docs: document image pipeline profiles

README.md

@@ -59,6 +59,17 @@ Operational notes:
 - Mirrored feeds are written under `out/feeds/<slug>/`.
   In production, expose `out/feeds/` directly from the reverse proxy at `/feeds/`.
 - `Feed URL` is used to generate absolute media URLs and `atom:link rel="self"` in exported feeds.
+- Image output is profile-driven. `REPUBLISHER_IMAGE` defines full-size
+  variants; the first profile is the canonical image URL used when feed image
+  URLs are rewritten.
+- Default image profiles keep source bytes under `images/source/`, write
+  full-size variants under `images/full/`, and write thumbnail profiles from
+  `REPUBLISHER_IMAGE_THUMBNAILS` under `images/thumbs/`.
+- Explicit item image media is exported as Media RSS image groups with named
+  thumbnails. Inline HTML images are mirrored and rewritten in content, but are
+  not promoted to item-level Media RSS.
+- Image profile names and transform settings are part of generated filenames.
+  Reordering `REPUBLISHER_IMAGE` changes canonical feed image URLs.
 - Job logs and stats artifacts are written under `out/logs/`.
 
 The legacy one-shot config-driven crawler is still available:
@@ -79,10 +90,9 @@ REPUBLISHER_FEED_URL = "https://mirror.example"
 - [x] Offlines RSS feed xml
 - [x] Downloads media and enclosures
 - [x] Rewrites media urls
-- [x] Image normalization (JPG, RGB)
+- [x] Profile-driven image normalization, compression, and thumbnails
 - [x] Audio transcoding
 - [x] Video transcoding
-- [ ] Image compression - Do we want this? -> DEFERED for now
 - [x] Download and rewrite media embedded in content/CDATA fields
 - [x] Config file to drive the program
 - [x] Add sqlite database and simple admin UI to replace config