board: a tag in the prose is coloured where it sits, not printed twice

A tagged note was showing its tag twice — once where it was typed, once as a
chip — and the duplicate was the loud copy. Now the chip row carries only what
the body cannot say (a tag lifted off its own line, a label from the picker),
and a `#tag` left mid-sentence is tinted in place.

Which characters are a tag is asked of the CORE, the way the card already asks
it which lines are checklist items: `extract_tag_spans` keeps the spans
`extract_tags` throws away, and `body_tags` hands them to Kotlin. Offsets are
UTF-16 code units, because `AnnotatedString` and JS both index that way and a
char index lands mid-token the first time somebody writes an emoji. The web
keeps its own matcher in markdown.ts, mirroring `line_tags` case for case.

The inline ink is its own table, one Tailwind step deeper than the chip's. A
chip brings its own -100 fill and reads against that alone; inline text sits on
whatever the card is, including a gray-tagged card at neutral-200 — where the
chip's -700 measured 3.98 (green), 4.11 (orange) and 4.34 (teal), under the 4.5
body text needs. At -800/-300 every hue lands 5.63-12.01 light and 7.20-10.84
dark across every palette and generated fill.

Chips now carry the `#` on every surface. The via_tag branch that used to
decide it is gone from the card, and Android's row said no hash at all.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>

core/src/local/derive.rs

@@ -62,6 +62,62 @@ pub fn extract_tags(body: &str) -> Vec<String> {
     out
 }
 
+/// One `#tag` and exactly where it sits, for a renderer drawing the body itself.
+///
+/// The card no longer prints a chip for a tag whose text is still in the note — it
+/// colours the token where it was typed instead. To do that a renderer needs the
+/// SPAN, not just the name, and asking it to find the name again would be a second
+/// grammar quietly disagreeing with this one about what `##a` or `#1` is.
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct DerivedTag {
+    /// Which body line it sits on, like [`DerivedItem::line`].
+    pub line: u32,
+    /// Offsets into that line, in UTF-16 code units — INCLUDING the leading `#`.
+    ///
+    /// UTF-16 rather than chars or bytes because the two languages that consume this
+    /// both index strings that way: Kotlin's `AnnotatedString` and JavaScript. A char
+    /// index is right up until somebody puts an emoji before a tag, and then it lands
+    /// mid-token with no error anywhere.
+    pub start: u32,
+    pub end: u32,
+    pub name: String,
+}
+
+/// Every `#tag` in `body` with its position — the same scan [`extract_tags`] does,
+/// keeping the spans instead of throwing them away.
+///
+/// Not deduped, unlike `extract_tags`: two mentions of `#todo` are two pieces of text
+/// to colour. Fences are not skipped either, and that is deliberate — `extract_tags`
+/// does not skip them, so a `#tag` inside a code block IS a label on the note, and a
+/// renderer that left it plain would be the only surface disagreeing.
+pub fn extract_tag_spans(body: &str) -> Vec<DerivedTag> {
+    let mut out = Vec::new();
+    for (n, line) in body.split('\n').enumerate() {
+        let chars: Vec<char> = line.chars().collect();
+        let spans = line_tags(&chars);
+        if spans.is_empty() {
+            continue;
+        }
+        // Prefix sums, built once per tagged line: char index -> UTF-16 offset.
+        let mut units: Vec<u32> = Vec::with_capacity(chars.len() + 1);
+        let mut total: u32 = 0;
+        units.push(0);
+        for c in &chars {
+            total += c.len_utf16() as u32;
+            units.push(total);
+        }
+        for (start, end, name) in spans {
+            out.push(DerivedTag {
+                line: n as u32,
+                start: units[start],
+                end: units[end],
+                name,
+            });
+        }
+    }
+    out
+}
+
 /// Whether a line opens or closes a fenced code block.
 fn is_fence(line: &str) -> bool {
     let trimmed = line.trim_start();
@@ -438,6 +494,36 @@ mod tests {
         assert!(extract_tags("").is_empty());
     }
 
+    // ── tag spans, for the renderer that draws them in place ─────────────────
+
+    #[test]
+    fn tag_spans_carry_the_hash_and_the_line() {
+        let spans = extract_tag_spans("buy milk #grocery\nand call #mom about #mom");
+        assert_eq!(spans.len(), 3);
+        assert_eq!((spans[0].line, spans[0].start, spans[0].end), (0, 9, 17));
+        assert_eq!(spans[0].name, "grocery");
+        // Not deduped: two mentions are two pieces of text to colour.
+        assert_eq!(spans[1].line, 1);
+        assert_eq!(spans[2].name, "mom");
+        assert_eq!((spans[2].start, spans[2].end), (20, 24));
+    }
+
+    #[test]
+    fn tag_spans_are_utf16_offsets_not_char_indices() {
+        // The emoji is ONE char and TWO UTF-16 code units. Kotlin and JS both index
+        // the second way, so a char index would highlight one character too early.
+        let spans = extract_tag_spans("🎁 #gift");
+        assert_eq!(spans.len(), 1);
+        assert_eq!((spans[0].start, spans[0].end), (3, 8));
+    }
+
+    #[test]
+    fn tag_spans_agree_with_extract_tags_about_what_a_tag_is() {
+        let body = "#1 nope a#b no but #Yes ##no";
+        let names: Vec<String> = extract_tag_spans(body).into_iter().map(|t| t.name).collect();
+        assert_eq!(names, extract_tags(body));
+    }
+
     // ── lifting standalone tags ──────────────────────────────────────────────
     //
     // The MIRROR of `split_body_tags` in the server's notes/tags.py, case for case.