Pre-Installation System Check

Before we book your Weld AI installation call, please complete the three steps below. The installation call itself is short and straightforward — but only once your system is confirmed ready in advance.

⚠️

Prerequisites must be in place before we book your installation call We cannot begin installation until your system passes the automated check. For Windows PC and Windows Server Direct installs, this includes WSL2 being fully configured and working. If IT prerequisites are not in place on the day of the call, we will need to reschedule.

ℹ️

Complete both the script and the form below Email the diagnostics zip and your answers to hello@weldai.uk with subject: Pre-Install — [your company name].

1 Choose Your Install Type

Weld AI supports several installation configurations depending on your server setup. Review the options below and identify which best matches your situation — include this in your email.

Not sure which applies to you? Email us at hello@weldai.uk and we'll advise.

Windows PC or Workstation

Most common

Weld AI installed directly on a Windows 10 or 11 PC or workstation. Documents stored in a local folder on the same machine.

Windows 10 (21H2+) or Windows 11
8GB RAM minimum, 16GB recommended
Documents on a local drive (C:, D: etc.)
Docker Desktop + WSL2

Windows Server + Ubuntu VM

Recommended for servers

Your IT team creates a small Ubuntu Linux virtual machine (VM) on your Windows Server using Hyper-V — Windows Server's built-in virtualisation platform. Weld AI runs inside the Ubuntu VM, completely isolated from your main server. This is the recommended approach for Windows Server as it avoids WSL2 complexity and supports network shares.

Windows Server 2022 with Hyper-V role enabled
16GB+ RAM on the server (4GB allocated to Ubuntu VM)
40GB spare disk space for the VM
IT creates the Ubuntu VM; we install Weld AI remotely
Full network share (Samba) support

Linux Server

Linux

Weld AI installed directly on an Ubuntu or Debian Linux server. Simplest possible install — Docker works natively with no virtualisation layer needed.

Ubuntu 22.04 LTS or Debian 12
8GB RAM minimum
Docker Engine (no WSL required)
Full network share support via NFS/Samba

Windows Server (Direct)

Alternative only

Weld AI installed directly on Windows Server 2022 without a Linux VM. Only recommended when Hyper-V is unavailable or IT policy prevents running Linux. Requires complex WSL2 configuration.

Windows Server 2022 Desktop Experience only
16GB RAM minimum
WSL2 requires significant manual IT setup
Network shares not supported
Use Ubuntu VM approach if at all possible

💡

What is Hyper-V? Hyper-V is Microsoft's built-in virtualisation platform — included free with Windows Server 2022. It allows multiple virtual machines (VMs) to run on a single physical server, each completely isolated from each other. Most Windows Server 2022 installations have Hyper-V available. Your IT team can check by opening Server Manager and looking for the Hyper-V role, or by running Get-WindowsFeature Hyper-V in PowerShell.

⚠️

Windows Server Direct — Desktop Experience required If using the Windows Server Direct option, the server must be running Windows Server 2022 with Desktop Experience (full graphical interface). Windows Server Core (command-line only) is not supported.

2 System Requirements

Requirement	Windows PC	Win Server + Ubuntu VM	Win Server Direct	Linux Server
OS	Windows 10 (21H2+) or 11	Windows Server 2022 + Ubuntu 22.04 VM	Windows Server 2022 Desktop Experience	Ubuntu 22.04 LTS or Debian 12
RAM	8GB min, 16GB rec.	16GB min (4GB for VM)	16GB min	8GB min, 16GB rec.
Disk	20GB free	60GB free (VM + app)	20GB free	20GB free
Virtualisation	WSL2 + VM Platform	Hyper-V on host	WSL2 (complex setup)	Not required
Docker	Docker Desktop	Docker Engine (in VM)	Docker Desktop	Docker Engine
Network shares	❌ Not supported	✅ Samba/NFS	❌ Not supported	✅ Samba/NFS
Internet	Outbound HTTPS (port 443) to api.anthropic.com — required for AI queries.
Network ports	80 and 443 inbound — only needed if accessed from other PCs on the network.

2b How Many Documents Can Weld AI Handle?

Weld AI can index any number of documents — the practical limit depends on your server's RAM and storage. Use the table below to check your setup is appropriate for your document collection.

Document collection	Min RAM	Recommended RAM	Initial index time*
Small — up to 50 documents	8GB	8GB	Under 5 minutes
Medium — 50–200 documents	8GB	16GB	5–20 minutes
Large — 200–500 documents	16GB	16GB+	20–60 minutes
Very large — 500+ documents	Contact us before purchase — we will advise on hardware and configuration.

* Estimated times for typical mixed documents on a 16GB RAM server with local storage. Actual times depend heavily on hardware — see measured results below.

Scenario	Documents	Chunks	Time (constrained)
First full build (incl. model download)	5 large PDFs	1,375	~18 minutes
Subsequent full rebuild	5 large PDFs	1,375	~17 minutes
Incremental update (1 new file added)	1 of 6	275	~3 minutes

3 Where Are Your Documents?

Understanding where your documents are stored determines the best install type. Weld AI reads documents from a folder on the server — the table below shows what is and isn't supported.

Document Location	Windows PC / Server (Direct)	Windows Server + Ubuntu VM	Linux Server
Local drive (C:\, D:\, E:\ etc.)	✅ Supported	✅ Supported	✅ Supported
USB / external drive	✅ Supported — must stay connected	✅ Supported	✅ Supported
OneDrive / SharePoint synced folder	✅ Supported — files must be set to "Always keep on this device"	⚠️ Complex	⚠️ Complex
Network mapped drive (e.g. U:\)	❌ Not supported	✅ Supported via Samba — domain credentials needed for corporate networks	✅ Supported via NFS/Samba — domain credentials needed for corporate networks
UNC path (\\server\share)	❌ Not supported	✅ Supported via Samba — domain credentials needed for corporate networks	✅ Supported via NFS/Samba — domain credentials needed for corporate networks
File server (documents only on server)	⚠️ Copy or sync needed	✅ Best option — install Weld AI on the file server	✅ Supported

⚠️

Network mapped drives on Windows — not supported Docker on Windows cannot access network mapped drives (e.g. U:\docs) or UNC paths (e.g. \\server\documents). This is a fundamental Docker on Windows limitation. If your documents are on a file server, the recommended solution is the Windows Server + Ubuntu VM install type — Ubuntu can mount Windows file server shares directly via Samba, giving Weld AI full access to your documents without copying. We will advise on the best approach during your pre-installation review.

✅

Windows Server + Ubuntu VM — full network share support (tested) The Ubuntu VM can mount Windows file server shares directly using Samba. Weld AI indexes documents live from the share — no copying or syncing required. Tested with corporate domain credentials. This is the recommended install type for customers whose documents are on a file server.

☁️

OneDrive / SharePoint users If your documents are in a OneDrive or SharePoint synced folder, you must ensure the folder is set to "Always keep on this device" before indexing. Cloud-only files (shown with a ☁️ icon in File Explorer) are not stored locally and cannot be indexed.

To fix: right-click your documents folder in File Explorer → Always keep on this device. All files will show a ✅ green tick when ready.

🔄

Documents on a file server — Windows PC workaround If you have a Windows PC install but documents are on a file server, they can be synced automatically using a scheduled robocopy task:

robocopy \\server\share C:\WeldAI\docs /MIR /Z

This mirrors your network share to a local folder nightly. Weld AI then indexes from the local copy. We can set this up during installation.

Better alternative: if you have a server, the Windows Server + Ubuntu VM install type gives live access to network shares with no copying needed.

4 Run the Automated System Check

The script checks your system automatically and saves a zip file to your Desktop. Takes about 30 seconds. Run this on the machine where Weld AI will be installed.

⬇ Download weldai-diagnostics.ps1

PowerShell script · No installation · No changes made to your system

🛡️

Your browser may warn "this file could harm your device" — this is normal and safe to dismiss Windows and browsers flag all PowerShell scripts downloaded from the web, regardless of their content. This script only reads information from your system and makes no changes. Click Keep (Edge/Chrome) or Keep anyway to proceed with the download.

Save to your Desktop

Download weldai-diagnostics.ps1 and save it to your Desktop. If your browser warns the file could be harmful, click Keep to proceed — this warning appears for all PowerShell scripts.

Right-click → Run with PowerShell

Right-click the file and select Run with PowerShell. If asked about execution policy, type R and press Enter. No administrator rights needed.

Email the zip file

A file called weldai-diagnostics.zip will appear on your Desktop. Email it to hello@weldai.uk with subject: Pre-Install — [your company name].

5 IT Team Checklist

The automated script cannot check these items without Administrator rights. Please ask your IT contact to confirm the following and include their answers in your email.

🖥️ All install types

🔌
Static IP address Recommended The server should have a fixed IP address so staff can always reach it. IT can assign a static IP or create a DHCP reservation. The current IP will appear in the diagnostics script output — please confirm whether it is static or DHCP.
🔥
Firewall rules — ports 80 and 443 Required for multi-user Required if multiple staff will access Weld AI from their own PCs. IT needs to open inbound TCP ports 80 and 443 on the server's Windows Firewall. Single-user installs on a local PC can skip this.
🌐
Outbound internet access Required The server needs outbound HTTPS (port 443) access to api.anthropic.com. Required during installation and for every AI query. If a corporate proxy is in place, IT will need to whitelist this domain.

🪟 Windows PC / Windows Server (Direct) — additional checks

🐧
WSL2 confirmed working Must be done before install call WSL2 must be fully installed and working before we book the installation call. To verify: open PowerShell as Administrator and run wsl --status. It must show Default Version: 2 and have a Linux distribution installed (e.g. Ubuntu).

To test fully: run wsl — you should get a Linux prompt. Type exit to return to PowerShell.

Windows Server note: WSL2 on Windows Server requires running Windows Update to completion first (30–60 minutes, multiple reboots). This must be done by IT in advance — not during the installation call.

→ Full step-by-step WSL2 setup instructions in the IT Setup Guide
⚙️
Virtual Machine Platform enabled Required Required for Docker. To check: open PowerShell as Administrator and run:
Get-WindowsOptionalFeature -Online -FeatureName VirtualMachinePlatform
Should show State: Enabled.
🐳
Docker service startup type Recommended Docker must start automatically after a reboot. To check and fix: open PowerShell as Administrator and run:
Set-Service -Name "com.docker.service" -StartupType Automatic
🖥️
Windows Server — Desktop Experience edition Required Windows Server Core (no GUI) is not currently supported. The server must be running Windows Server 2022 with Desktop Experience (full graphical interface).

🖥️🐧 Windows Server + Ubuntu VM — additional checks

📦
Hyper-V enabled on Windows Server Required Hyper-V must be enabled as a server role. To check: open Server Manager and look for Hyper-V in the roles list, or run Get-WindowsFeature Hyper-V in PowerShell. IT will need to create a Ubuntu 22.04 VM (4GB RAM, 40GB disk) before the installation call.

→ Step-by-step Hyper-V VM setup instructions in the IT Setup Guide
💾
Available RAM and disk Required The Ubuntu VM needs 4GB RAM and 40GB disk. Please confirm the host server has sufficient spare resources — check via Task Manager or Get-CimInstance Win32_OperatingSystem.
🔗
Network share access from Ubuntu VM If using network shares If documents are on a Windows file server, Ubuntu can mount the share directly via Samba — no copying needed. IT will need to confirm:
- The UNC share path (e.g. \\fileserver\documents)
- A Windows username and password with read access
- Whether the account is a domain account (e.g. CORP\username) or local account
- That the Ubuntu VM can reach the file server on the network (same subnet or firewall rule allowing TCP 445)
We will configure and test the Samba mount during the installation call. See IT Setup Guide for details.

✓ Before You Book

Please confirm all of the following before emailing us to book your installation call. Include this confirmation in your email.

🖥️
Diagnostics script run and zip file ready to attach
The zip file from the automated script (Section 4) must be attached to your email.
🌐
Outbound internet access confirmed
The server can reach api.anthropic.com on port 443 — confirmed in the diagnostics output.
🔑
Anthropic API key obtained
API key created at console.anthropic.com with a payment method added.
🐧
WSL2 working — Windows PC and Windows Server Direct installs only
wsl --status shows Default Version: 2 and running wsl gives a Linux prompt. Not required for Ubuntu VM or Linux installs.
📂
Documents folder identified and accessible
You know where your documents are stored and the path is supported (see Section 3).
👤
Administrator access available for the installation call
Someone with Administrator rights will be available on the day, or screen-sharing access arranged.

✅

Once all boxes are ticked — email us to book Send your diagnostics zip and completed form below to hello@weldai.uk with subject: Pre-Install — [your company name].

6 Installation Information

Please provide the following information in your email. This allows us to prepare your configuration in advance so the installation call is as quick as possible.

📋 Please include in your email to hello@weldai.uk

Install type

Select the option that best matches your setup from Section 1 above.

Anthropic API key *

Weld AI uses the Anthropic Claude API to answer questions from your documents. You will need an API key before installation.

How to get your API key:

Go to console.anthropic.com and create an account (or sign in).
Go to API Keys and create a new key.
Copy the key — you will only see it once. Keep it safe.
Add a payment method under Billing — usage costs are typically a fraction of a penny per query.

🛡️

Data protection note When you create an Anthropic API account and accept their Commercial Terms, Anthropic's Data Processing Agreement (DPA) is automatically included. This DPA covers UK data transfers via the UK International Data Transfer Agreement (IDTA). Anthropic does not use API data to train its models. If your firm has a Data Protection Officer or compliance requirements, you may wish to share Anthropic's DPA with them before installation.

Company name (for branding)

The name shown in the app header.

Number of staff who will use the system

Approximate number — helps us advise on hardware and access configuration.

Documents folder path

The full path to the folder containing your company documents. See Section 3 for supported locations.

Approximate number of documents

4 suggested questions (optional)

The app shows up to 4 suggested question buttons on the home screen. Leave blank and we will suggest some based on your documents.

Any known IT constraints

✉️

Email your answers + diagnostics zip to hello@weldai.uk Subject: Pre-Install — [your company name]. Not sure about any of these? Don't worry — email what you can and we will work through the rest on a quick pre-call.