Hermes Agent Configuration

Personal configuration for Hermes Agent (NousResearch).

Setup

Hermes is installed at /opt/hermes-agent and runs as a systemd user service.

LLM Backend

• Host: Mac Studio 1 via Tailscale (100.100.241.110)
• Routing proxy: port 8081 (Docker, auto-start via launchd)
• Fast model (default): mlx-community/Qwen3.5-122B-A10B-4bit — MLX backend, port 8082
• Quality model (delegation/fallback): qwen3.5-397b — llamacpp, port 8084

Messaging

• Telegram: enabled (bot: rHermesMS_bot)

Memory

Three-layer memory system:

1. Built-in — MEMORY.md / USER.md (file-based, in system prompt)
2. Vector — Qdrant MCP server (semantic search, local storage)
3. Session search — full-text over past transcripts

Gateway Service

# manage
systemctl --user {start,stop,restart,status} hermes-gateway

# diagnostics
hermes doctor

# update
cd /opt/hermes-agent && git pull origin main
uv pip install -e . --python /opt/hermes-agent/venv/bin/python
systemctl --user restart hermes-gateway

Files

File	Description
cli-config.yaml	Main CLI/agent configuration
config.yaml	Runtime config (model routing, MCP servers, TTS, etc.)
SOUL.md	Persona / system prompt
hermes-gateway.service	Systemd unit for gateway
.env.example	Environment variables template (secrets redacted)