AIOPS — Local AI Operations Server

   ____  _   _   _    _   _ _____ ___   ____  ___  ____
  / __ \| | | | / \  | \ | |_   _/ _ \ / ___|/ _ \/ ___|
 | |  | | | | |/ _ \ |  \| | | || | | | |   | | | \___ \
 | |__| | |_| / ___ \| |\  | | || |_| | |___| |_| |___) |
  \___\_\\___/_/   \_\_| \_| |_| \___/ \____|\___/|____/

  Local AI Operations Server  ·  Quantocos AI Labs  ·  v5.3.0

A single-command installer that turns any Ubuntu machine or WSL2 environment into a fully operational local AI stack. Chat UI, workflow automation, vector database, agent framework, and optional GTM tooling — all running on your own hardware with no Docker and no cloud dependency.

All configuration lives in one file. All ports are changeable. Works on any username, any machine, any network.

---

Quick Start

Linux / WSL2

bash <(curl -fsSL https://raw.githubusercontent.com/quantocos/AIOps/main/aiops.sh)

Windows (fresh machine)

1. Download aiops-windows.bat
2. Right-click → Run as administrator
3. Installs WSL2 + Ubuntu 24.04, configures .wslconfig, sets up port forwarding, then launches aiops.sh inside Ubuntu automatically

---

What Gets Installed

Core Stack — Part 1 (automatic)

Service	Version	Default Port	Purpose
Ollama	latest	11434	Local LLM inference
OpenWebUI	0.8.10	8080	Chat UI → Ollama
n8n	latest	5678	Workflow automation
Qdrant	v1.17.0	6333	Vector database + dashboard
CrewAI Studio	latest	8501	Visual agent builder
PM2	latest	—	Process manager
Caddy	v2.11+	80	Reverse proxy + LAN routing
Avahi	—	—	mDNS .local hostname on LAN
pnpm	latest	—	Required for monorepo addons

Python venvs (isolated per tool): CrewAI 1.10.1 · Aider 0.86.2 · Open Interpreter 0.4.3 · Scrapy · Playwright

All default ports are defined in ~/aiops-server/aiops.conf and can be changed without reinstalling.

---

Addons — Part 2 (interactive menu)

Key	Service	Default Port	Purpose
d	Shared deps	—	PostgreSQL · Redis · Firecrawl venv
1	Twenty CRM	3000	Lead and deal management
2	Listmonk	9000	Email campaigns (single binary)
3	OpenFang	—	AI agent Hands (Lead · Browser · Researcher · Twitter)
4	Mautic	8100	Full marketing automation
5	Chatwoot	3100	Unified inbox
6	Cal.com	3002	Booking and scheduling
7	htop + nvtop	—	CPU and GPU/VRAM monitoring
8	Netdata	19999	Full system dashboard
a	Core GTM bundle	—	d + 1 + 2 + 3
b	Full GTM bundle	—	d + 1 + 2 + 3 + 4 + 5 + 6
m	All monitors	—	7 + 8

Multiple addons can be installed in one step by space-separating choices: d 1 2 3

---

Requirements

Linux / WSL2

	Minimum	Recommended
OS	Ubuntu 22.04	Ubuntu 24.04 LTS
RAM	8GB	16GB+
Disk	20GB free	100GB free
GPU	None (CPU inference)	NVIDIA with 8GB+ VRAM

Windows (for aiops-windows.bat)

• Windows 10 Build 19041+ or Windows 11
• Virtualisation enabled in BIOS (Intel VT-x / AMD-V)
• Administrator rights

---

Configuration

All runtime settings are stored in a single file generated at install time:

~/aiops-server/aiops.conf

What it contains

AIOPS_DOMAIN="myserver.local"    # your chosen .local hostname

AIOPS_PORT_OPENWEBUI=8080        # change any port here
AIOPS_PORT_N8N=5678
AIOPS_PORT_QDRANT=6333
AIOPS_PORT_CREWAI=8501
# ... all service ports

AIOPS_SCRIPTS_DIR="~/scripts"
AIOPS_VENVS_DIR="~/.venvs"
AIOPS_QDRANT_DIR="~/qdrant-data"

Changing ports or domain

ai-config          # opens aiops.conf in nano
# edit the values
pm2 restart all    # launchers re-read config on next start

No reinstall needed. Every launcher sources this file at runtime — nothing is baked into script text.

---

Models

Pull models after install based on available VRAM:

# Embeddings — pull this regardless of GPU
ollama pull nomic-embed-text

# 4–6GB VRAM or CPU-only
ollama pull qwen3:4b             # fast general chat
ollama pull qwen2.5:7b           # content, email, general
ollama pull qwen2.5-coder:7b     # coding
ollama pull deepseek-r1:8b       # reasoning
ollama pull llama3.1:8b          # agent tool use

# 10GB+ VRAM
ollama pull qwen2.5:14b
ollama pull qwen2.5-coder:14b

List available models: ollama list

---

Shell Commands

After source ~/.bashrc:

# Service management
ai-status                # PM2 process list
ai-start                 # resurrect all saved processes
ai-stop                  # stop all
ai-restart               # restart all
ai-logs                  # live log tailing
ai-urls                  # print current service URLs (resolved at runtime)
ai-config                # edit ~/aiops-server/aiops.conf
aiops-caddy-regen        # regenerate Caddyfile after port changes

# AI tools
aider                    # code assistant
crew <script.py>         # run a CrewAI crew
interpreter              # Open Interpreter session
scrape <script.py>       # Scrapy runner
automate <script.py>     # Playwright runner

# Chat
chat                     # ollama run qwen3:4b
chat-coder               # ollama run qwen2.5-coder:7b
chat-reason              # ollama run deepseek-r1:8b
models                   # ollama list

---

Directory Structure

~/
├── aiops-server/
│   ├── aiops.conf           # single source of truth for all config
│   ├── install.log
│   └── addons.log
├── agents/
│   ├── crews/               # CrewAI scripts
│   ├── tasks/
│   ├── tools/
│   ├── outputs/
│   └── configs/
├── scripts/
│   ├── run-openwebui.sh     # all launchers source aiops.conf at runtime
│   ├── run-n8n.sh
│   ├── run-qdrant.sh
│   ├── run-crewai-studio.sh
│   ├── run-aider.sh
│   ├── run-crew.sh
│   ├── run-interpreter.sh
│   ├── run-scrapy.sh
│   └── run-playwright.sh
├── qdrant-data/
│   ├── static/              # Qdrant web UI
│   └── storage/             # vector data
├── .venvs/
│   ├── crewai/
│   ├── aider/
│   ├── interpreter/
│   ├── scrapy/
│   └── playwright/
└── CrewAI-Studio/

---

Architecture

Routing

Caddy listens on :80 and routes by path prefix. Every service runs on its own port — no subpath proxying for WebSocket-heavy applications.

Browser / LAN device
        │
   Caddy :80
        │
   /n8n*     → strip /n8n    → :PORT_N8N      (WebSocket: Connection "Upgrade")
   /agents*  → strip /agents → :PORT_CREWAI   (WebSocket: Connection "Upgrade")
   /qdrant*  → strip /qdrant → :PORT_QDRANT   (REST API only via Caddy)
   /monitor* → strip         → :PORT_NETDATA
   /ollama*  → strip         → :PORT_OLLAMA
   /*                        → :PORT_OPENWEBUI

   Direct port access (bypass Caddy):
   :PORT_QDRANT    → Qdrant UI + API (SPA served at root)
   :PORT_OLLAMA    → Ollama API
   :PORT_TWENTY    → Twenty CRM
   :PORT_LISTMONK  → Listmonk
   :PORT_MAUTIC    → Mautic
   :PORT_CHATWOOT  → Chatwoot
   :PORT_CALCOM    → Cal.com
   :PORT_NETDATA   → Netdata

All PORT_* values come from aiops.conf. After changing ports, run aiops-caddy-regen to regenerate the Caddyfile.

WebSocket

Caddy uses Connection "Upgrade" as a literal string for all WebSocket-capable routes. The {http.headers.Connection} placeholder does not reliably forward upgrade requests and causes white screens on n8n and Streamlit.

Streamlit (CrewAI Studio)

Launched with --server.baseUrlPath /agents so Streamlit generates WebSocket URLs at /agents/_stcore/stream. Caddy strips the /agents prefix before proxying, so Streamlit sees requests at its own root.

Config-file model

No IPs, domain names, or usernames are baked into launcher scripts. Every launcher does:

source ~/aiops-server/aiops.conf
exec service --port ${AIOPS_PORT_SERVICE}

IP is never stored anywhere — it is discovered at display time via ip route get 1. The stack works correctly after reboots, network changes, and on any machine regardless of username or DHCP assignment.

---

Troubleshooting

Services not starting

ai-status                         # check which processes are errored
pm2 logs <service> --lines 50     # read the actual error

White screen on n8n or CrewAI Studio

Check Caddy and restart:

sudo caddy validate --config /etc/caddy/Caddyfile
sudo systemctl restart caddy
pm2 restart n8n crewai-studio

If you changed ports in aiops.conf without regenerating the Caddyfile:

aiops-caddy-regen

OpenWebUI not starting

pm2 logs openwebui --lines 30
# Binary missing — reinstall:
pip install open-webui==0.8.10 --break-system-packages
pm2 restart openwebui

Ollama not responding

sudo systemctl status ollama
sudo systemctl restart ollama
curl http://localhost:11434/api/tags

Port conflict

sudo fuser -k <PORT>/tcp    # free the port
pm2 restart <service>
# Or change the port in aiops.conf and restart

Twenty CRM or Cal.com install errors

Both are pnpm monorepos. If you see ERESOLVE or workspace:* errors:

cd ~/twenty && pnpm install && pm2 restart twenty
cd ~/calcom  && pnpm install && pm2 restart calcom

Chatwoot — missing gems

cd ~/chatwoot
RUBY_MINOR=$(ruby -e 'puts RUBY_VERSION.split(".")[0..1].join(".")')
export PATH="$HOME/.gem/ruby/${RUBY_MINOR}.0/bin:$PATH"
gem install bundler --user-install
bundle install
pm2 restart chatwoot chatwoot-worker

WSL2 memory pressure

Edit %USERPROFILE%\.wslconfig on Windows:

[wsl2]
memory=16GB
processors=8
swap=8GB

Then from PowerShell: wsl --shutdown, reopen Ubuntu.

---

Version History

v5.3.0 — Current

• Architecture: Config-file model — all ports, domain, and paths live in ~/aiops-server/aiops.conf; every launcher sources it at runtime; nothing hardcoded into script text
• Added: ai-config, ai-urls, aiops-caddy-regen aliases
• Added: write_caddyfile() callable standalone to regenerate Caddy config after port changes
• Fixed: IP never stored; discovered at runtime via ip route get 1

v5.2.0

• Fixed: OpenWebUI binary discovery fully runtime — find_owu() function in launcher searches all known pip locations; no install-time path resolution

v5.1.0

• Fixed: Per-port architecture — WebSocket apps no longer white-screened by subpath proxy
• Fixed: n8n, CrewAI Studio — Connection "Upgrade" literal + correct strip_prefix
• Fixed: Qdrant UI — SPA served at root on its own port
• Fixed: Twenty CRM / Cal.com — pnpm install (both are pnpm monorepos; npm fails)
• Fixed: Chatwoot — gem install bundler --user-install for system Ruby on Ubuntu 24.04
• Fixed: OpenFang — PATH exported in-session so hand activation runs without shell reload
• Fixed: Mautic — PHP routes through public/index.php Symfony front controller
• Added: pnpm to core stack; aiops-windows.bat Windows bootstrap

v5.0.0

• Initial release — core stack + GTM addon suite

---

License

MIT — see LICENSE

---

Quantocos AI Labs
"Build with intelligence. Operate with precision."