@vincemakes/kiso-tui 0.16.4 → 0.16.6

package/dist/ask-panel.js

@@ -263,7 +263,7 @@ export function askBlockRows(view, state, W, maxRows) {
     const counter = spec.questions.length > 1 ? `${p.dim} ‹ ${state.qIndex + 1}/${spec.questions.length} ›${p.reset}` : "";
     const rows = [];
     // R2: the same dashed rule the composer and the approval panel use.
-    rows.push(`${p.dim}${"\u254c".repeat(Math.max(0, W))}${p.reset}`);
+    rows.push(`${p.dim}${"\u2500".repeat(Math.max(0, W))}${p.reset}`);
     rows.push(`  ${cutLine(`${p.bold}${escapeTerminal(q.question)}${p.reset}${counter}`, Math.max(1, W - 2))}`);
     const header = q.header === undefined ? "the question" : escapeTerminal(q.header.slice(0, ASK_HEADER_CAP));
     // R2: the divider row is gone (the opening rule says a block starts
@@ -317,7 +317,7 @@ export function askBlockRows(view, state, W, maxRows) {
     // TUI2-R1.5 ⑪ had already replaced a two-cell `\u2514 ` stub with a
     // real rule for the reason that stub read as the cut-notice prefix it
     // collides with; this keeps that finding and only changes which rule.
-    rows.push(`${p.dim}${"\u254c".repeat(Math.max(0, W))}${p.reset}`);
+    rows.push(`${p.dim}${"\u2500".repeat(Math.max(0, W))}${p.reset}`);
     return rows;
 }
 /** The status row's right-hand hint — the phase's keys. */

package/dist/at-picker.js

@@ -265,6 +265,6 @@ export function atPanelRows(state, W) {
  */
 export function bandHeader(label, W) {
     const p = palette();
-    const head = `\u254c\u254c\u254c ${label} `;
-    return `${p.dim}${widthCut(`${head}${"\u254c".repeat(Math.max(1, W - head.length))}`, Math.max(1, W))}${p.reset}`;
+    const head = `\u2500\u2500\u2500 ${label} `;
+    return `${p.dim}${widthCut(`${head}${"\u2500".repeat(Math.max(1, W - head.length))}`, Math.max(1, W))}${p.reset}`;
 }

package/dist/components.d.ts

@@ -2,3 +2,5 @@
  *  Amendment 4) — this module is the re-export shim: the compositor's
  *  and index.ts's imports (./components.js) stay verbatim. */
 export * from "@vincemakes/kiso-tui-cells/components";
+export { foldTerms } from "@vincemakes/kiso-tui-cells/components";
+export { MOTION_FRAMES, TWINKLE, breathFrame, twinkleFrame } from "@vincemakes/kiso-tui-cells/render";

package/dist/components.js

@@ -2,3 +2,7 @@
  *  Amendment 4) — this module is the re-export shim: the compositor's
  *  and index.ts's imports (./components.js) stay verbatim. */
 export * from "@vincemakes/kiso-tui-cells/components";
+// R3 (design §5.2): the two motion cycles reach the compositor through
+// the same shim every other cell primitive does.
+export { foldTerms } from "@vincemakes/kiso-tui-cells/components";
+export { MOTION_FRAMES, TWINKLE, breathFrame, twinkleFrame } from "@vincemakes/kiso-tui-cells/render";

package/dist/compositor.js

@@ -52,7 +52,7 @@ import { MOUSE_OFF } from "./editor.js";
 import { atPanelRows, bandHeader } from "./at-picker.js";
 // TUI2-R2 ②: the session picker's rows — the band's third occupant.
 import { sessionPickerRows } from "./session-picker.js";
-import { Container, ROLLUP_NOUN, SPINNER, MdStream, bodySpacing, boxBottom, boxTop, cellComponent, exploreCounts, focusToken, exploreRows, foldLine, isExploreTool, pendingQueueRows, statusLine, turnFold, visibleWidth, } from "./components.js";
+import { Container, ROLLUP_NOUN, MOTION_FRAMES, MdStream, bodySpacing, boxBottom, boxTop, cellComponent, exploreCounts, foldTerms, focusToken, exploreRows, foldLine, isExploreTool, pendingQueueRows, statusLine, turnFold, visibleWidth, } from "./components.js";
 import { bannerLines, escapeTerminal, foldResult, foldThinking, palette, renderTerminalGap, renderToolSummary, toolTarget } from "./render.js";
 import { displayVerb, keysSheetRows } from "./strings.js";
 /** The cursor marker — an APC private sequence the focus component
@@ -69,6 +69,123 @@ const NOT_PAINTED = "\u0000never";
  *  that a single resize still feels immediate. */
 const RESIZE_SETTLE_MS = 80;
 const CHROME_ROWS = 4; // box top + input + box bottom + status — the design §03 chrome (V6-3; the box is W6)
+/** W13 / TUI2-R1 (B) — a rolled run's TITLE: the exploration sentence on
+ *  a mixed run, W13's verb+count on a single-name one. */
+function rolledTitle(cell) {
+    const r = cell.rolled;
+    if (r.parts !== undefined)
+        return `explored ${exploreCounts(r.parts)}`;
+    return `${displayVerb(cell.name)} ${r.count} ${ROLLUP_NOUN[cell.name] ?? "calls"}`;
+}
+/** W13 / TUI2-R1 (B) — a rolled run's DETAIL rows: one row per tool with
+ *  its subjects on a mixed run, one `└ target` per call on a single-name
+ *  one.
+ *
+ *  Extracted at R3b so the segment fold's expansion opens a run to the
+ *  SAME rows `ctrl+r` on the run itself would have opened. Two copies of
+ *  this would be two answers to "show me that run". */
+function rolledDetail(cell, W) {
+    const p = palette();
+    const r = cell.rolled;
+    if (r.parts !== undefined)
+        return exploreRows(r.parts, W);
+    return r.targets.map((t) => `  ${p.dim}└ ${escapeTerminal(t)}${p.reset}`);
+}
+/**
+ * W13 / TUI2-R1 (B) — the rollup's own projection of a run: the count,
+ * the lines, the elapsed, the targets, and — on a MIXED run only — the
+ * per-tool parts. A single-name run keeps W13's row byte for byte,
+ * which is the "the generalization adds, it never rewrites" rule.
+ *
+ * Extracted at R3b because the segment fold's EXPANSION renders the run
+ * through this same projection rather than reimplementing it — so the
+ * expanded rows cannot drift from the ones the commit path would have
+ * drawn.
+ */
+function rolledOf(members) {
+    let total = 0;
+    const targets = [];
+    for (const m of members) {
+        // the lines count, excluding the tool's OWN truncation note
+        // (read_file's "… N more lines") — the per-cell meta's rule
+        const noteAt = m.resultText.lastIndexOf("\n… ");
+        const shown = noteAt >= 0 ? m.resultText.slice(0, noteAt) : m.resultText;
+        const rows = shown.split("\n");
+        total += rows[rows.length - 1] === "" ? rows.length - 1 : rows.length;
+        let input = {};
+        try {
+            input = JSON.parse(m.inputFull);
+        }
+        catch {
+            // the full JSON is always parseable (stringified at toolStart)
+        }
+        const target = toolTarget(m.name, input);
+        targets.push(target.split("/").pop() ?? target);
+    }
+    const parts = exploreParts(members);
+    const first = members[0];
+    const last = members[members.length - 1];
+    const elapsed = first.startedAt !== null && last.doneAt !== null ? ((last.doneAt - first.startedAt) / 1000).toFixed(1) : "?";
+    return { count: members.length, lines: total, elapsed, targets, ...(parts.length > 1 ? { parts } : {}) };
+}
+/**
+ * TUI2-R1 (B) / R3b — the per-tool parts of an explore run, in
+ * first-call order. A search's subject is the PATTERN it looked for
+ * (quoted); a read's or a list's is the path it named.
+ *
+ * Extracted at R3b because TWO paths need it now: the commit-time
+ * rollup, which has always built it, and the segment fold's EXPANSION,
+ * which shows the rollup's rows rather than one row per call. Two
+ * copies of this would be two answers to "what did that run do".
+ */
+function exploreParts(members) {
+    const parts = [];
+    for (const m of members) {
+        let input = {};
+        try {
+            input = JSON.parse(m.inputFull);
+        }
+        catch {
+            // the full JSON is always parseable (stringified at toolStart)
+        }
+        const target = toolTarget(m.name, input);
+        const subject = m.name === "search_text" ? `"${String(input.pattern ?? "")}"` : target;
+        const part = parts.find((x) => x.name === m.name);
+        if (part === undefined)
+            parts.push({ name: m.name, subjects: [subject] });
+        else
+            part.subjects.push(subject);
+    }
+    return parts;
+}
+/** R3b — a segment's terms, for the expand header. The fold line's own
+ *  wording comes from `turnFold`; this is the same facts in the header
+ *  idiom the other expands use. */
+function foldMeta(seg) {
+    const parts = foldTerms(seg.reads, seg.edits, [...seg.others]);
+    return parts.length === 0 ? "thinking" : parts.join(" · ");
+}
+/** R3b — the turn's open segment, opened on demand at the first cell of
+ *  work that follows a text block (or the turn's start). Returns null
+ *  only when there is no turn at all, which is the pipe path's shape. */
+function openSegment(turn, now) {
+    if (turn === undefined)
+        return null;
+    const last = turn.segments[turn.segments.length - 1];
+    if (last !== undefined && last.closedAt === null)
+        return last;
+    const fresh = { openedAt: now, closedAt: null, reads: 0, edits: 0, others: new Map(), folded: false, spilled: false, headCell: null, cells: [] };
+    turn.segments.push(fresh);
+    return fresh;
+}
+/** R3b — close the turn's open segment, if it has one. Idempotent: text
+ *  arriving twice in a row closes nothing the second time, which is what
+ *  keeps a zero-cell segment from ever existing. */
+function closeSegment(turn, now) {
+    const last = turn?.segments[turn.segments.length - 1];
+    if (last !== undefined && last.closedAt === null)
+        last.closedAt = now;
+}
 /** W20 — the whole-table-replace comparison: the live task block only
  *  redraws when the items actually changed (the task extension's
  *  idempotent shape — an unchanged replace is a no-op, no frame). */
@@ -163,6 +280,11 @@ export class Body {
     #lastThinking = null;
     #lastTool = null;
     #pendingCalls = new Map();
+    /** R3b — cell index → the index of the segment it belongs to, for
+     *  thinking/tool cells; -1 for every other kind. Parallel to #cells,
+     *  because a segment is the COMPOSITOR's bookkeeping and does not
+     *  belong on the cell type the renderer sees. */
+    #cellSegment = [];
     #pipeBuf = ""; // the passthrough's thinking buffer
     /** TUI2-MD ⑤ — the markdown scanner of the message currently
      *  streaming, and the cell index its first block landed at. Null
@@ -320,7 +442,7 @@ export class Body {
         // W14: the turn boundary — the record the fold-hold's release
         // state machine reads; the cell carries the record's index. A9:
         // the user's own words ride the record — the fold's leading chip.
-        this.#turns.push({ ended: false, hasText: false, thoughtSeconds: 0, reads: 0, edits: 0, others: new Map(), words: text, folded: false });
+        this.#turns.push({ ended: false, hasText: false, thoughtSeconds: 0, reads: 0, edits: 0, others: new Map(), words: text, folded: false, segments: [] });
         this.#cells.push({ kind: "user", text, done: true, turn: this.#turns.length - 1 });
         this.#mark();
     }
@@ -335,6 +457,17 @@ export class Body {
         }
         else {
             this.#cells.push({ kind: "thinking", text, done: false, turn: this.#turns.length - 1 });
+            // R3b: thinking is WORK, so it opens a segment too — a turn that
+            // thinks, speaks, then thinks again has two segments, and the
+            // second one's clock starts here rather than at a tool call it
+            // may never make.
+            //
+            // OPEN then STAMP, in that order: the stamp records the segment
+            // the cell belongs to, and a stamp taken first records the
+            // PREVIOUS segment (or none at all) — which left the thinking
+            // row standing outside the fold it should have led.
+            openSegment(this.#turns[this.#turns.length - 1], Date.now());
+            this.#stampSegment();
         }
         this.#mark();
     }
@@ -393,7 +526,19 @@ export class Body {
                 turn.edits += 1;
             else
                 turn.others.set(name, (turn.others.get(name) ?? 0) + 1);
+            // R3b: and into the SEGMENT, which opens here when this is the
+            // first work since the last text block.
+            const seg = openSegment(turn, Date.now());
+            if (seg !== null) {
+                if (name === "read_file")
+                    seg.reads += 1;
+                else if (name === "edit_file")
+                    seg.edits += 1;
+                else
+                    seg.others.set(name, (seg.others.get(name) ?? 0) + 1);
+            }
         }
+        this.#stampSegment();
         this.#mark();
     }
     toolApproval(callId, diff) {
@@ -515,6 +660,11 @@ export class Body {
         const turn = this.#turns[this.#turns.length - 1];
         if (turn !== undefined)
             turn.hasText = true;
+        // R3b: text CLOSES the open segment. This is the boundary design.md
+        // §8 names — "folding at every text boundary changes what commits
+        // and when" — and it is the whole mechanism: a segment is what sits
+        // between two of these.
+        closeSegment(turn, Date.now());
         // TUI2-MD ⑤: assistant body text is MARKDOWN, scanned as it
         // streams. The scanner yields CLOSED blocks (final source, final
         // render) and one OPEN tail block; each becomes a cell, and the
@@ -598,6 +748,9 @@ export class Body {
             return;
         turn.ended = true;
         turn.thoughtSeconds = thoughtSeconds;
+        // R3b: the settle closes the last open segment — the turn's end is
+        // a boundary exactly as a text block is.
+        closeSegment(turn, Date.now());
         // W20: the turn's live task block settles HERE — the ONE recap
         // block for the turn ("`task done · N items · <duration>", the
         // duration clocked compositor-side from the block's first call —
@@ -801,6 +954,89 @@ export class Body {
         const idx = this.#collapsed[this.#expandPtr % this.#collapsed.length];
         this.#expandPtr += 1;
         const cell = this.#cells[idx];
+        // R3b — a folded SEGMENT expands to the work it stands for.
+        //
+        // The fold line collapses a run of thinking and tool cells into
+        // one row; without this the run would be unreachable, which is
+        // hiding a durable record behind a summary. The rows are APPENDED
+        // (ADR-0046 — history is never rewritten), exactly as every other
+        // expand in this method does, and they are the cells' OWN renders,
+        // so the expansion cannot drift from what was folded.
+        const seg = this.#segmentOf(idx);
+        if (seg !== null && seg.headCell === idx) {
+            const p = palette();
+            const turnsBack = this.#cells.slice(idx + 1).filter((c) => c.kind === "user").length;
+            const back = `${turnsBack} ${turnsBack === 1 ? "turn" : "turns"} back`;
+            const W = this.#opts.width();
+            const ctx = { spinnerI: this.#spinnerI, now: Date.now(), height: this.#opts.height() };
+            // R3b (owner ruling): the ROLLUP is the expansion. TUI2-R1 built
+            // a richer projection of an explore run than a fold line can
+            // carry — the per-tool counts, and one row per tool with its
+            // subjects — and the segment fold would have retired it by
+            // simply arriving first. So the run's own rows are what the key
+            // opens: explore tools group the way the rollup groups them,
+            // everything else renders as itself.
+            // The segment's cells IN ORDER, with consecutive explore tools
+            // grouped exactly as the rollup groups them — a write, a shell
+            // or anything else BREAKS the run, which is TUI2-R1's own rule
+            // and the reason two explore runs on either side of a write
+            // stay two runs. Merging every explore tool of the segment
+            // would have been simpler and would have quietly deleted that
+            // rule.
+            const rows = [];
+            let run = [];
+            const flush = () => {
+                if (run.length === 0)
+                    return;
+                // the same threshold the commit-time rollup uses: below it a
+                // "run" is just some rows
+                if (run.length > 2) {
+                    // the run renders through the ROLLUP's own projection —
+                    // literally the same function the commit path uses — so
+                    // a single-name run keeps W13's row and a mixed one gets
+                    // the exploration line, exactly as they would have if the
+                    // segment had never folded.
+                    const head = run[0];
+                    const saved = head.rolled;
+                    head.rolled = rolledOf(run);
+                    // the run OPENS. The fold's key already asked to see the
+                    // work, so what lands is the same rows `ctrl+r` on the
+                    // run itself would have opened — its title, then its
+                    // detail — never its collapsed row, which would make the
+                    // reader press a second time for what the first press
+                    // was for.
+                    rows.push(`  ${p.dim}${escapeTerminal(rolledTitle(head))}${p.reset}`);
+                    rows.push(...rolledDetail(head, W));
+                    head.rolled = saved;
+                }
+                else {
+                    for (const c of run)
+                        rows.push(...cellComponent(c).render(W, ctx));
+                }
+                run = [];
+            };
+            for (const j of seg.cells) {
+                if (j < idx)
+                    continue;
+                const c = this.#cells[j];
+                if (c.kind === "tool" && isExploreTool(c.name)) {
+                    run.push(c);
+                    continue;
+                }
+                flush();
+                rows.push(...cellComponent(c).render(W, ctx));
+            }
+            flush();
+            // the header NAMES the segment. When the segment is exactly one
+            // explore run, "explored 8 files · 14 searches" is what that run
+            // is called everywhere else in the product, and the header says
+            // the same thing rather than a second wording of it.
+            // the header states what the SEGMENT did, in the fold line's own
+            // terms; each run below states what IT did, in the rollup's. Two
+            // scales, one wording each — the header used to borrow the run's
+            // sentence, which read as the same run twice.
+            return { kind: "appended", lines: [`${p.bold}✦${p.reset} expanded · ${escapeTerminal(foldMeta(seg))} · ${back}`, ...rows] };
+        }
         if (cell.kind !== "tool")
             return { kind: "none" };
         if (cell.rolled !== null) {
@@ -814,15 +1050,10 @@ export class Body {
             // TUI2-R1 (B): an EXPLORATION head lists per TOOL — the counts
             // the row showed, then one row per tool with its subjects. The
             // header keeps W15's shape; only the subject changes.
-            if (cell.rolled.parts !== undefined) {
-                const header = `${p.bold}✦${p.reset} expanded · ${escapeTerminal(`explored ${exploreCounts(cell.rolled.parts)}`)} · ${back}`;
-                return { kind: "appended", lines: [header, ...exploreRows(cell.rolled.parts, this.#opts.width())] };
-            }
-            const noun = ROLLUP_NOUN[cell.name] ?? "calls";
-            const header = `${p.bold}✦${p.reset} expanded · ${escapeTerminal(`${displayVerb(cell.name)} ${cell.rolled.count} ${noun}`)} · ${back}`;
+            const rolledHead = rolledTitle(cell);
             return {
                 kind: "appended",
-                lines: [header, ...cell.rolled.targets.map((t) => `  ${p.dim}└ ${escapeTerminal(t)}${p.reset}`)],
+                lines: [`${p.bold}✦${p.reset} expanded · ${escapeTerminal(rolledHead)} · ${back}`, ...rolledDetail(cell, this.#opts.width())],
             };
         }
         let input = {};
@@ -1232,15 +1463,32 @@ export class Body {
         }, this.#frameMs);
         this.#frameTimer.unref();
     }
-    /** The spinner: a ONE-SHOT re-armed ONLY while a running tool exists —
-     *  no tool → no timer → zero bytes (the #14/#15 contract, structural). */
+    /**
+     * The spinner: a ONE-SHOT re-armed ONLY while something is MOVING —
+     * nothing moving → no timer → zero bytes (the #14/#15 contract,
+     * structural).
+     *
+     * R3: "moving" used to mean a running TOOL and nothing else. But the
+     * same counter drives the status row's `working` twinkle, and the
+     * model spends whole stretches thinking with no tool open at all —
+     * so the one indicator whose entire job is "I am still alive" froze
+     * solid exactly when the user most needed it to move. It was
+     * reported as "I thought it had failed", which is precisely the
+     * message a frozen liveness mark sends.
+     *
+     * An OPEN THINKING cell counts as movement now, and so does a tool
+     * parked at an approval — that one is waiting on a human rather than
+     * working, but the mark is still the proof the process is alive.
+     */
     #armSpinner() {
         if (this.#spinnerTimer !== null)
             return;
         this.#spinnerTimer = setTimeout(() => {
             this.#spinnerTimer = null;
-            if (this.#cells.some((c) => c.kind === "tool" && c.state === "running" && !c.done)) {
-                this.#spinnerI = (this.#spinnerI + 1) % SPINNER.length;
+            const moving = this.#cells.some((c) => (c.kind === "tool" && (c.state === "running" || c.state === "approval") && !c.done) ||
+                (c.kind === "thinking" && !c.done));
+            if (moving) {
+                this.#spinnerI = (this.#spinnerI + 1) % MOTION_FRAMES;
                 this.#dirty = true;
                 this.#scheduleFrame();
                 this.#armSpinner();
@@ -1603,7 +1851,11 @@ export class Body {
         // unshift: the cells commit oldest-first, so the NEWEST cut lands
         // at the front — the expand pointer's "newest back" walk starts
         // where the user's last key press would aim.
-        if (cell.kind === "tool" && lines.some((l) => l.includes("ctrl+r")))
+        // R3b: a fold HEAD joins the ring too. The test used to demand a
+        // tool cell, and a segment's fold can be emitted at a thinking
+        // cell — which would have left the whole segment unreachable by
+        // the very key its own row advertises.
+        if ((cell.kind === "tool" || this.#segmentOf(i)?.headCell === i) && lines.some((l) => l.includes("ctrl+r")))
             this.#collapsed.unshift(i);
         this.#lineCache[i] = lines;
         const placed = this.#space(i, i > 0 ? this.#lineCache[i - 1] : null, lines);
@@ -1611,6 +1863,60 @@ export class Body {
         this.#committedLines += placed.length;
         this.#committedLinesThisFrame.push(...placed);
     }
+    /** R3b — record which segment the cell just pushed belongs to. Called
+     *  right after the push, so #cells.length-1 is that cell. */
+    #stampSegment() {
+        const turn = this.#turns[this.#turns.length - 1];
+        const idx = turn === undefined ? -1 : turn.segments.length - 1;
+        const at = this.#cells.length - 1;
+        this.#cellSegment[at] = idx;
+        if (turn !== undefined && idx >= 0)
+            turn.segments[idx].cells.push(at);
+    }
+    /**
+     * R3b — does the segment hold a call that FAILED or was DENIED?
+     *
+     * Such a segment does not fold. Routine work is what the fold is for;
+     * a refusal and an error are the opposite of routine, and putting
+     * either behind a key hides the one thing on the screen that most
+     * needs a human's eye. Law 1.3 makes the same call about marks — a
+     * failure keeps its colour AND its words — and this is that rule at
+     * the scale of a run.
+     *
+     * The cost, accepted: a turn that reads twenty files and hits one
+     * denial keeps all twenty rows. The alternative is a screen that says
+     * `✦ thought 3s · 20 reads` while a write was refused inside it.
+     */
+    #segmentHasTrouble(seg) {
+        return this.#segmentTools(seg).some((c) => c.isError || c.reason !== null);
+    }
+    /** R3b — the segment's TOOL cells, in order. */
+    #segmentTools(seg) {
+        const out = [];
+        for (const j of seg.cells) {
+            const c = this.#cells[j];
+            if (c.kind === "tool")
+                out.push(c);
+        }
+        return out;
+    }
+    /** R3b — how many cells the segment holds. The fold's threshold reads
+     *  it; nothing else needs it, so it is counted rather than tracked. */
+    #segmentCells(seg) {
+        return seg.cells.length;
+    }
+    /** R3b — the segment a committed cell belongs to, or null when it has
+     *  none (a cell of the pipe path, or a kind that is not work). */
+    #segmentOf(i) {
+        const cell = this.#cells[i];
+        if (cell.kind !== "thinking" && cell.kind !== "tool")
+            return null;
+        const turn = cell.turn >= 0 ? this.#turns[cell.turn] : undefined;
+        const si = this.#cellSegment[i];
+        if (turn === undefined || si === undefined || si < 0)
+            return null;
+        return turn.segments[si] ?? null;
+    }
     /** W14 — the fold-hold: a thinking/tool cell of the OPEN quiet turn
      *  (no text yet) does not commit — its committed form is decided at
      *  the release. The cell's OWN turn must be the CURRENT one (a cell
@@ -1623,10 +1929,32 @@ export class Body {
         const turn = cell.turn >= 0 ? this.#turns[cell.turn] : undefined;
         if (turn === undefined || turn !== this.#turns[this.#turns.length - 1])
             return false;
+        // R3b (owner, 2026-08-27) — the hold is the SEGMENT's, not the
+        // quiet turn's.
+        //
+        // W14 held a turn's work only while the turn had produced no text
+        // at all, because the fold existed only for a turn that never
+        // spoke. The owner ruled that a segment folds the moment text
+        // arrives, so the unit whose committed form is undecided is the
+        // SEGMENT: while it is open, its cells must not reach the
+        // scrollback, because a committed row cannot be replaced by the
+        // fold line that is going to stand for it.
+        //
+        // This is exactly the change design.md §8 warned about — "folding
+        // at every text boundary changes what commits and when" — and the
+        // warning is why the hold is stated here, once, rather than
+        // spread across the callers.
+        //
+        // The quiet turn is the same rule seen from one side: its single
+        // segment never closes until the settle, so it holds exactly as
+        // it always did.
+        const seg = this.#segmentOf(i);
+        if (seg !== null)
+            return seg.closedAt === null;
+        // no segment (the pipe path's shape) — W14's original test, kept
+        // so a cell that never got a segment behaves as it used to.
         if (!turn.ended && !turn.hasText)
             return true;
-        // the turn's END releases every hold — the settle is where the run
-        // is decided, and a held cell at settle would never commit at all.
         if (turn.ended)
             return false;
         return this.#growingRun(i);
@@ -1679,18 +2007,53 @@ export class Body {
     #foldOrRollup(cell, i, W, ctx) {
         if (cell.kind === "thinking" || cell.kind === "tool") {
             const turn = cell.turn >= 0 ? this.#turns[cell.turn] : undefined;
-            if (turn !== undefined && turn.ended && !turn.hasText) {
-                if (!turn.folded) {
+            const seg = this.#segmentOf(i);
+            // R3b — the SEGMENT folds, and it folds once.
+            //
+            // A cell only reaches here once its segment is closed (the hold
+            // above keeps an open segment out of the commit loop entirely),
+            // so the first cell of a closed segment emits the fold and every
+            // cell after it renders nothing. `folded` is the latch: without
+            // it the second cell would emit a second fold line for the same
+            // work.
+            //
+            // A SPILLED segment is the honest degradation. The force-commit
+            // path does not consult the hold — the screen's hard cap wins —
+            // so a segment too big for the screen already has rows in the
+            // scrollback that cannot be taken back. It renders normally and
+            // says nothing false; it simply does not collapse.
+            // R3b — a ONE-CELL segment does not fold. Collapsing one row into
+            // one row gains no space and costs the row's subject: `✦ thought
+            // 0s · 1 shell` says strictly less than `shell make build ·
+            // exit 0`. The fold exists to stop a screen filling with work
+            // rows, and one row is not that.
+            if (turn !== undefined && seg !== null && seg.closedAt !== null && !seg.spilled && this.#segmentCells(seg) >= 2 && !this.#segmentHasTrouble(seg)) {
+                if (!seg.folded) {
+                    seg.folded = true;
+                    seg.headCell = i;
                     turn.folded = true;
                     // A9 (ruling R2, mock A): the user chip rides the fold —
-                    // the words take the fold's width budget (turnFold is
-                    // W-aware — the ONE row never trips invariant ①).
+                    // but ONLY on a quiet turn, where the fold stands for the
+                    // whole turn and the chip has nowhere else to be. In a
+                    // turn WITH text the chip cell commits on its own, so a
+                    // fold that repeated the words would put the user's line
+                    // on screen twice. The words take the fold's width budget
+                    // (turnFold is W-aware — the ONE row never trips
+                    // invariant ①).
+                    const quiet = turn.ended && !turn.hasText;
+                    // A QUIET turn (no text at all) keeps `thoughtSeconds` —
+                    // the CLI's own measure, taken at the settle, and the
+                    // number every W14 gate pins. A mid-turn segment cannot
+                    // have it (endTurn has not run), so it reports its own
+                    // wall clock, which is the only honest number available
+                    // at the moment it folds.
+                    const seconds = quiet ? turn.thoughtSeconds : Math.max(0, Math.round(((seg.closedAt ?? 0) - seg.openedAt) / 1000));
                     return turnFold({
-                        words: turn.words,
-                        thoughtSeconds: turn.thoughtSeconds,
-                        reads: turn.reads,
-                        edits: turn.edits,
-                        others: [...turn.others],
+                        words: quiet ? turn.words : "",
+                        thoughtSeconds: seconds,
+                        reads: seg.reads,
+                        edits: seg.edits,
+                        others: [...seg.others],
                     }, W);
                 }
                 return [];
@@ -1758,37 +2121,7 @@ export class Body {
             // exploration row's counts and its expanded list both read them.
             // A search's subject is the PATTERN it looked for (quoted); a
             // read's or a list's is the path it named.
-            const parts = [];
-            for (const m of members) {
-                // the lines count, excluding the tool's OWN truncation note
-                // (read_file's "… N more lines") — the per-cell meta's rule
-                const noteAt = m.resultText.lastIndexOf("\n… ");
-                const shown = noteAt >= 0 ? m.resultText.slice(0, noteAt) : m.resultText;
-                const rows = shown.split("\n");
-                total += rows[rows.length - 1] === "" ? rows.length - 1 : rows.length;
-                let input = {};
-                try {
-                    input = JSON.parse(m.inputFull);
-                }
-                catch {
-                    // the full JSON is always parseable (stringified at
-                    // toolStart) — the empty fallback never fires
-                }
-                const target = toolTarget(m.name, input);
-                targets.push(target.split("/").pop() ?? target);
-                const subject = m.name === "search_text" ? `"${String(input.pattern ?? "")}"` : target;
-                const part = parts.find((x) => x.name === m.name);
-                if (part === undefined)
-                    parts.push({ name: m.name, subjects: [subject] });
-                else
-                    part.subjects.push(subject);
-            }
-            const first = members[0];
-            const last = members[members.length - 1];
-            const elapsed = first.startedAt !== null && last.doneAt !== null ? ((last.doneAt - first.startedAt) / 1000).toFixed(1) : "?";
-            // TUI2-R1 (B): `parts` rides ONLY a mixed run — a single-name
-            // run keeps W13's row, byte for byte (the generalization adds).
-            cell.rolled = { count: members.length, lines: total, elapsed, targets, ...(parts.length > 1 ? { parts } : {}) };
+            cell.rolled = rolledOf(members);
             return cellComponent(cell).render(W, ctx);
         }
         // a MEMBER of an already-rolled run → [] (its rows live in the

package/dist/editor.js

@@ -88,7 +88,13 @@ const N_MAX = 6;
 const OSC_MAX = 1024;
 /** The dim "…" — the ONE truncation mark: the horizontal scroll's
  *  prefix (unchanged) and the viewport's hidden-rows markers. */
-const ELLIPSIS = "\x1b[2m…\x1b[0m";
+/** R3: built per call from the palette — `dim` is an absolute grey once
+ *  the ground is known, so a frozen SGR 2 here would be the one span
+ *  that ignores it. */
+const ellipsis = () => {
+    const p = palette();
+    return `${p.dim}\u2026${p.reset}`;
+};
 /**
  * The editor. Raw mode + bracketed paste (?2004h) on enter, restored on
  * exit. The input row is rendered by `onRender` (the CLI wires it to the
@@ -459,9 +465,9 @@ export class Editor {
             // the cursor's own row carries the horizontal scroll (and its
             // "…"); the other rows render whole and cap at the frame's wall
             const from = i === cursorLine ? b.start + this.#scroll : b.start;
-            const scrolled = i === cursorLine && this.#scroll > 0 ? ELLIPSIS : "";
-            const above = i === first && first > 0 ? ELLIPSIS : "";
-            const below = i === first + n - 1 && first + n < bounds.length ? ELLIPSIS : "";
+            const scrolled = i === cursorLine && this.#scroll > 0 ? ellipsis() : "";
+            const above = i === first && first > 0 ? ellipsis() : "";
+            const below = i === first + n - 1 && first + n < bounds.length ? ellipsis() : "";
             lines.push(`${above}${scrolled}${String.fromCodePoint(...this.#chars.slice(from, b.end))}${below}`);
         }
         const cursorRow = cursorLine - first;

package/dist/status.d.ts

@@ -21,9 +21,21 @@
  * deliberate exception: the running row's interrupt hint, which KC2 §2
  * widens to name the new gesture.
  */
-/** v3 §03/§05 — the working glyph family; the CLI's 200ms spinner walks
- *  it and hands each glyph back to `runningStatus`. */
-export declare const STATUS_GLYPHS: readonly ["▖", "▘", "▝", "▗"];
+/**
+ * R3 (design §5.2) — the working glyph family is the TWINKLE, and the
+ * CLI's 200ms spinner walks it exactly as it walked the four quadrant
+ * blocks it replaces.
+ *
+ * Two reasons the quadrants had to go. §5.3: they ROTATED, and a mark
+ * that turns implies progress a call of unpredictable duration does not
+ * have. §4.1: the twinkle settles onto `✦`, which is the mark the
+ * folded segment keeps — so the glyph a human watches while the model
+ * thinks is the glyph left behind when the thought collapses into a
+ * record, and nothing new appears at the transition.
+ *
+ * Glyphs only, no colour: intact under NO_COLOR and on any ground.
+ */
+export declare const STATUS_GLYPHS: readonly ["✧", "✦", "✶", "✸", "✺", "✸", "✦"];
 /**
  * The RUNNING row: the rotating glyph, the wall seconds since `since`
  * (never below 1 — a run that just started still reads "1s", so the row

package/dist/status.js

@@ -22,9 +22,22 @@
  * widens to name the new gesture.
  */
 import { kUnit } from "./render.js";
-/** v3 §03/§05 — the working glyph family; the CLI's 200ms spinner walks
- *  it and hands each glyph back to `runningStatus`. */
-export const STATUS_GLYPHS = ["▖", "▘", "▝", "▗"];
+import { TWINKLE } from "@vincemakes/kiso-tui-cells/render";
+/**
+ * R3 (design §5.2) — the working glyph family is the TWINKLE, and the
+ * CLI's 200ms spinner walks it exactly as it walked the four quadrant
+ * blocks it replaces.
+ *
+ * Two reasons the quadrants had to go. §5.3: they ROTATED, and a mark
+ * that turns implies progress a call of unpredictable duration does not
+ * have. §4.1: the twinkle settles onto `✦`, which is the mark the
+ * folded segment keeps — so the glyph a human watches while the model
+ * thinks is the glyph left behind when the thought collapses into a
+ * record, and nothing new appears at the transition.
+ *
+ * Glyphs only, no colour: intact under NO_COLOR and on any ground.
+ */
+export const STATUS_GLYPHS = TWINKLE;
 /** The ~ctx estimate as the whole-percent LEFT. A non-finite ratio (no
  *  window, no estimate) yields null and the row prints "~null%" — the
  *  long-standing shape, kept on purpose: an honest null beats an

package/package.json

@@ -1,6 +1,6 @@
 {
 	"name": "@vincemakes/kiso-tui",
-	"version": "0.16.4",
+	"version": "0.16.6",
 	"description": "kiso tui — the pure terminal layer (cell renderer, dock, raw editor, diff, palette). Zero runtime dependencies: input is data, output is bytes.",
 	"type": "module",
 	"license": "MIT",
@@ -35,6 +35,6 @@
 	},
 	"homepage": "https://github.com/vincemakes/kiso/tree/main/packages/tui#readme",
 	"dependencies": {
-		"@vincemakes/kiso-tui-cells": "0.16.4"
+		"@vincemakes/kiso-tui-cells": "0.16.6"
 	}
 }