TicketAssistant (Laravel 13 + pgvector + external Ollama)

AI-powered customer support ticket system using semantic retrieval + LLM re-ranking.

Requirements

Docker Desktop
Docker Compose v2
Ollama running on a reachable network server

Quick Start (Docker)

Create environment file

cp .env.docker.example .env

Set the external Ollama URL in .env

OLLAMA_BASE_URL=http://<OLLAMA_SERVER_IP_OR_HOSTNAME>:11434

Build and start containers

docker compose up -d --build

Install PHP dependencies in app container

docker compose exec app composer install

Generate app key and run database setup

docker compose exec app php artisan key:generate
docker compose exec app php artisan migrate --seed

API base URL: http://localhost:8090/api Admin dashboard: http://localhost:8090/admin/dashboard

Admin Dashboard (Blade + Livewire)

/admin/dashboard: KPI cards + recente tickets + recente AI decisions
/admin/articles: artikel aanmaken + artikeloverzicht
/admin/tickets: tickets met AI-classificatie-overzicht

Business logic is isolated in service classes:

App\Services\AdminDashboardService
App\Services\AdminArticleService
App\Services\AdminTicketService

Services

web: Nginx (public endpoint)
app: PHP-FPM Laravel runtime
worker: queue worker for async jobs
postgres: PostgreSQL 17 + pgvector

Notes

No external AI APIs are used.
Ollama runs outside Docker on your network.
Set QUEUE_EMBEDDINGS=true for async article embedding generation.

Helpdesk Scraper Command

Importeer artikelen vanaf InternetToday Helpdesk (categorieen + subcategorieen):

docker compose exec app php artisan helpdesk:import

Opties:

# test zonder schrijven
docker compose exec app php artisan helpdesk:import --dry-run

# verwerk alleen eerste N artikelen
docker compose exec app php artisan helpdesk:import --limit=25

# custom base-url
docker compose exec app php artisan helpdesk:import --base-url=https://www.internettoday.nl/helpdesk

2.0 KiB Raw Blame History