Derived from this repository's documentation.
Arabic ITSM Server is a FastAPI inference and demo layer for classifying Arabic IT support tickets, with support for multiple model checkpoints and dynamic task heads.
Core Scope
The project provides a model catalog, dedicated page per model, dashboard analytics, monitoring, and export utilities, while exposing API endpoints for single-model and multi-model inference.
What You Can Do
- Browse discovered checkpoints in the catalog and open any model-specific demo page.
- Run classification with one model (`/api/classify`) or compare all loaded models (`/api/classify/all`).
- Track usage metrics, recent predictions, and feedback from the dashboard.
- Monitor CPU, memory, uptime, and visitor activity from the monitoring page.
- Export data as CSV/ZIP or SQL backups for reporting and recovery.
Deployment & Reproducibility Notes
- Cloud deployment options in docs include Hugging Face Spaces, ngrok demos, Modal, Render, and Coolify VPS.
- The server records model IDs and supports per-model workflows for clearer academic comparisons.
- Tokenizer consistency safeguards were added to reduce inference mismatch risks across environments.