The current version is #ident "@(#)$Format:LocalFoodAI_lanfr144:INSTALL_WSL.md:%an:%ae:%ad:%cn:%ce:%cd:%H:%D:%N$"

🚀 WSL2 Port-Shifted Installation Guide (Local Food AI)

This guide provides step-by-step instructions to install and run the Local Food AI system on Windows Subsystem for Linux (WSL2).

To prevent port conflicts with standard local services (such as existing MySQL databases, custom Nginx web ports, Zabbix suites, or Airflow installations), all public host ports in this deployment have been increased by +20 from their original defaults.

📊 Port Mapping Reference

Service Container	Original Port	Shifted Port (+20)	Type / Purpose
Nginx Proxy	`80`	`100` (or `8020`)	Main Web Gateway (Reverse Proxy)
Streamlit Application	`8502`	`8522`	Direct Streamlit web port
MySQL Database	`3306`	`3326`	Database external connection listener
SearXNG Search	`8085`	`8105`	Anonymous meta-search gateway
Zabbix Web UI	`8081` / `8444`	`8101` / `8464`	Monitoring dashboard (HTTP / HTTPS)
Zabbix Server Daemon	`10051`	`10071`	Active telemetry monitoring trap listener
Airflow Webserver	`8082`	`8102`	Airflow data workflow manager

🛠️ Step-by-Step Installation Runbook

Step 1: Open Your WSL2 Ubuntu Terminal

Ensure you have WSL2 enabled and are using an Ubuntu 24.04 shell instance.

Step 2: Clone the Git Repository

Run the following commands inside your WSL Ubuntu home directory to clone the project:

Primary Repository (Internal Network):

git clone https://git.btshub.lu/lanfr/LocalFoodAI_lanfr144.git
cd LocalFoodAI_lanfr144

Alternative Repository (Worldwide Access - Clone of the Primary):

git clone https://github.com/lanfr144/LocalFoodAI_lanfr144.git
cd LocalFoodAI_lanfr144

Step 3: Setup Local Environment Variables

Create the required .env file at the root of the repository to feed standard local credentials to the containers: Configure your database credentials, active network mode, and the target model name in a .env file at the root of the repository. A generic template is provided below:

# NETWORK_MODE: local (offline) or server (online)
NETWORK_MODE=local
LLM_MODEL=llama3.2:3b

# DATABASE CREDENTIALS (MySQL)
MYSQL_ROOT_PASSWORD=your_secure_root_password
DB_READER_PASS=your_secure_reader_password
DB_LOADER_PASS=your_secure_loader_password
DB_APP_AUTH_PASS=your_secure_auth_password
MYSQL_ZABBIX_PASSWORD=your_secure_zabbix_password

# ZABBIX & SNMP CREDENTIALS
ZABBIX_URL=your_zabbix_url
ZABBIX_USER=your_zabbix_user
ZABBIX_PASS=your_zabbix_pass
ZABBIX_SNMP_USER=your_snmp_user
ZABBIX_SNMP_AUTHKEY=your_snmp_authkey
ZABBIX_SNMP_PRIVKEY=your_snmp_privkey
DISCORD_WEBHOOK=your_discord_webhook

# EMAIL ALERTS CONFIGURATION
EMAIL_USER=your_email_user
EMAIL_PASS=your_email_pass

# TAIGA CREDENTIALS
TAIGA_URL=https://192.168.130.161/taiga
TAIGA_USER=your_taiga_user
TAIGA_PASS=your_taiga_password

Step 4: Launch the Docker Container Stack

Deploy the entire 10-container system in the background using the custom port-shifted WSL configuration file:

docker compose -f docker-compose-wsl.yml up -d

Step 5: Pull the Quantized Reasoning LLM Model

Download the high-capacity, reasoning-optimized local model directly inside the running Ollama container instance:

docker exec -it $(docker ps -q -f name=ollama) ollama pull $( grep '^[ \t]*LLM_MODEL[ 	]*=' .env | sed 's/^.*=//' )

Step 6: Ingest the OpenFoodFacts Database Records

Initialize the database tables and trigger the ingestion pipeline to parse local dataset records:

docker compose -f docker-compose-wsl.yml run --rm ingest python ingest_csv.py

🌐 Verifying and Accessing the Services

Once the stack is fully running, you can connect to all system components in your web browser:

🍏 Streamlit Application UI: Open http://localhost:100 (uses Nginx reverse proxy on Port 100) or bypass the proxy directly at http://localhost:8522.
📊 Zabbix Monitoring Suite: Open http://localhost:8101 (Default Credentials: Username Admin / Password zabbix).
🌀 Airflow Dag Orchestrator: Open http://localhost:8102 (Default Credentials: Username admin / Password admin).
🔍 Database Server: Connect using your preferred SQL client (DBeaver, MySQL Workbench) via Host localhost and Port 3326.

⚡ Developer Productivity & Troubleshooting

1. Dynamic LLM Pulls (Non-Interactive)

To pull updates to your reasoning models in a single line without entering an interactive shell, use:

docker exec -it $(docker ps -q -f name=ollama) ollama pull $(grep '^[ \t]*LLM_MODEL[ \t]*=' .env | cut -d'=' -f2)

2. Path Resolutions inside WSL

If you need to configure container volumes or load local files (like the datasets or PDF fonts), remember that the host's C: drive is mounted under /mnt/c/ in WSL:

Windows Path: C:/Users/lanfr144/Documents/DOPRO1/Antigravity/Food/docs/fonts/
WSL Path: /mnt/c/Users/lanfr144/Documents/DOPRO1/Antigravity/Food/docs/fonts/
Tip: Always use forward slashes / in path configurations, as they are natively supported by both PowerShell on the Windows host and Bash inside WSL.

3. Git Attributes Clean/Smudge Loops

If running a clean checkout (e.g. git checkout -f) triggers an error about a missing git-ident-filter.py script, restore the filter script first before executing the checkout:

git checkout HEAD -- local_tools/git-ident-filter.py
git checkout -f

Prepared by Francois Lange for the Local Food AI Delivery.

INSTALL_WSL.md 5.6 KB Trvalý odkaz Historie Surový