• One-click setup for GPU-enabled VPS (Linux)
• Production-ready: Secure, robust, and fully automated
• OpenAI-compatible API with RAG (ChromaDB + Instructor-XL)
• Telegram admin bot: Button-based control for all features
• Child bot management: Add/revoke, usage_scope, isolation
• System monitoring: Alerts, watchdog, resource stats
• Exception logging for audit/debug
• No hardcoded secrets: All config via environment variables

• Linux VPS (recommended: Ubuntu 20.04+)
• Python 3.10+
• NVIDIA GPU (for LLM acceleration)
• sudo/root access

1. Clone the repo:

   git clone https://github.com/iceyxsm/all-in-one-llm-server.git && cd all-in-one-llm-server

2. Run the setup script:

   bash setup.sh

3. Follow prompts to configure environment variables and select/download your LLM model.
4. Wait for setup to finish (NGINX, SSL, Python venv, models, etc.).
5. Start all services:

   sudo systemctl start api
   sudo systemctl start ollama
   sudo systemctl start telegram-bot

Copy .envtemplate to .env and fill in:

• MODEL_PATH, MODEL_NAME: Path/name of your GGUF model
• SUPABASE_URL, SUPABASE_SERVICE_KEY: Your Supabase project
• TELEGRAM_BOT_TOKEN, TELEGRAM_OWNER_ID: For admin bot
• LLM_API_URL: (optional) Override API endpoint

• POST /v1/chat/completions (OpenAI-compatible)
• POST /upload (file upload for RAG)
• All endpoints require a valid API key (managed via Supabase)

• Start/stop services, manage admins, manage API keys, upload files, view logs/stats, and more—all via Telegram buttons
• Only the main owner can add/revoke admins and child bots
• All admin actions are protected

• Add a child bot: Provide token, username, and usage scope (everyone/admins/custom)
• Each child bot is isolated, gets its own API key and RAG collection
• Revoking a child bot stops its process and deletes its folder
• Usage scope is enforced (admins only, custom coming soon)

• All sensitive actions require admin rights
• All system commands use sudo (passwordless sudo recommended for specific commands)
• No hardcoded secrets
• Exception logging to server.log and child_bot.log

• Check server.log and child_bot.log for errors
• Use Telegram admin bot to view system logs and status
• Ensure all environment variables are set
• For GPU issues, check nvidia-smi and driver installation

• Add rate limiting, custom usage scopes, or more logging as needed
• Automated tests recommended for future changes

Ready for production.