MAC v0.0 — MBM AI Cloud

MAC v0.0 — MBM AI Cloud

MBM Engineering College's self-hosted AI inference platform.

Run local LLMs, speech-to-text, and AI tools on your own hardware — no internet or external API keys required.

---

What's included

• FastAPI backend — OpenAI-compatible /v1/chat/completions API
• vLLM inference — Qwen2.5-7B-Instruct-AWQ (RTX 3060 12GB, ~5GB VRAM)
• Whisper STT — Speech-to-text via faster-whisper (CPU)
• PostgreSQL + Redis — Persistent storage and rate limiting
• Qdrant — Vector database for RAG (retrieval-augmented generation)
• SearXNG — Self-hosted web search
• Nginx — Reverse proxy with SSL termination
• pgAdmin — Database admin UI

Features

• Role-based access: Admin / Faculty / Student
• Copy-check (plagiarism detection) with PDF/image upload
• Attendance management with QR codes
• Doubt tracking and notifications
• File sharing
• Jupyter notebooks with live kernels
• RAG document ingestion
• Web search integration
• Agent mode with tool use
• PWA (installable on mobile)
• i18n (English + Hindi)

Windows Installer

MAC-v0.0-Setup.exe sets up the full stack on a Windows PC:

1. Run the installer as Administrator
2. Choose HOST role (full server) or WORKER role (GPU node)
3. Hardware scan runs automatically
4. SSL certificate is generated for LAN access
5. Firewall rules are configured
6. Services start via Docker Compose

Requirements: Windows 10/11, Docker Desktop, NVIDIA GPU (12GB+ VRAM recommended)

Docker image

docker pull ghcr.io/mbmuniversity2026/mac:v0.0

Default credentials (dev/test)

Role	Username	Password
Admin	abhisek.cse@mbm.ac.in	Admin@1234
Faculty	raj.cse@mbm.ac.in	Faculty@1234
Student	21CS045	Student@1234

Change all passwords before production use.