local-ai-environment

Operations Guide

Daily usage, model management, and maintenance for the Local AI Environment.

Starting the Environment

Ollama

Ollama installs a system tray application that starts automatically with Windows. If it is not running:

# Start Ollama service (runs in background, binds to localhost:11434)
ollama serve

Or double-click the Ollama icon in the Start menu. Look for the llama icon in the system tray to confirm it is running.

Verify Ollama is up:

Invoke-RestMethod http://localhost:11434
# Expected: "Ollama is running"

Open-WebUI

Open-WebUI does not run automatically. Start it using the convenience script created during installation:

cd C:\Local-AI-Environment
.\Start-OpenWebUI.ps1

Or start manually:

# Set Ollama endpoint, then start Open-WebUI
$env:OLLAMA_BASE_URL = "http://localhost:11434"
open-webui serve

Default port is 3000. To use a different port:

open-webui serve --port 3001

Open the browser at: http://localhost:3000

First Launch

On first visit to http://localhost:3000, you will be prompted to create an admin account. This account is stored locally — no cloud registration required.

Checking System Health

Run the validation script whenever something feels wrong or after updates:

.\scripts\07-Validate-Environment.ps1

This checks all 12 components and shows a table with ✅ PASS, ⚠️ WARNING, or ❌ FAIL for each.

For detailed output:

.\scripts\07-Validate-Environment.ps1 -Verbose

To skip the slow LLM inference test:

.\scripts\07-Validate-Environment.ps1 -SkipLLMTest

Results are saved to reports/validation-report.md.

Managing Models

Command Reference

Operation	Command
List installed models	`ollama list`
Download a model	`ollama pull qwen2.5:7b`
Remove a model	`ollama rm qwen2.5:7b`
Run model interactively	`ollama run qwen2.5:7b`
Show model details	`ollama show qwen2.5:7b`
Copy/rename a model	`ollama cp qwen2.5:7b my-custom-name`

Recommended Models by Use Case

Use Case	Recommended Model	Notes
General chat (fast)	`qwen2.5:1.5b`	1GB, works on low RAM
General chat (quality)	`qwen2.5:7b`	5GB, needs 16GB RAM
Coding assistant	`qwen2.5:7b` or `phi4:latest`	Strong code generation
Fast responses	`qwen2.5:0.5b`	400MB, CPU-friendly
Best reasoning	`phi4:14b`	9GB, needs 32GB RAM
Multilingual	`qwen2.5:7b`	Excellent non-English support

Downloading a New Model

# Re-run the model configuration script for guided selection
.\scripts\05-Configure-Local-LLM.ps1

# Or pull directly if you know the model tag
ollama pull llama3.2:3b

The script shows estimated download size and hardware requirements before pulling.

Interactive Model Testing

# Start a chat session with any installed model
ollama run qwen2.5:7b

# Type your message and press Enter
# Type /bye to exit

Using Open-WebUI

Web Interface

Open http://localhost:3000 in any browser.

Key features:

Model selector — dropdown at top of chat window; lists all Ollama models
New conversation — click the pencil icon or press Ctrl+Shift+O
System prompt — click the settings icon on any conversation
Conversation history — left sidebar, saved locally

Managing Models from the UI

Click your avatar (top right) → Settings
Go to Models
Under “Pull a model from Ollama.com”, type the model tag (e.g., gemma3:4b) and click the download button
Installed models appear in the model selector immediately

Open-WebUI API

Open-WebUI exposes an OpenAI-compatible REST API at http://localhost:3000/api/. This means any tool that supports OpenAI API format can use it locally:

# Example: list available models
Invoke-RestMethod http://localhost:3000/api/models

# Example: chat completion
$body = @{
    model = "qwen2.5:7b"
    messages = @(@{ role = "user"; content = "Hello" })
} | ConvertTo-Json

Invoke-RestMethod -Method POST -Uri http://localhost:3000/api/chat/completions `
    -ContentType "application/json" -Body $body

Data Storage

Open-WebUI stores conversations, settings, and user accounts in:

$env:APPDATA\open-webui\   (typical location)

Updating Components

Update Open-WebUI

pip install --upgrade open-webui

# Verify new version
open-webui --version

Restart Open-WebUI after updating.

Update Ollama

Re-run the install script — it detects the existing version and upgrades:

.\scripts\04-Install-Ollama.ps1

Or download the latest installer from https://ollama.ai/download and run it.

Update a Model

Ollama models are versioned. Pull again to get the latest:

ollama pull qwen2.5:7b   # downloads only changed layers

Update All Installed Models

# List all installed models and pull each
ollama list | Select-String -Pattern "^\S+" | ForEach-Object {
    $model = $_.Matches[0].Value
    Write-Host "Updating $model..."
    ollama pull $model
}

Stopping the Environment

Stop Open-WebUI

Press Ctrl+C in the terminal where open-webui serve is running
Or close the PowerShell/terminal window

Stop Ollama

Right-click the llama icon in the system tray → Quit Ollama
Or from PowerShell:

# If running as a service
Stop-Service Ollama -ErrorAction SilentlyContinue

# If running as a process
Get-Process | Where-Object { $_.Name -like "*ollama*" } | Stop-Process

Viewing Logs

Script Logs

Each script creates a timestamped log file:

.\logs\00-System-Check_2026-09-06_143022.log
.\logs\03-Install-OpenWebUI_2026-09-06_144501.log

Format: [YYYY-MM-DD HH:mm:ss] [LEVEL] Message

To view the latest log for a script:

Get-ChildItem .\logs\ | Sort-Object LastWriteTime -Descending | Select-Object -First 5

Open-WebUI Logs

Logs are printed to the terminal where open-webui serve is running. There is no separate log file by default.

Ollama Logs

# View Ollama log file
Get-Content "$env:LOCALAPPDATA\Ollama\ollama.log" -Tail 50

# Follow in real time
Get-Content "$env:LOCALAPPDATA\Ollama\ollama.log" -Wait -Tail 20

Re-running Scripts

All scripts are idempotent — safe to re-run at any time. They check existing state before taking action:

If a tool is already installed at the correct version → skip install
If a model is already downloaded → skip download
If configuration is already correct → skip configuration

Re-run the relevant script after any of these situations:

After a failed installation (fix the error, re-run)
After updating Windows or Python
After moving the project to a new location
When adding a new model

Reports

Reports are saved in .\reports\ after each script run:

File	Generated by
`system-report.json` / `.md`	00-System-Check
`docker-audit.json` / `.md`	01-Docker-Audit
`docker-before-clean.json`	02-Docker-Clean (FULL mode)
`docker-after-clean.json`	02-Docker-Clean
`openwebui-install-report.json` / `.md`	03-Install-OpenWebUI
`ollama-install-report.json` / `.md`	04-Install-Ollama
`llm-config-report.json` / `.md`	05-Configure-Local-LLM
`openwebui-config-report.json` / `.md`	06-Configure-OpenWebUI
`validation-report.json` / `.md`	07-Validate-Environment
`FINAL-ENVIRONMENT-REPORT.md`	99-Generate-Report

Generate a fresh final report at any time:

.\scripts\99-Generate-Report.ps1