inkmap c8194e5: docs(inklang): converge placement documentation

README.md

@@ -21,6 +21,7 @@ can read back.
 
 | do this | to |
 | --- | --- |
+| describe a placement, such as **on the left forearm** | resolve an exact spot with InkLang; choose from buttons if the wording is ambiguous |
 | click a design, then click the body | place it — a ghost follows your pointer until you click |
 | type a subject under **Generate a design** | have a new flash design drawn for you (a few seconds; it queues on a shared GPU) |
 | **A** / **D** | rotate on the skin (hold **Shift** for bigger steps) |
@@ -37,14 +38,21 @@ to your own computer. Only the subject you type is sent to the generator,
 which draws with the open-weights [Z-Image-Turbo](https://huggingface.co/Tongyi-MAI/Z-Image-Turbo)
 model and keeps nothing.
 
+InkLang is the placement system behind the description box. It normalizes the
+words, combines them with the selected body, and records one exact point on the
+body's canonical surface. It does not generate artwork. A phrase like
+**forearm** needs a left/right choice and nothing is generated or placed until
+you make it. The older “a fine line octopus on the left knee ditch” sentence
+also works; its design words and placement clause stay separate internally.
+
 ## What is in a saved layout
 
 Each tattoo is stored as a **surface anchor** — a spot on the body mesh — plus
 a rotation and a size in millimetres, and the file records exactly which body
-it was made on. That is deliberately not a picture: it is the information a
-robot would need to draw the same design in the same place. A layout is a
-sketch, not an instruction. Nothing here operates a machine or tattoos a
-person.
+it was made on. The original placement description, normalized meaning, and
+body-surface resolution travel with that anchor. That is deliberately not a
+picture. A layout is a sketch, not a robot instruction, reachability result, or
+contact qualification. Nothing here operates a machine or tattoos a person.
 
 ## Credits