Enodios

Ἐνόδιος — Hermes of the road. One-script Linux setup for Hermes Agent on your GPU via vLLM.

Local inference. Fast tool calling. No cloud API keys.

Documentation: https://dataknifeai.github.io/enodios/

---

Overview

Enodios installs vLLM, tunes it for Hermes 3 (64k context, tool parser), and wires Hermes to http://127.0.0.1:8000/v1.

Install order: Enodios → start vLLM → Hermes Agent → hermes setup.

Default model: solidrust/Hermes-3-Llama-3.1-8B-AWQ served as hermes3:8b (~20GB VRAM @ 64k on a 4090).

More: Advanced setup & reference — wizard walkthrough, distributed Hermes, tuning, troubleshooting.

---

Quick start

# 1. Enodios + vLLM
curl -fsSL https://raw.githubusercontent.com/DataKnifeAI/enodios/main/install.sh | bash
enodios recommend --apply && source ~/.local/share/enodios/recommended.env
enodios start -b && enodios bench

# 2. Hermes (after vLLM is up)
curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash
hermes setup    # Full setup → Inference Provider → Custom endpoint
# URL: http://127.0.0.1:8000/v1  model: hermes3:8b  context: 65536

# Or skip wizard:
enodios configure && hermes chat

# Free GPU for games / heavy tasks:
enodios stop      # resume: enodios start -b

Update: enodios update

---

Commands

Command	Description
enodios deps	Check OS GPU deps (apt/dnf/yum/zypper/pacman); --install
enodios install	OS deps check + venv + vLLM + CLI link
enodios update	git pull + upgrade vLLM
enodios recommend	GPU → model/settings (--apply to save)
enodios start -b	Background vLLM (loopback)
enodios start --lan	Expose on LAN for remote Hermes
enodios stop / pause	Free GPU VRAM
enodios configure	Wire Hermes to local vLLM
enodios configure --url URL	Wire Hermes to remote vLLM
enodios firewall	Check / allow port 8000 (--allow)
enodios urls	Local + LAN API URLs
enodios status	/v1/models + URLs
enodios bench	Tool-call smoke test
enodios doctor	GPU, venv, endpoint health
enodios help	Full CLI help

---

Architecture

Same machine (default):

flowchart LR
  U[You] --> H[Hermes Agent]
  H -->|127.0.0.1:8000/v1| V[vLLM]
  V --> G[NVIDIA GPU]
  H --> T[Tools]

Loading

Distributed — see advanced guide:

flowchart LR
  U[You] --> C[Controller Hermes]
  C -->|LAN :8000/v1| V[vLLM GPU host]
  V --> G[NVIDIA GPU]

Loading

Orchestrator + sub-agents — see advanced guide:

flowchart TB
  U[You] --> O[Hermes Orchestrator]
  O -->|delegate| C1[Client Hermes]
  O -->|delegate| C2[Client Hermes]
  O -->|delegate| C3[Client Hermes]
  O -->|plan :8000/v1| VO[vLLM]
  C1 -->|infer| V[vLLM via Enodios]
  C2 -->|infer| V
  C3 -->|infer| V
  C1 --> T1[Sub-task tools]
  C2 --> T2[Sub-task tools]
  C3 --> T3[Sub-task tools]
  VO --> G[NVIDIA GPU hosts]
  V --> G

Loading

---

License

MIT — see LICENSE.