Hardware
GPU (Required)
RecordEngine requires an NVIDIA GPU with at least 16 GB VRAM. The GPU runs all AI inference locally — document extraction, summarisation, confidence scoring, and chat all happen on the GPU.CPU, RAM, and Storage
Storage grows with your document archive. Plan for approximately 1–5 MB per document including the original file and extracted data. At 500 documents per month, 200 GB lasts several years — but NVMe SSD is strongly recommended over SATA SSD for faster model loading on startup.
Software
Operating System
Required Software
All other dependencies (Python, FastAPI, Streamlit, Ollama, etc.) are included inside the Docker containers — nothing needs to be installed separately on the host OS.
Network
Inbound Ports
Ports 8501 (Streamlit) and 8510 (FastAPI) should not be exposed publicly — all external access goes through Nginx on ports 80/443.
Internet Access
RecordEngine operates fully offline after initial setup. Internet access is only required for:- Initial setup: Downloading Docker images and the AI model (~7–10 GB total)
- Software updates: Pulling new code and model versions (infrequent, on your schedule)
Firewall
If your server is behind a corporate firewall, the following outbound connections are needed only during setup:
After initial setup, all of these can be blocked if your security policy requires it.
Recommended Server Configurations
Standard Office Deployment
Lenovo ThinkStation or Dell Precision
- CPU: Intel Core i9 or Xeon W
- GPU: NVIDIA RTX 4090 (24 GB)
- RAM: 64 GB DDR5
- Storage: 1 TB NVMe SSD
- OS: Ubuntu 24.04 LTS
High-Volume Deployment
AWS g5.xlarge or equivalent
- GPU: NVIDIA A10G (24 GB)
- vCPU: 4
- RAM: 16 GB
- Storage: 500 GB+ EBS gp3
Client Requirements
Users access RecordEngine through a web browser — no software installation is required on client machines.
The mobile chat interface works on any modern mobile browser (iOS Safari, Android Chrome).