Posts on Sanyam Chhabra's Blog

Building a Home Server

Tue, 28 Jul 2026 00:00:00 +0000

Introduction

Proxmox VE is a type 1 hypervisor which provides a headless web-based UI to manage VMs and LXCs. I am going to setup 2 devices, for a multi-node cluster.

Requirements

Minimum Req. : A PC with virtualization support and sufficient ram and cpu cores to be able to run VMs/containers. (If the device does not support Hardware virtualization, it can still run LXC Containers)
A Wired Internet Connection - It is not recommended to use wifi.

Proxmox Installation

Download latest Proxmox VE ISO from here
Flash the image to a pendrive (8gb or greater) using Rufus/Balena Etcher
Before installing Proxmox, change these settings in the BIOS:
- Turn off Secure Boot
- Turn on Hardware Virtualisation (VT-x/AMD-V or Intel VT-d/AMD IOMMU)
- Also turn on Hyper Threading if present
Now, boot using the Proxmox install media (pendrive)
Follow the Proxmox Installation GUI:
- Select the disk to install Proxmox on (it would be wiped, so double check)
- The password you set will be used as the root password, and for the Web-ui login
- In the Network Configuration, set hostname as {node_name}.local, and set the static IP (CIDR) for the machine (the IP addresses following the machine’s IP will be used for VMs/LXCs, so choose the IP with ample free addresses afterwards as per your router configuration). ex. - 192.168.1.200
After installation, reboot and you’d be greeted with the shell. You may now disconnect all external accessories from that PC (monitor, KBM). There is no need to mess with the shell, as everything would be managed using the integrated web-ui, which also provides web shell access.
Now, navigate to the web interface using the IP of the device set earlier. The default port for the web-ui is 8006. eg. - https://192.168.1.200:8006

Proxmox Setup

The default username is root and password is the one set during installation.

Post Install Setup

Proxmox VE Helper-Scripts

These Helper scripts provides a lot of prebuild templates for self hosting various services with ease. Run these scripts using the web shell proxmox provides.

This script provides options for managing Proxmox VE repositories, including disabling the Enterprise Repo, adding or correcting PVE sources, enabling the No-Subscription Repo, adding the test Repo, disabling the subscription nag, updating Proxmox VE, and rebooting the system.

By default, you will not receive proxmox updates without an active enterprise subscription, to fix that this script adds the no subscription (community) repo to proxmox.

Here are my recommended options for the script: Disable Enterprise repos Enable No-Subscription repos Select no to PVE-Test repo for maximum stability Disable Subscription nag Don’t disable High Availability service if you want to create a HA cluster using 2 or more nodes, for replication and backup. If not, then disable it to free up system resources. It can always be reactivated later aswell, under the system tab of the node Update and reboot PVE Also, you can always rerun the script again in the future to apply any other optimizations.

Making the default bridge VLAN-aware

Proxmox VLANs offer a multitude of benefits for your virtual environment. By leveraging VLANs in your Proxmox network, you can enhance security, improve performance, and streamline network management. Know more about VLANs

To VLAN enable your default bridge:

Go to your node under datacenter, then under System -> Network -> default Linux bridge (vmbr0), click edit, and check the VLAN aware checkbox. Click ok, then Apply Configuration. Restart the proxmox server to apply the changes.

Setting up ssl certificates for Proxmox

The proxmox webui can be secured by configuring ssl certificates which will also remove the warning popups saying the website is not secure. The typical way to do this is using LetsEncrypt, but it is a bit long and complicated process, and is not completely free. You will need a domain name, and then will need to use cloudflare dns for managing the domain. It is also required to regularly renew the Lets Encrypt certificate, which can be automated using certbot, but requires additional setup. Do this only if you plan on exposing the web-ui to the internet, for remote access. For the detailed process, refer to this

Setting up Two-Factor Authentication

If you plan to expose your proxmox server to the internet, it is recommended to setup multifactor auth. To set up 2FA: Datacenter -> permissions -> Two Factor -> Add -> TOTP. Use any Authenticator app (Google/Microsoft Authenticator) to set it up using the QR code

Bonus (For Laptops)

Prevent Laptop from going to sleep on closing the Lid

Open shell, and edit the service login file /etc/systemd/logind.conf Use nano/vim as per preference
Find and uncomment the 3 HandleLidSwitch settings, and set each of them to ignore (from suspend)

1
2
3

HandleLidSwitch=ignore
HandleLidswitchExternalPower=ignore
HandleLidswitchDocked=ignore

Restart the login service to apply changes systemctl restart systemd-logind.service

Screensaver (Turn off display on idle)

Even on closing the lid of the laptop, the screen stays on continuously, which may cause the screen to burn out over time, making it unusable

Using the shell, edit the Grub file /etc/default/grub
To put the screen to sleep after 1 min, set GRUB_CMDLINE_LINUX="consoleblank=60"
To apply the changes update-grub

Setting up Storage

There are basically two different classes of storage types:

File level storage: They are in general more flexible than any Block level storage (see below), and allow you to store content of any type. On file based storages, snapshots are possible with the qcow2 format.
Block level storage: Allows to store large raw images. It is usually not possible to store other files (ISO, backups, ..) on such storage types.

First lets talk about the major storage types proxmox supports:

LVM

It is a block level storage thus it can cannot store anything other than VM disk images and containers. It has very little overhead and preferred over directory to store VMs and Containers. But LVM normally allocates blocks when you create a volume, i.e. storage is pre allocated for volumes in LVM, which can result in a lot of wasted storage. Also, It is possible to use LVM on top of an iSCSI or FC-based storage. That way you get a shared LVM storage.

LVM-thin

It is similar to LVM and is a block type storage. But, LVM-thin but fully supports snapshots and clones efficiently, and storage is dynamically allocated for volumes. This behavior is called thin-provisioning, because only the blocks that the guest system actually use will be written to the storage i.e. With thin provisioning activated, volumes can be much larger than physically available space.

Say for instance you create a VM with a 32GB hard disk, and after installing the guest system OS, the root file system of the VM contains 3 GB of data. In that case only 3GB are written to the storage, even if the guest VM sees a 32GB hard drive. In this way thin provisioning allows you to create disk images which are larger than the currently available storage blocks. You can create large disk images for your VMs, and when the need arises, add more disks to your storage without resizing the VMs’ file systems.

This means that for example if you have 100 gb storage available, but you can still assign a 150 gb hard disk to a VM and as long as it actually uses less than 100 gb of storage everything will work fine.

Warning: If the storage runs full, all guests using volumes on that storage receive IO errors. This can cause file system inconsistencies and may corrupt your data. So it is advisable to avoid over-provisioning of your storage resources, or carefully observe free space to avoid such conditions.

It must be mentioned that LVM thin pools cannot be shared across multiple nodes, so it can only be used as local storage. Even though thin provisioning does cause a negligible performance loss i would recommend using this over lvm in most cases.

ZFS

ZFS is technically a file level storage, but Disk images for VMs are stored in ZFS volume (zvol) datasets, which provide block device functionality. Hence it provides a hybrid between both file and block level storage. ZFS is probably the most advanced system, and it has full support for snapshots and clones. ZFS supports storage pools with varying levels of data redundancy. When a bad data block is detected, ZFS fetches the correct data from another redundant copy and repairs the bad data, replacing it with the correct data. ZFS is used as an alternative to raid which is used a way of storing the same data in different places on multiple hard disks to protect data in the case of a drive failure.

ZFS is only recommended when you have multiple harddisks. In a ZFS setup, for example if you have two 1tb drives (2 tb total) and use one as a mirror drive, then only 1tb of net storage will be available for use. ZFS is more suited for more advanced users like developers who constantly move data around different disks and servers. For actually utilising benefits of ZFS, you may need and enterprise class ssd for dedicated cache/ log disk. Also, ZFS depends heavily on memory, so it is not recommended with less than 8GB ram, and over that you should have at least approx. 1 gb of ram extra for every 1tb of harddisk storage to be used. To prevent data corruption, it is recommend to use high quality ECC RAM.

ZFS may start causing issues when your hardrive gets filled upto 80-90% and the the performance also gets degraded. Also ZFS adds a major overhead and can cause significant performance loss on old low end systems. Hence it is not recommended for basic user to setup zfs.

(If you are experimenting with an installation of Proxmox VE inside a VM (Nested Virtualization), don’t use virtio for disks of that VM, as they are not supported by ZFS. Use IDE or SCSI instead (also works with the virtio SCSI controller type).)

Note: Do not use ZFS on top of a hardware RAID controller which has its own cache management. ZFS needs to communicate directly with the disks. An HBA adapter or something like an LSI controller flashed in “IT” mode is more appropriate.

Ceph

It is also an advanced storage type which is used as shared storage in a High Availability cluster.

For more information: https://pve.proxmox.com/wiki/Storage

Configuring Drives

First we need to setup our storage device because by default proxmox is not configured to use all the storage of our system. We need to perform this step before deploying any virtual machines or containers on our system.

Initially, when you go to your node you can see that only your boot drive has been configured and there are 2 partitions of that storage.

One being local which can be used to store everything but is currently only configured to be able to store iso images and container templates. it currently can’t store VMs or Containers. It has the smaller fraction of your storage right now The other one is local-lvm, which actually has the major part of your storage but it can only be configured to be able to store Disk images of VMs and containers.

I have multiple drives, so what I am going to do is delete the local-lvm partition and actually assign all the storage to local volume then configure it to store everything including iso files, container templates and backups. Basically i am gonna use the smaller boot drive for storage, and the other bigger drive for disk images of VMs and LXCs.

Now, follow these steps to resize our partitions:

We will start by deleting that volume. To do that, under Datacenter -> Storage, select local-lvm and remove it. This deleted the local-lvm volume, but that storage is still not available for use. To fix that we need to use shell commands.
Perform the following commands:

// To delete local-lvm from cli
lvremove /dev/pve/data

// To reallocate the space
lvresize -l +100%FREE /dev/pve/root

// To confirm resize of main storage
resize2fs /dev/mapper/pve-root

You could also do this, if you have a single HDD, and use a single partition for everything. Don’t do this if using a single SSD, as it may degrade its health faster.

Now when you check the local directory under your node, you can see that the size of that volume is now utilizing the full drive. But, we still have one step left to give the local directory permissions of the type of data it can store.

To do that, go to Datacenter -> storage then, select local and click edit. In the edit menu we need to specify what type of content can be allowed in this directory. By Default Disk images and containers are not allowed.

If you only have this single drive in your system, i will recommend you to select every thing in the contents section. Otherwise, if you have another storage drive for VMs and containers, you can leave Disk Image and Containers Unchecked for simplicity.

If you have other storage devices in your system aswell, follow these steps to configure them :

Go to your node -> Disks, under there you should see all the storage drives connected to your computer. Other than the boot drive, select the one you want to add and click wipe disk
This should delete all existing partitions and data on that disk. Now select it again and click initialize disk with GPT
Now, under the Disks dropdown menu, select the storage type you want to use between directory, lvm, lvm-thin and zfs according to your use case. If unsure, I would recommend using LVM-thin
Click on lvm-thin, then click Create Thinpool and under disk, select the disk you want to add, then give it a name. Make sure Add Storage is Checked. Then click Create
If you have other storage devices you want to add aswell, then you can create a new thin pool for lvm-thin which is the easier option.

Just make sure to edit the Content type allowed on the newly configured drives aswell. Disk image and Container should be checked, to store VMs and LXCs on that drive.

Bonus (For advanced use cases)

It is possible to add other storage device in an existing thin pool aswell but that is a tedious process and has risks because if not done correctly, it can corrupt your proxmox installation. Make sure to do this before storing anything on the drives to prevent data loss

Here are a few commands to check the current configuration in host shell:

pvs
vgs
lvs
lsblk

First make sure your volume group (VG) called pve takes up almost the entire partition. After that, extending local-lvm is quite simple. Just make sure you don’t forget to also extend metadata.

// Find new disk
fdisk -l

// Format the disk
cfdisk /dev/sdc    // use the correct name for the disk to add
// > New -> Primary -> Specify size in MB  
// > Write  
// > Quit  

// Create volume
pvcreate /dev/sdc1    // use correct partition name

// Add new volume to lvm
vgextend pve /dev/sdc1

// Increase volume of lvm to take advantage of new HDD
lvextend /dev/pve/data -L +251.50g    // use *vgs* command to find correct size of new volume 

lvresize --poolmetadatasize +1GB pve/data

If the partition already has the correct size, the following command can also be used instead of the 2nd last command, but keep in mind that they delete the LVM partition. If you do that all your data will be lost if you don’t have a backup.

First extend your metadata and then use, lvresize -l +100%FREE pve/data

For more details, Follow this guide

Creating a Cluster (Optional)

If you have multiple computers you want to install proxmox on, to increase your total available resources, then, you can add them in a cluster for easy management. Just make sure to give each of your proxmox servers different nodenames during the setup process in the Hostname section. But first lets under stand what Cluster, High Availability and replication is:

Clustering: Clustering in Proxmox involves connecting multiple Proxmox nodes together to create a distributed system. The main benefit is scalability; as your resource needs grow, you can add more nodes to the cluster. Clustering allows the system to manage multiple nodes from a single interface and aids in load balancing. However, clustering doesn’t inherently provide redundancy. If a node fails, the virtual machines (VMs) or containers running on that node will be affected unless you’ve implemented additional measures.

High Availability (HA): HA, on the other hand, is specifically designed for redundancy. With HA enabled, in the event of a node failure, Proxmox can automatically start the affected VMs or containers on another node in the cluster, minimizing downtime. It does this by monitoring the status of VMs/containers across the cluster and managing their states to ensure they remain operational. HA requires a cluster to work, but a cluster doesn’t inherently include HA.

Replication: Replication allows you to migrate VMs with less downtime by copying the contents of the HDD between local storage on different nodes. If you have shared storage (NFS, Ceph etc) then you don’t need to run replication. HA monitors the state of the VM and either starts or migrates it as required to meet the desired state.

HA and replication are used together and they do require you to setup a zfspool storage, so it is only recommended for Advance users to set this up.

In summary, clustering in Proxmox provides a way to scale your resources and manage them from a single interface, while high availability adds a layer of redundancy to maintain service continuity in the event of a node failure. They often work together to create a more robust, scalable, and resilient system.

Even If you don’t want to use High Availability and replication, Clustering offers the following benefits:

Centralized web-based management that allows you to view all the nodes and their statistics from any node.
Seamless migration of virtual machines and containers between physical hosts.
Faster deployment of resources.
Cluster-wide services like a firewall and High Availability.
Use of pmxcfs, a database-driven file system. This is a file system for storing configuration files. It is replicated in real-time across all the proxmox nodes using corosync.

Note: You should Create a cluster before creating any VMs or containers, as it may cause problems while joining a cluster if multiple nodes have VMs or Containers with the same ID.

Setting up the Cluster

First, go to the web ui of your primary proxmox node. Then go to datacenter and then click on cluster. Now click on Create Cluster
Now give your cluster a name and select the server’s static ip under cluster network then click create. Wait for it to finish, when it says TASK OK, close the Task Viewer
Now click on Join Information on the top, then click on Copy Information
Now, go to the Web UI of your other Proxmox Nodes that you want to add to the cluster, and go to datacenter, then click on cluster, and then click Join Cluster and paste the join information you copied earlier from the primary node. Now type your primary node’s password in the password section and click Join
Now, it is going to take a while to finish setting up the cluster and the process may get stuck inbetween and show and error or connection lost. Just refresh the web interface pages of all your nodes and after logging in again you will see under the Cluster Information Section, that they joined the cluster successfully

Now, under datacenter you can see and manage all your nodes from any single node’s web ui, aswell as effortlessly migrate any VM or container from one node to another.

There is an issue where if you have setup any VM or Container before joining a cluster, it may not work and give an error. To fix that issue you can refer to this

High Availability and Replication

To setup HA and replication, you can refer to this

Bonus Tip

Proxmox cluster has the concept of “quorum” and it’s ensuring stability of the cluster’s state. By default, you will need at least greater than (total number of nodes/2) nodes online in order to start/use containers or VMs. Each node will have 1 vote, and to operate your cluster, you need to have majority numbers. If the minimum numbers of votes are not met, it will throw an error “TASK ERROR: cluster not ready - no quorum?”, whenever your proxmox tries to autostart VMs or containers on boot or you manually try to start them. In some cases even the shell would not open. To fix this, first you need access to the shell, either via the web ui or using ssh. Then you can use the command: pvecm expected 1 This will maually set expected votes to 1. But, If it gives the error” Cannot initialize CMAP service”, make sure corosync service is running using the command: systemctl status corosync.service

But this is only a one time solution, and you will need to run the command every-time to fix the issue. This step can also be automated by making a Bashscript and autorunning it on boot.

Make a new script using: nano quorum.sh
Paste the following

1
2
3

#!/bin/bash

pvecm expected 1

Make the script executable using: chmod +x quorum.sh
To make it autorun on boot, create a new service file using: nano /etc/systemd/system/quorum.service
Paste the following

[Unit]
Description=Quorum skip
After=network.target

[Service]
ExecStartPre=/bin/sleep 30
ExecStart=/root/quorum.sh

[Install]
WantedBy=multi-user.target

Refresh the services list using: systemctl daemon-reload
Enable the autostart on boot using: systemctl enable quorum.service

On the next reboot, the quorum will automatically be met.

Setting up VMs and LXCs

Virtual Machines

The process of setting up a VM remains the same for all linux distros but there are a few extra setup steps required for windows, especially windows 11 due to its requirements.

Download the iso installer files for an OS, and upload it to proxmox, using the webui, under: your_node->local(storage)->ISO Images
Click Create VM button on top, and just follow the VM creation wizard.
Under the OS section, when you select the iso file, proxmox should automatically detect the Guest OS type, else it can be set manually. This is important, because according to the Guest OS, the optimal system settings would be configured automatically.

VirtIO Drivers

VirtIO Drivers are paravirtualized drivers for kvm/Linux. In short, they enable direct (paravirtualized) access to devices and peripherals for virtual machines using them, instead of slower emulated ones. This massively improves the performance of a VM.

Linux has them preinstalled, but Windows does not have native support for VirtIO devices included, so we have to manually install them. To do that we need the official kvm driver disk.

You can get the latest stable VirtIO Drivers ISO from the official github repo

You just need to download the VirtIO ISO and upload it to proxmox. Then, check the add additional drive for VirtIO drivers and then select the VirtIO ISO under that, in the VM creation Wizard

LXC

These are a bit different than the more popular docker containers. Linux containers work on the same containerization principle as docker. They are lightweight as it shares the host’s kernel so only linux containers are possible. Windows can not be used as LXC. These work pretty similar to a CLI based Linux distro but require much less resources than running it on a VM. But a VM is much more secure with far less risk of harming the Host OS (proxmox) if in case it gets compromised. This is applicable more if you are going to expose that to the internet. Proxmox provides a few LXC templates, but the best part is, it also supports the universal container template format, and allows downloading those templates from the OCI registry. That means, it is possible to use the docker container templates for LXCs aswell

To setup an LXC:

Get container templates from under you_node->local(storage)->CT templates.
Click Create CT button on the top, and just follow the LXC creation wizard.

Bonus Tip

Instead of setting up all VMs and Containers manually, check out the Proxmox VE Helper Scripts It provides a lot of scripts to easily setup popular services to self host, with no hassle.

Bonus

Windows on Proxmox

https://youtu.be/fupuTkkKPDU?si=ZV7H_grfZNZ68rBa

Proxmox Networking

https://youtu.be/zx5LFqyMPMU?si=qlU4RtaiZhqPmK5h

SDN

https://youtu.be/_lIk9p_SyvU?si=rKtMx-mMwstiysM4

Setup Email alerts

https://youtu.be/85ME8i4Ry6A?si=30ICKcrIh-D5ZVHr

PCIe Passthrough

It is also possible to pass through a GPU to a VM in proxmox. Also, this only works for dedicated GPU. It is also possible to passthrough other PCIe devices such as network cards. But there are some requirements to do that. First you should have a CPU that supports IOMMU. There are different names for that technology such as VT-d/VT-x, etc.

For more information you can refer to these:

https://www.reddit.com/r/homelab/comments/b5xpua/the_ultimate_beginners_guide_to_gpu_passthrough/

https://youtu.be/_hOBAGKLQkI?si=OtquZnS-REAIfWGz https://youtu.be/S6jQx4AJlFw?si=VO86VPvJzjLLFdiv

To passthrough a gpu to multiple LXCs : https://youtu.be/0ZDr5h52OOE?si=YacPpr5RT3d73inC

Remote Access

For remotely accessing your self hosted services, and or the proxmox webui, there are much safer options than port forwarding.

Cloudflare Tunnels

You require a domain name with Cloudflare as the dns manager. This method is not completely free, since domains need to be renewed. All you need to do is setup a cloudflare tunnel connector on the proxmox server, then cloudflare makes it very easy to manage which ip and ports you want to expose, using a subdomain, with no other manual firewall configuration required. Learn more

VPNs

Not the traditional VPNs or reverse proxy, which still requires manual config. Services like ZeroTier, Tailscale, Twingate provide easy to setup Zero Trust, Mesh VPNs, and are free for personal use (with a max device limit on the free tier). These solutions are better & more secure than traditional VPNs, as they use Split-tunneling (Split DNS), i.e. it only routes internal traffic through the vpn. Setting these up is as simple as installing a node connector on the server, and having a client app on the device, which would be used to access the server, for the VPN connection.

Building a Voice Assistant

Fri, 20 Jun 2025 00:00:00 +0000

Introduction

I am going to setup Home Assistant for local home automation on a VM, then create a voice pipeline to setup a local AI voice assistant. I am also going to train a custom voice model for the Voice Assistant. I am going to be cloning the voice of Yoda (from star wars)

Setting up Home Assistant

Install Home Assistant : https://www.home-assistant.io/installation
Access the WebUI : http://homeassistant.local:8123
Create an account and then setup any or all of your IoT devices under Settings->Devices & services.
Additionally, it is also possible to setup the HA companion app on mobile, and create Automations.
Also, it is possible to setup Tailscale or ZeroTier to access HA remotely, by installing their respective Add-ons from the Addon store.

Setting up Voice Pipeline

Under Settings->Add-ons , Install the following Addons :
- Whisper (STT)
- Piper (TTS)
- Assist Microphone
- openWakeWord
Start all the Addons, then under Devices & services, add the 4 Wyoming Protocol Devices
Also, add the Ollama Integration to connect and use an LLM as the conversation agent for the assistant, to make it smarter and remember context. Ollama could be setup on the HA host or any other PC with a GPU for better speed. I am using llama3.2 Note: If running ollama on the same machine, the IP for the host is usually the gateway address of the subnet of the VM (if not using bridged adapter). Also, use this command for ollama to allow commections from remote services :

`1`	`OLLAMA_HOST=0.0.0.0 ollama serve`

Under Settings->Voice Assistants, Add Assistant,
- Change the name, then under Conversation Agent, select the LLM added using the Ollama Integration, which is way smarter than the default Home Assistant agent. Then in settings of ollama conversation agent, Turn Assist on under Control Home Assistant
- For Speech-to-Text, select faster-whisper, and for Text to Speech, select piper.
- Also, to use a wake word, Add streaming wake word, then under Wake word engine select openwakeword, and select a Wake Word. Note: To activate assistant through a wake word, a usb speaker and mic needs to be used (passthrough to the HAOS VM ,if applicable), and selected under the Assist Microphone addon Configuration
To access the Assistant, use the Assist icon on HA dashboard. In the webui, it is only possible to talk to the assistant using text. To use voice, an SSL certificate needs to be added to HA to get a HTTPS connection.
Here’s a guide to get ssl certs for HomeAssistant: https://novamostra.com/2023/04/29/add-ssl-tls-certificate-to-homeassistant-web-interface/
The easier way to converse with the assistant using voice is to use the Home Assistant Companion App on mobile.
It is also possible to setup a Wyoming Satellite on another device to access the Voice Assistant with wake word trigger. https://github.com/rhasspy/wyoming-satellite.git Note: It is also possible to offload Whisper (SST) and Piper (TTS) to another device with a GPU to improve the Assistant’s response speed. https://github.com/rhasspy/wyoming-addons Then add the service using the Wyoming Protocol
It is also possible to train a custom wake word : https://www.home-assistant.io/voice_control/create_wake_word/

Training a custom voice clone model

I will be training the voice model using Piper, which will give the model in a standard format (.onnx), which can be used in many TTS services and for the voice assistant aswell. This process requires the use of linux and an Nvidia GPU. I used WSL with Ubuntu-22.04

Step 1: Prerequisites

Piper works best with Python-3.10 and cuda-11.7

sudo apt update && sudo apt upgrade
sudo apt install python3-venv python3-dev gcc
sudo apt-get install g++ freeglut3-dev build-essential libx11-dev libxmu-dev libxi-dev libglu1-mesa libglu1-mesa-dev

## Adding GPU Drivers repo

sudo add-apt-repository ppa:graphics-drivers/ppa  
sudo apt update
# Install the compatible GPU drivers if not preinstalled

## Installing Cuda

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-ubuntu2204.pin  

sudo mv cuda-ubuntu2204.pin /etc/apt/preferences.d/cuda-repository-pin-600  
  
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/3bf863cc.pub  
  
sudo add-apt-repository "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/ /"  
  
sudo apt-get update  
  
sudo apt install cuda-11-7  

echo 'export PATH=/usr/local/cuda-11.7/bin:$PATH' >> ~/.bashrc  
  
echo 'export LD_LIBRARY_PATH=/usr/local/cuda-11.7/lib64:$LD_LIBRARY_PATH' >> ~/.bashrc  
  
source ~/.bashrc  
  
sudo ldconfig

sudo reboot

## To verify install

nvcc -V

nvidia-smi

Step 2: Data Prep

The key to any AI project is getting CLEAN DATA. There are two options to get the data,

Record your voice using the Piper Recording Studio : https://github.com/rhasspy/piper-recording-studio.git

Pull audio data from YouTube videos :

Download YT Videos : Around 1 hr of clips should be good enough

cd ~
mkdir dataprep
cd dataprep

python3 -m venv .venv
source .venv/bin/activate

pip install yt-dlp

yt-dlp -x --audio-format wav "YT-URL"

Cleanup your audio samples : Piper requires audio samples around 10sec each. Use Audacity to trim clips. When exporting from Audacity, Piper requires the audio in this format:
- Format: wav
- Channels: mono
- Sample Rate: 22050 Hz
Remove Silence & Cut up your files :

for file in ./*.wav; do
    ffmpeg -i "$file" -af "silenceremove=stop_periods=-1:stop_duration=3:stop_threshold=-25dB" "./${file%.wav}_nosilence.wav"
done


mkdir wav 

for file in *.wav; do
  ffmpeg -i "$file" -f segment -segment_time 15 -c copy "./wav/split_${file%.*}_%03d.wav"
done

Transcribe your audio samples with Whisper (Using a python script) :

1
2
3

pip install git+https://github.com/openai/whisper.git

nano transcribe.py

import os
import whisper

model = whisper.load_model("base")

audio_dir = "./wav"
output_csv = "./metadata.csv"

audio_files = [f for f in os.listdir(audio_dir) if f.endswith(".wav")]
audio_files.sort()

with open(output_csv, "w") as f:
    for audio_file in audio_files:
        audio_path = os.path.join(audio_dir, audio_file)
        
        result = model.transcribe(audio_path)

        transcription = result["text"].strip()

        file_id = os.path.splitext(audio_file)[0]
        
        f.write(f"{file_id}|{transcription}\n")

print(f"Metadata saved to {output_csv}")

python3 transcribe.py

deactivate

cd ~/

Step 3: Training Prep

Prepare the Training Env : 11.8(Nvidia 40xx) Note: For Nvidia 40 series, also follow this: https://github.com/rhasspy/piper/issues/295

mkdir training
cd training

git clone https://github.com/rhasspy/piper.git

python3 -m venv .venv
source .venv/bin/activate

python3 -m pip install pip==23.3.1
pip install numpy==1.24.4
pip install torchmetrics==0.11.4

cd piper/src/python

python3 -m pip install --upgrade wheel setuptools

pip3 install -e .

./build_monotonic_align.sh

Process dataset to the format piper requires:

python3 -m piper_train.preprocess \
  --language en \
  --input-dir ~/dataprep/ \
  --output-dir ~/training/dataset \
  --dataset-format ljspeech \
  --single-speaker \
  --sample-rate 22050

Download a starting point (checkpoint) : https://huggingface.co/datasets/rhasspy/piper-checkpoints/tree/main/en/en_US/lessac . I would recommend using (lessac) Medium

`1`	`wget https://huggingface.co/datasets/rhasspy/piper-checkpoints/resolve/main/en/en_US/lessac/medium/epoch%3D2164-step%3D1355540.ckpt`

Step 4: Training the model

python3 -m piper_train \
    --dataset-dir ~/training/dataset/ \
    --accelerator 'gpu' \
    --gpus 1 \
    --batch-size 16 \
    --validation-split 0.0 \
    --num-test-examples 0 \
    --max_epochs 6000 \
    --resume_from_checkpoint "~/training/epoch=2164-step=1355540.ckpt" \
    --checkpoint-epochs 1 \
    --precision 16 \
    --max-phoneme-ids 400 \
    --quality medium

Pausing and Resuming
- The training can be interrupted and picked up where it was left off by ctrl-c
- To resume training, use the same command but change the checkpoint file to the most recent checkpoint file, created under : dataset/lightning_logs
Exporting the voice model : After the final epoch is reached, training is done, and model can be exported

# Export the model

python3 -m piper_train.export_onnx \
    "~/training/dataset/lightning_logs/version_0/checkpoints/epoch=5999-step=xxxxxx.ckpt" \
    ~/output/directory/voicemodel.onnx

1
2
3

# Copy the training json file to your model file directory

cp ~/training/dataset/config.json ~/output/directory/voicemodel.onnx.json

Testing the Voice Model

1
2
3

pip install piper-tts

echo "Completed, the training is" | piper -m voicename.onnx --output_file test.wav

Using the Voice Model for HA voice assistant

The Voice Model can be transferred to Home Assistant using the SAMBA share plugin.
Inside the share folder, Create a new folder : piper , and paste the voicemodel.onnx & voicemodel.onnx.json files inside this folder.
To edit the name this voice model shows as inside HA, edit the dataset & quality inside the voicemodel.onnx.json file.
After restarting Home Assistant, this voice model should be ready for use, and can be selected by editing the Voice Assistant settings, and under Text-to-speech, a new language should be created (English), and under Voice the new Voice Model should be present.
It is also possible to change the system prompt of the LLM in its settings to match the persona of who’s voice was cloned.

Self-Hosting an AI server

Wed, 15 Jan 2025 00:00:00 +0000

Introduction

I am going to install Ubuntu Server on a PC with a GPU, to make a Headless server, to run all my AI locally.
I will run a local LLM using Ollama and Image generation using stable diffusion, on a GUI using Open WebUI.
I will be using my old laptop with an NVIDIA gpu, to dualboot Ubuntu server, with the windows 10 pre-installed.
Note: All the AI tools mentioned, can also be installed and run on any computer (with any os), and a dedicated linux gpu server is not compulsory.
Prerequisites:
- A Computer with atleast 8gb ram & a GPU (preferably NVIDIA)
- An empty USB drive (atleast 8gb)
- An extra drive on the computer to install Ubuntu Server on (If dual-booting with previously installed OS)

Preparing the Server (setting up dependencies)

Installation of Ubuntu Server

Create a bootable usb, flashed with ubuntu server image : The ISO file can be downloaded from Ubuntu’s Website. Software such as Rufus or BalenaEtcher can be used to flash the image to the usb drive.
Boot from the usb drive with Ubuntu Server Image : Turn off the computer and boot into bios, by spamming the bios key for your pc. From the boot menu, select the USB drive.
Install Ubuntu Server : Press enter to select Try or Install Ubuntu Server. It is a pretty standard OS installation. Press Enter to Confirm, Arrow keys to move between the options, and space to select.
- Choose your language
- Select Keyboard layout
- Choose installation type as Ubuntu Server & also select Search for 3rd-party drivers
- Setup your Network Configuration. It is recommended to use LAN and also manually bind your pc a static ip instead of using DHCP
- Set proxy if required
- Let Ubuntu test default Mirror and proceed
- Setup your storage configuration. It is recommended to allot a full disk for installation
- Re-confirm your Installation drive, so as to not wipe anything important, then continue
- Fill in your information including your name, name for the server, username and password
- If Ubuntu found any GPU drivers, it will prompt for install confirmation
- Skip Ubuntu Pro for now
- Select install OpenSSH server to access your server remotely using another pc over ssh. Also import SSH keys using GitHub, to not have to authenticate everytime you ssh into the server. Also select Allow Password Authentication, to allow other devices without Authorised keys, to login by using credentials
- Select any snaps(packages) you want to be installed automatically
- Let Ubuntu server install on your PC, then reboot. also remove the installation media (pendrive) while rebooting.

Post-Install Setup

Now whenever the PC turns on, you will be greeted the the grub bootloader menu, where you can select which os you want to boot into, Ubuntu or any previously installed OS. Select Ubuntu and after it boots up, login using the username and password set during installation.

Updating Repositories and packages :

`1`	`sudo apt update && sudo apt upgrade -y`

Updating the Main partition to use all the free storage available : by default, only about half of the drive is available as usable storage

`1`	`sudo lvextend -l +100%FREE /dev/ubuntu-vg/ubuntu-lv && sudo resize2fs /dev/ubuntu-vg/ubuntu-lv`

Prevent Laptop from sleeping when lid closed & screensaver (optional, for laptops only) :

`1`	`sudo nano /etc/systemd/logind.conf`

Uncomment & set the HandleLidSwitch settings to ignore, then save & exit : HandleLidSwitch=ignore HandleLidSwitchExternalPower=ignore
HandleLidSwitchDocked=ignore

`1`	`sudo systemctl restart systemd-logind.service`

`1`	`sudo nano /etc/default/grub`

Set the time for display to sleep (in seconds), then save & exit : GRUB_CMDLINE_LINUX="consoleblank=60"

`1`	`sudo update-grub`

Setting up GPU

Installing GPU drivers : Ubuntu Documentation

`1`	`sudo ubuntu-drivers install`

Reboot the computer : sudo reboot
Testing (nvidia) drivers : nvidia-smi

Installing NVIDIA container toolkit (Nvidia GPUs only) : Official Documentation

curl -fsSL https://nvidia.github.io/libnvidia-container/gpgkey | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-container-toolkit-keyring.gpg \
  && curl -s -L https://nvidia.github.io/libnvidia-container/stable/deb/nvidia-container-toolkit.list | \
    sed 's#deb https://#deb [signed-by=/usr/share/keyrings/nvidia-container-toolkit-keyring.gpg] https://#g' | \
    sudo tee /etc/apt/sources.list.d/nvidia-container-toolkit.list

`1`	`sudo apt-get update`

`1`	`sudo apt-get install -y nvidia-container-toolkit`

Docker installation & setup (for running containers)

Install Docker :

# Add Docker's official GPG key:
sudo apt-get update
sudo apt-get install ca-certificates curl
sudo install -m 0755 -d /etc/apt/keyrings
sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyrings/docker.asc
sudo chmod a+r /etc/apt/keyrings/docker.asc

# Add the repository to Apt sources:
echo \
  "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \
  $(. /etc/os-release && echo "$VERSION_CODENAME") stable" | \
  sudo tee /etc/apt/sources.list.d/docker.list > /dev/null
sudo apt-get update

`1`	`sudo apt-get install docker-ce docker-ce-cli containerd.io docker-buildx-plugin docker-compose-plugin`

Post-installation steps :

## Manage Docker without prefacing the docker command with sudo

sudo groupadd docker
sudo usermod -aG docker $USER
newgrp docker

## Configure Docker to start on boot

sudo systemctl enable docker.service
sudo systemctl enable containerd.service

Configuring (nvidia) GPU-acceleration support for Docker Containers :

1
2

sudo nvidia-ctk runtime configure --runtime=docker
sudo systemctl restart docker

Also, reboot the PC : sudo reboot

Install Portainer (GUI for managing Docker Containers)

Documentation

`1`	`docker volume create portainer_data`

`1`	`docker run -d -p 8000:8000 -p 9443:9443 --name portainer --restart=always -v /var/run/docker.sock:/var/run/docker.sock -v portainer_data:/data portainer/portainer-ce:latest`

To check the container status : docker ps

Access the Portainer GUI : https://{server_ip}:9443

Installation & Setup of AI tools

Ollama (for local LLMs)

Install Ollama

`1`	`curl -fsSL https://ollama.com/install.sh \| sh`

To change Ollama config :

`1`	`systemctl edit ollama.service`

Here’s the config I use :

[Service]
Environment="OLLAMA_HOST=0.0.0.0"
Environment="OLLAMA_CONTEXT_LENGTH=98304"
Environment="OLLAMA_KEEP_ALIVE=15m"
Environment="OLLAMA_ORIGINS=*"
Environment="OLLAMA_FLASH_ATTENTION=1"
Environment="OLLAMA_KV_CACHE_TYPE=q4_0"

Then restart ollama :

1
2

systemctl daemon-reload
systemctl restart ollama

To Download a Model : Models list

`1`	`ollama pull qwen3.8`

To Run the Model :

`1`	`ollama run qwen3.8`

Ollama Integrations : https://github.com/ollama/ollama#extensions--plugins

LiteLLM

To create and expose an OpenAI compatible gateway

Install LiteLLM:

`1`	`uv add litellm`

Create config : ~/.config/litellm/config.yaml

model_list:
  - model_name: "qwen3.8"
    litellm_params:
      model: "ollama_chat/qwen3.8"
      api_base: "http://localhost:11434"
    model_info:
      supports_function_calling: true
litellm_settings:
  drop_params: true
general_settings:
  disable_admin_ui: true
  disable_env_credential_login: true
  master_key: sk-$(openssl rand -hex 32)

Create a service : sudo nano /etc/systemd/system/litellm.service

[Unit]
Description=LiteLLM Proxy
After=network-online.target ollama.service
Wants=network-online.target
Requires=ollama.service

[Service]
Type=simple
User=$(whoami)
WorkingDirectory=/home/$(whoami)

ExecStart=$(which litellm) --config /home/$(whoami)/.config/litellm/config.yaml

Restart=always
RestartSec=5

StandardOutput=null
StandardError=journal

Environment=PYTHONUNBUFFERED=1

[Install]
WantedBy=multi-user.target

Start the Proxy :

1
2

sudo systemctl daemon-reload
sudo systemctl enable --now litellm

Automatic1111 (Stable Diffusion: for Local Image Generation)

Documentation

Pre-requisites :

`1`	`sudo apt install -y make build-essential libssl-dev zlib1g-dev libbz2-dev libreadline-dev libsqlite3-dev wget curl llvm libncurses5-dev libncursesw5-dev xz-utils tk-dev libffi-dev liblzma-dev git`

Setup Pyenv

`1`	`curl -fsSL https://pyenv.run \| bash`

echo 'export PYENV_ROOT="$HOME/.pyenv"' >> ~/.bashrc
echo '[[ -d $PYENV_ROOT/bin ]] && export PATH="$PYENV_ROOT/bin:$PATH"' >> ~/.bashrc
echo 'eval "$(pyenv init - bash)"' >> ~/.bashrc
echo 'eval "$(pyenv virtualenv-init -)"' >> ~/.bashrc

`1`	`source ~/.bashrc`

1
2

pyenv install 3.10.6
pyenv global 3.10.6

Install StableDiffusion WebUI

`1`	`wget -q https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui/master/webui.sh`

1
2

chmod +x webui.sh
./webui.sh

Download SD models : https://civitai.com/models

1
2
3

## Transfer model to ubuntu server

scp /path/to/downloaded/model/{model_name}.safetensors {username}@{server.ip}:~/stable-diffusion-webui/models/Stable-diffusion/

Note : You may also delete the default SD 1.5 model

Additional Configuration :

1
2

cd stable-diffusion-webui
nano webui-user.sh

Uncomment the COMMANDLINE_ARGS line set the following arguments :

`1`	`export COMMANDLINE_ARGS="--xformers --medvram --opt-split-attention --listen --api"`

--xformers : Only for NVIDIA GPUs

Note : Use the webui.sh script present inside the stable-diffusion-webui to run StableDiffusion with the specified Arguments The Stable Diffusion WebUI should be accessible at : {server_ip}:7860

Open WebUI (for an interactive GUI)

Installation :

`1`	`docker run -d --network=host -v open-webui:/app/backend/data -e OLLAMA_BASE_URL=http://127.0.0.1:11434 --name open-webui --restart always ghcr.io/open-webui/open-webui:main`

The UI should be accessible at : {server_ip}:8080

Additional Configuration : You can also set a default model, and enable Memory under Settings -> Personalization. Some other settings to enable are under Admin Panel -> Settings
- Under Documents, Enable PDF Extract Images (OCR)
- Under Web Search, enable Web Search and set Engine to DuckDuckGo Note: OpenWebUI also has the inbuilt functionality to use Documents for context aswell. It is also possible to add System Prompts to customize the model’s responses
To Integrate Image Generation (with prompt generation) :
- Go to Admin Panel -> Settings -> Images
- Set Image Generation Engine to Automatic1111, and set base url to http://127.0.0.1:7860
- Select the Default model and enable Image Generation
- Now just press the + icon in chat and select image, to generate images.
Note: If you have low vram (<=4GB), Image generation might cause Out of Memory errors. To help with that, Ollama can be set to unload the model from memory everytime, after generating a response.

`1`	`systemctl edit ollama.service`

Add this Environment Variable under [Service]

`1`	`Environment="OLLAMA_KEEP_ALIVE=0"`

Note : It can result in slower response times, as ollama needs to load the model into memory everytime

Building a Blog Website for free

Fri, 10 Jan 2025 00:00:00 +0000

Introduction

I am going to build a free website using a static site generator HUGO and hosting it on Cloudflare pages.
The workflow will be like this: The posts will be made in markdown format (using Obsidian), then it will be converted into html (using Hugo), then it will be pushed to a github repository, from where a free static hosting provider (cloudflare pages) will pull the public directory files, and host the website.
I will also provide a bash script to automate deploying the website.
Note : The CLI commands mentioned will be valid for Linux, MacOS & WSL.

The Installation & Setup

Step 1 : Build a static site using HUGO

Install Dependencies : Git & Go
Install Hugo

1
2
3

## Verify if Hugo works

hugo version

Create a new site :

## Generate a new site directory (First navigate to a directory where site should be stored)

hugo new site {websitename}
cd {websitename}

Initialize a git repository (Make sure you are in your Hugo website directory)

git init

# Set global username and email parameters for git

git config --global user.name "{Your Name}"
git config --global user.email "{your.email@example.com}"

Install a Hugo theme : Find a theme.
- Follow the theme instructions on how to download. The BEST option is to install as a git submodule. Once downloaded it should be in your Hugo themes folder

`1`	`git submodule add https://github.com/{theme_repo} themes/{theme_name}`

Adjust Hugo Theme settings :
- Every theme has a different setup configuration. Refer to the theme’s page for specific documentation.
- To set the theme, edit the hugo.toml present in the site directory, and add theme = "{theme_folder_name}". Also edit the other settings in the toml file accordingly.
- Most themes will also have an example configuration or an exampleSite folder. Copy over the contents to the Hugo site directory. This is usually the best way to make sure Hugo works well and out of the box.
- The exampleSite should also contain sample posts, which should be used for reference while creating new posts. The example posts will be present in a “post/posts” folder, under the content folder, in the site directory.
Test Hugo site

`1`	`hugo server -t {themename}`

Step 2 : Push the Website code to GitHub

Create a new private repository on Github with the name for your website
Add your SSH public key to your Github account, to push the site to the remote repo just created. Official Documentation here
Push site to GitHub :
- Navigate to the Hugo site directory

1
2
3

## Connect the remote repo to the local repo

git remote add origin git@github.com:{github_username}/{repository_name}.git

## Push site code

git add .
git commit -m "{commit message}"
git push -u origin main

Step 3 : Host the site using Cloudflare Pages

SignIn/SignUp at Cloudflare Pages
Select Pages, and connect to github account
Select the Hugo site repository and begin setup
Change the project name to desired site name, and select the production branch as main
Under Build Settings, change the Framework preset to Hugo
Also add an Environment Variable : HUGO_VERSION = v{your.hugo.version}+extended
- (use command hugo version) in terminal to find your Hugo version)
Finally, Save and Deploy. Let it deploy the site, then Continue to project. Now you can Visit the site. It should look exactly the same as it did when deployed locally.
Some other settings to change in the project would be, if you want to add Custom domains, or turn on web analytics. You can also adjust other Settings as per requirement.
By default, Cloudflare Pages will automatically Deploy the newest version of the site, whenever changes are made and pushed to the GitHub site repository.

Step 4 : Creating posts in Obsidian

Get Obsidian here, or any other markdown editor will work aswell.
Create a vault in Obsidian (if not done already), to store markdown files, and make a folder called ‘Posts’, where the blog posts will be stored.
Also, right click on the Posts folder and reveal its location. This directory will be required later.
- Note : Any notes/posts can be created in obsidian, but only ones moved to the Posts folder will be published.
The posts in the “Posts” folder should be made following the layout of the “post/posts” folder under content directory of the site, as per the Theme used for proper working. Other assets (such as images) should also be stored as per the Theme’s layout. Use the exampleSite for reference.
Some settings you should change in Obsidian would be :
- Turn off Use Wikilinks & set New link format to relative path to file
- Also set Default location for new attachments, according to the Theme
To create new posts, take reference from sample post. It often provides guides to syntax and layout structure for the posts.
Also add some Frontmatter in the beginning of file. A Template can also be created in Obsidian to create each new post with frontmatter automatically. Example :

---
title: blogtitle
date: 20245-1-10
draft: false
tags:
  - tag1
  - tag2
---

After making the Posts, rsync command will be used to sync the contents of the posts folder in Obsidian and the Hugo site.

`1`	`rsync -av --delete "/path/to/obsidian/posts/" "/path/to/hugo/posts/"`

Note : A lot of useful plugins are available for Obsidian to make Markdown easier. A few I would recommend are : Editing toolbar, Paste URL into selection, Templater

The Workflow

All we have to do is, create posts in the Posts folder of the Obsidian Vault, in the format according to the Hugo Theme. Then run this automatic deployment script below.
Make sure to change the paths in the script according to your configuration.

#!/bin/bash

cd "/path/to/site/directory"

read -p "Enter commit message: " message

rsync -av --delete "/path/to/obsidian/posts/" "/path/to/hugo/posts/"

git add .
git commit -m "$message"
git push origin main

printf "\n \e[32m	Webiste Deployed!	\e[39m \n"

Just copy this and create a bash script (with extension ‘.sh’), in your site directory.
To Deploy changes in the site, to Cloudflare Pages, just run this bash script : bash {script_name}.sh

Posts on Sanyam Chhabra's Blog

Building a Home Server

Introduction

Requirements

Proxmox Installation

Proxmox Setup

Post Install Setup

Proxmox VE Helper-Scripts

PVE Post Install Script

Making the default bridge VLAN-aware

Setting up ssl certificates for Proxmox

Setting up Two-Factor Authentication

Bonus (For Laptops)

Prevent Laptop from going to sleep on closing the Lid

Screensaver (Turn off display on idle)

Setting up Storage

Directory

LVM

LVM-thin

ZFS

Ceph

Configuring Drives

Bonus (For advanced use cases)

Creating a Cluster (Optional)

Setting up the Cluster

High Availability and Replication

Bonus Tip

Setting up VMs and LXCs

Virtual Machines

VirtIO Drivers

LXC

Bonus Tip

Bonus

Windows on Proxmox

Proxmox Networking

SDN

Setup Email alerts

PCIe Passthrough

Remote Access

Cloudflare Tunnels

VPNs

Building a Voice Assistant

Introduction

Setting up Home Assistant

Setting up Voice Pipeline

Training a custom voice clone model

Step 1: Prerequisites

Step 2: Data Prep

Record your voice using the Piper Recording Studio : https://github.com/rhasspy/piper-recording-studio.git

Pull audio data from YouTube videos :

Step 3: Training Prep

Step 4: Training the model

Testing the Voice Model

Using the Voice Model for HA voice assistant

Self-Hosting an AI server

Introduction

Preparing the Server (setting up dependencies)

Installation of Ubuntu Server

Post-Install Setup

Setting up GPU

Docker installation & setup (for running containers)

Install Portainer (GUI for managing Docker Containers)

Installation & Setup of AI tools

Ollama (for local LLMs)

LiteLLM

Automatic1111 (Stable Diffusion: for Local Image Generation)

Open WebUI (for an interactive GUI)

Building a Blog Website for free

Introduction

The Installation & Setup

Step 1 : Build a static site using HUGO

Step 2 : Push the Website code to GitHub

Step 3 : Host the site using Cloudflare Pages

Step 4 : Creating posts in Obsidian

The Workflow