docs(readme): production in the fleet deploys from infra/komodo, make deploy stays for a checkout deployed by hand

template/{{ part_dir }}/Makefile.jinja

@@ -1,5 +1,5 @@
 COMPOSE := {% if layout == 'part' %}cd .. && {% endif %}docker compose
-IMAGE := {{ project_slug }}/{{ part }}
+IMAGE := {{ project_slug | replace('_', '-') }}/{{ part }}
 changed = ! git diff --quiet $(1) -- $(2) 2>/dev/null
 stamp = { git update-ref $(1) HEAD 2>/dev/null || true; }
 

template/{{ part_dir }}/docker-compose.yml.jinja

@@ -1,7 +1,7 @@
 services:
   {{ service }}:
     build: {{ part_dir }}
-    image: {{ project_slug }}/{{ part }}
+    image: {{ project_slug | replace('_', '-') }}/{{ part }}
     profiles: [{{ service }}, services]
     restart: unless-stopped
     env_file:

template/{{ part_dir }}/{% if layout == 'root' %}README.md{% endif %}.jinja

@@ -10,3 +10,5 @@ make recreate
 `make env` creates `.env` and `docker-compose.override.yml` from their examples. `make recreate` starts the selected profiles, `make rebuild` builds the image first, `make logs`, `make restart` and `make down` do what they say.
 
 To deploy, `git pull && make deploy`. It runs `make pre-deploy`, which rebuilds the image only when something in the repository changed, then `docker compose up -d`. A deploy system with its own `up` step calls `make pre-deploy` before it and `make post-deploy` after. What was built is remembered as `refs/deploy/*` in the clone.
+
+In the fleet, production is deployed from `infra/komodo` instead (git mode of Komodo, see `docs/architecture.md` there): Komodo clones this repo, and a layer plus `deploy.sh` in the stack's directory tag images by the hash of what they bake in and write `.env` from sops. The repo only has to stay a plain recipe: images built from `build:` in compose, `caddy/site.caddy`. `make deploy` and `refs/deploy/*` are for a checkout deployed by hand.