Skip to content

Resource Requirements

TruSpace runs as a set of Docker containers. The resources required depend heavily on whether AI features are enabled and how many concurrent users and documents the deployment needs to support.

Estimates in progress

These are initial estimates based on development and testing environments. Requirements will be refined as more production data becomes available. If you have specific workloads, monitor performance and scale accordingly.

General Requirements

Component Minimum Recommended
CPU 1 vCPU 2+ vCPU
RAM 8 GB 16 GB
Storage 10 GB 100 GB
OS Linux 64-bit Ubuntu 22.04 LTS
Docker 20.10+ Latest stable

RAM is the critical resource

The local AI model (Ollama) is the largest consumer. A llama3.2:3b model alone requires ~4 GB of RAM at runtime. Plan accordingly or disable AI on constrained hardware.

Per-Deployment Profile

Profile RAM Storage Notes
Minimal (no AI) 4 GB 10 GB --no-ai flag or DISABLE_ALL_AI_FUNCTIONALITY=true
Standard 8 GB 50 GB Small team, lightweight model (phi3, llama3.2:3b)
Recommended 16 GB 100 GB+ Multi-user, larger models, IPFS replication
Raspberry Pi 4 GB 32 GB SD + SSD Pi ⅘ only, use tinyllama or no-AI mode

AI Model RAM Usage

Model Disk Size RAM at Runtime
tinyllama 637 MB ~2 GB
phi3 2.2 GB ~4 GB
llama3.2:3b 2.0 GB ~4 GB
llama3.2:7b 4.7 GB ~8 GB
mistral 4.1 GB ~8 GB

For faster AI inference, a GPU with NVIDIA Container Toolkit is recommended. Without a GPU, inference runs on CPU and can take minutes per document on smaller hardware.

Storage Growth

IPFS stores all document versions immutably. Storage usage grows over time:

  • Each uploaded document version is pinned permanently until explicitly unpinned
  • Multiple connected nodes each store a copy of pinned content
  • AI perspectives and metadata are also stored in IPFS

Monitor storage with:

# Total Docker volume usage
docker system df

# IPFS-specific repo stats
docker exec ipfs0 ipfs repo stat

Container Resource Limits

For constrained environments, cap individual containers with a docker-compose.override.yml:

services:
  backend:
    deploy:
      resources:
        limits:
          cpus: '1'
          memory: 512M

  ipfs0:
    deploy:
      resources:
        limits:
          cpus: '2'
          memory: 1G
    environment:
      - IPFS_PROFILE=lowpower

  ollama:
    deploy:
      resources:
        limits:
          cpus: '4'
          memory: 4G

  frontend:
    deploy:
      resources:
        limits:
          memory: 256M