GPUgraphics processorPC specs
GPU & RAM Requirements for Odysseus
Google Trends shows many users searching for GPU info after PewDiePie's launch. Here's what you actually need — split by the app vs local models.
Two separate things
Odysseus app — chat UI, agents, email, documents. Runs on CPU. 8 GB RAM is enough.
Local AI models — powered by Ollama or Cookbook. GPU optional but recommended for speed and larger models.
Hardware tiers
| Tier | Specs | What you can run |
|---|---|---|
| Minimum | 8 GB RAM, no GPU | Odysseus app + Ollama tiny models (0.8B–3B) on CPU |
| Recommended | 16 GB RAM + 8 GB VRAM (NVIDIA) | 7B–13B models smoothly via Ollama or Cookbook |
| Power user | 32 GB+ RAM + 12–24 GB VRAM | Large models, vLLM serving, multi-model workflows |
Minimum requirements
- → 8 GiB RAM or more (16 GiB+ recommended for local models)
- → Python 3.11+ (native install) or Docker Desktop / Docker Engine
- → Optional: NVIDIA GPU + drivers for Cookbook GPU serving on Linux
- → Optional: Ollama for the easiest local model setup on any OS
NVIDIA GPU in Docker (optional)
For Cookbook GPU model serving inside Docker on Linux or WSL2.
terminal
# From the Odysseus repo root (Linux / WSL2):
# 1. Diagnose / install NVIDIA Container Toolkit (optional helpers):
scripts/check-docker-gpu.sh --print-install-commands
# scripts/check-docker-gpu.sh --install-nvidia-toolkit
# 2. Enable the official overlay in .env (after passthrough works):
# COMPOSE_FILE=docker-compose.yml:docker/gpu.nvidia.yml
# Or run:
scripts/check-docker-gpu.sh --enable-nvidia-overlay
# 3. Recreate:
docker compose down
docker compose up -d --build
# Verify GPU inside the container:
docker compose exec odysseus nvidia-smi -L
# Stack UIs that need a single file: use docker-compose.gpu-nvidia.yml instead.No GPU? Use Ollama
Ollama is the easiest path for local models on any OS — including CPU-only machines. Start with a small model.
FAQ