, AnythingLLM v1.16.0 explicitly lists Node.js 18 in BARE_METAL.md, and its official Dockerfile also uses Node.js 18. The dev script therefore currently differs from the upstream bare-metal requirements.
Would it make sense to test or patch this as follows?
I cannot yet confirm whether Node.js 22 is the direct cause of the sharp failure, but using the upstream-specified Node.js version would remove one compatibility variable.
The reported IPv4 Internet Not Connected message appears to be based on failed ICMP probes to public DNS servers. It does not by itself prove that IPv4 HTTP connectivity is unavailable. Unfortunately, I removed the failed container; on the next test I will keep it for direct IPv4/IPv6 diagnostics if the download fails again.
Node18 is since many many months EOL and a Security risk
You are right, Node.js 18 is EOL and should not be used for a new installation. The upstream BARE_METAL.md requirement is therefore outdated, and keeping Node.js 22 makes sense.
I have now completed the installation successfully with Node.js 22. The remaining issue was specifically the sharp libvips download.
I downloaded the same asset successfully with curl on the Proxmox host and copied it into the container at:
After that, yarn setup detected the cached archive, completed successfully, and AnythingLLM is now running together with the collector.
So Node.js 22 was not the cause. The reproducible failure was limited to the libvips download performed during the sharp 0.32.6 installation. A possible installer-side workaround would be to prefetch this asset using the existing curl download logic before running yarn setup, although the hardcoded sharp/libvips version would need to be handled carefully.
The application is ignoring the STORAGE_DIR flag in /opt/anythingllm/server/.env
During update, from 1.16.0 to 1.16.1, my workspace gets wiped as everything is saved in /opt/anythingllm/server/storage/ instead of /opt/anythingllm_data/storage.
With gemma4:12b running on Ollama, I get only the Thoughts status, but no answer.
I tried the same thing with the desktop app from AnythingLLM and I got the answer no problem.
Activity
Julian612 commentedon Aug 26, 2026
The original AnythingLLM source archive now downloads and deploys successfully. I also tested the exact archive URL from both my Mac and the Proxmox host on the same connection; the complete 35.2 MB archive downloaded in about two seconds in both cases.
On the second installation attempt, the script proceeded to yarn setup but failed while sharp 0.32.6 was trying to download:
https://github.com/lovell/sharp-libvips/releases/download/v8.14.5/libvips-8.14.5-linux-x64.tar.br
I also noticed that the dev installer uses Node.js 22:
NODE_VERSION="22" NODE_MODULE="yarn" setup_nodejs
However, AnythingLLM v1.16.0 explicitly lists Node.js 18 in BARE_METAL.md, and its official Dockerfile also uses Node.js 18. The dev script therefore currently differs from the upstream bare-metal requirements.
Would it make sense to test or patch this as follows?
-NODE_VERSION="22" NODE_MODULE="yarn" setup_nodejs
+NODE_VERSION="18" NODE_MODULE="yarn" setup_nodejs
I cannot yet confirm whether Node.js 22 is the direct cause of the sharp failure, but using the upstream-specified Node.js version would remove one compatibility variable.
The reported IPv4 Internet Not Connected message appears to be based on failed ICMP probes to public DNS servers. It does not by itself prove that IPv4 HTTP connectivity is unavailable. Unfortunately, I removed the failed container; on the next test I will keep it for direct IPv4/IPv6 diagnostics if the download fails again.
MickLesk commentedon Aug 26, 2026
Node18 is since many many months EOL and a Security risk
Julian612 commentedon Aug 26, 2026
You are right, Node.js 18 is EOL and should not be used for a new installation. The upstream BARE_METAL.md requirement is therefore outdated, and keeping Node.js 22 makes sense.
I have now completed the installation successfully with Node.js 22. The remaining issue was specifically the sharp libvips download.
I downloaded the same asset successfully with curl on the Proxmox host and copied it into the container at:
/root/.npm/_libvips/libvips-8.14.5-linux-x64.tar.br
After that, yarn setup detected the cached archive, completed successfully, and AnythingLLM is now running together with the collector.
So Node.js 22 was not the cause. The reproducible failure was limited to the libvips download performed during the sharp 0.32.6 installation. A possible installer-side workaround would be to prefetch this asset using the existing curl download logic before running yarn setup, although the hardcoded sharp/libvips version would need to be handled carefully.
MickLesk commentedon Aug 26, 2026
here you can see, all over 18 is fine;
https://github.com/Mintplex-Labs/anything-llm/blob/a764c6d439e4f3c2ccdf05fbf1b9ea6261e21995/package.json#L17-L19
the libvips issue i can check in some mins
Ambre00 commentedon Aug 28, 2026
/ | ____ __ / // / () ____ / / / / / |/ /
/ /| | / __ / / / / __/ __ / / __ / __ `/ / / / / /|/ /
/ ___ |/ / / / // / // / / / / / / / // / // // / / /
// |// //_, /_// //// //_, //// //
// /_/
⚙️ Using Default Settings on node proxmox-3
💡 PVE Version 9.2.10 (Kernel: 7.0.14-11-pve)
🆔 Container ID: 111
🖥️ Operating System: debian (13)
📦 Container Type: Unprivileged
💾 Disk Size: 20 GB
🧠 CPU Cores: 4
🛠️ RAM Size: 4096 MiB
🎮 GPU Passthrough: Enabled
🚀 Creating a AnythingLLM LXC using the above default settings
✔️ Storage local (Free: 37.5GB Used: 22.9GB) [Template]
✔️ Storage local-lvm (Free: 3.6TB Used: 47.9GB) [Container]
✔️ Storage 'local-lvm' (lvmthin) validated
✔️ Template storage 'local' validated
✔️ Cluster is quorate
✔️ Template search completed
✔️ Template debian-13-standard_13.6-1_amd64.tar.zst [local]
✔️ LXC Container 111 was successfully created.
🎮 Detected NVIDIA GPU
🎮 Found 6 NVIDIA device(s) for passthrough
✔️ Automatically configuring NVIDIA GPU passthrough
✔️ NVIDIA GPU passthrough configured (1/1 GPU(s), 6 devices) - install drivers in container if needed
✔️ Started LXC Container
✔️ Network in LXC is reachable (ping)
✔️ GPU passthrough configured (video:44, render:992)
✔️ Customized LXC Container
✔️ Set up Container OS
✔️ Network Connected: 192.168.1.147
✔️ IPv4 Internet Connected
✔️ IPv6 Internet Connected
✔️ Git DNS: github.com:(✔️ ) raw.githubusercontent.com:(✔️ ) api.github.com:(✔️ )
✔️ Updated Container OS
✔️ Installed Dependencies
✔️ Setup Node.js 22
✔️ Installed yarn
✔️ Deployed: anythingllm (1.16.1)
✔️ Configured AnythingLLM
✖️ in line 48: exit code 1 (General error / Operation not permitted): while executing command yarn setup
--- Last 20 lines of log ---
CXX(target) Release/obj.target/sharp-linux-x64/src/common.o
../src/common.cc:13:10: fatal error: vips/vips8: No such file or directory
13 | #include <vips/vips8>
| ^~~~~~~~~~~~
compilation terminated.
make: *** [sharp-linux-x64.target.mk:140: Release/obj.target/sharp-linux-x64/src/common.o] Error 1
make: Leaving directory '/opt/anythingllm/server/node_modules/sharp/build'
gyp ERR! build error
gyp ERR! stack Error:
makefailed with exit code: 2gyp ERR! stack at ChildProcess. (/usr/lib/node_modules/npm/node_modules/node-gyp/lib/build.js:219:23)
gyp ERR! System Linux 7.0.14-11-pve
gyp ERR! command "/usr/bin/node" "/usr/lib/node_modules/npm/node_modules/node-gyp/bin/node-gyp.js" "rebuild"
gyp ERR! cwd /opt/anythingllm/server/node_modules/sharp
gyp ERR! node -v v22.23.2
gyp ERR! node-gyp -v v11.5.0
gyp ERR! not ok
info Visit https://yarnpkg.com/en/docs/cli/install for documentation about this command.
error Command failed with exit code 1.
info Visit https://yarnpkg.com/en/docs/cli/run for documentation about this command.
[2026-08-28 12:31:18] [ERROR] in line 48: exit code 1 (General error / Operation not permitted): while executing command yarn setup
✖️ Installation failed in container 111 (exit code: 1)
📋 Installation log: /tmp/anythingllm-111-532463c7.log
Error: General error / Operation not permitted
What would you like to do?
Select option [1-4] (default: 1, auto-remove in 60s):
TheOne299 commentedon Aug 28, 2026
Installation of LXC hangs on installing "puppeteer
Novasty commentedon Sep 3, 2026
The application is ignoring the STORAGE_DIR flag in /opt/anythingllm/server/.env
During update, from 1.16.0 to 1.16.1, my workspace gets wiped as everything is saved in /opt/anythingllm/server/storage/ instead of /opt/anythingllm_data/storage.
MickLesk commentedon Sep 3, 2026
Then its an upstream Bug, i think
Pokornz commentedon Sep 8, 2026
With gemma4:12b running on Ollama, I get only the Thoughts status, but no answer.
I tried the same thing with the desktop app from AnythingLLM and I got the answer no problem.
github-actions commentedon Oct 9, 2026
A PR has been created for anythingllm: community-scripts/ProxmoxVE#17802
MickLesk commentedon Oct 10, 2026
Merged with #17802 in ProxmoxVE
github-actions commentedon Oct 10, 2026
Files deleted with PR #2408