Hermes Agent Setup Notes

Reference for current Hermes Agent configuration: installed skills, provider setup, known Windows quirks, and onboarding steps.

Status

• Status: reference only / public-safe
• This repo must never contain API keys, tokens, or private paths.

At a glance

• Runtime: Hermes desktop GUI
• Local model: hermes3:8b Q4_K_M GGUF via llama-cpp-python + CUDA Toolkit 12.4
• Default provider: Nous Portal Plus (deepseek/deepseek-v4-flash)
• Terminal: bash/MSYS on Windows 10
• Flutter: 3.44.3 in C:\src\flutter
• Android SDK: C:\Android\Sdk, build-tools 36, JDK 17

Skills convention

Skills live under:

• ~/AppData/Local/hermes/skills/<category>/<name>/SKILL.md

Categories: mlops, productivity, web-development, autonomous-ai-agents, etc.

Onboarding checklist

1. Install Hermes desktop GUI
2. Configure provider: Nous Portal token + default model
3. Setup local model (hermes3:8b GGUF) + CUDA toolkit on NVIDIA
4. Install essential skills: hermes, productivity, web-development, mlops
5. Avoid committing config.yaml, .env, credentials/, or tokens

Known issues

• Ollama GPU path requires CUDA Toolkit runtime libraries on Windows
• Next.js on Windows: prefer npx next start over npm run dev

Changelog

• v1.0.0 initial: setup reference sanitized