Hermes is a self-improving AI agent that creates skills from experience and
improves them during use. It runs as a CLI on a developer's machine and
supports a pluggable toolset system, multiple LLM backends (OpenAI, Anthropic,
Mistral, Bedrock, ...), web search and crawling integrations, voice / TTS /
messaging connectors, and an agent-control protocol (ACP) adapter.

This port installs the core CLI plus the optional [web] dashboard
backend (FastAPI + Uvicorn).  The dashboard's static frontend assets
(hermes_cli/web_dist/) are produced by an upstream npm pipeline that is
not run in the ports build environment.  Other upstream extras such as
Modal, Daytona, Vercel, faster-whisper, Telegram, Discord, Slack and
the [rl] training loop are not packaged.

Because Hermes ships top-level Python packages with generic names
(tools, agent, gateway, plugins, ...), the source tree is installed
under ${PREFIX}/lib/hermes-agent/ rather than site-packages, matching
upstream's own Docker / Nix / Homebrew packaging.  The wrappers in
${PREFIX}/bin prepend that directory to sys.path before invoking each
entry point.

The skills/ and optional-skills/ template directories ship under
${DATADIR}/.  Hermes reads its working set from ${HOME}/.hermes/.
