Real, runnable configs and scripts referenced by the cookbook
and the agent design in
../architecture/agent-integration.md.
They split two ways: cloud agent bindings (JSON you pass to
bullmoose admin agent bind --config, run by the deployed agent worker) and
the homelab agent bridge (a box you run yourself, Pattern A).
| file | what it is |
|---|---|
on-your-own-domain.md |
Start here. Four worked paths from a bare domain to the first working thing — one person, a family, a business role address, a homelab — with every command on example.com rather than ours (#2). |
editor-emily.config.json |
Cloud reply binding — persona + a model-alias menu (cheap/llama on Workers AI, opus4.8/sonnet/… via AI Gateway). Powers EditorEmily (cookbook §2); see also the agent-accounts playbook. |
analyst-allen.config.json |
Cloud ledger binding — pipeline: "ledger", digest targets, spend categories, chart threshold. Powers Allen the Analyst (cookbook §3). |
hermes-bridge.sh |
The homelab bridge: bullmoose watch → a local agent (the hermes CLI) → reply out via popcorn SMTP, with a sender allowlist, RFC 3834 loop guards, and a watchdog. No cloud binding — Pattern A (cookbook §4). |
cc.bullmoose.hermes-bridge.plist |
macOS launchd agent that keeps the bridge running (RunAtLoad + KeepAlive) on the homelab box. __TOKEN__ is a placeholder for the bm_… app-password. |
The two *.config.json files are the server-side BindingConfig (persona,
defaultModel, modelAliases); the local agent serve runtime takes a
different schema — see the
agent-accounts section
of the top-level README for both.