diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 2ef64da8..83fef2b6 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -711,12 +711,11 @@ jobs:
       # invocation in the cross-platform `test` matrix does not carry a
       # Chrome dependency. This dedicated job opts in explicitly.
       - run: cargo test --test headless_zero_egress --locked -- --ignored
-      # RFC 4180 CSV parser correctness (cute-dbt#66) — also requires
-      # headless Chrome and reuses the Chrome install above. Splitting
-      # into a separate `cargo test` invocation keeps test names clean
-      # and isolates failure modes (a parser regression is unrelated
-      # to a network regression).
-      - run: cargo test --test headless_csv_parser --locked -- --ignored
+      # RFC 4180 CSV parser correctness (cute-dbt#66) moved to the Rust
+      # `parse_csv_rows` unit tests (`src/domain/unit_test_table.rs`,
+      # `g22`–`g26`) in cute-dbt#138: the JS `parseCsvRows` twin was retired
+      # when the Current view started rendering the Rust-computed FixtureTable
+      # POD, so there is no longer a headless CSV parser to exercise.
       # Updated-only ↔ All-tests toggle interaction (cute-dbt#91) — drives
       # the inlined report JS in a real browser (selector visibility,
       # toggle-dependent counts, 0-updated hint, auto-All landing). Reuses
diff --git a/AGENTS.md b/AGENTS.md
index bcd1a989..e9bb3b67 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -98,12 +98,14 @@ verified 2026-05-26:
 | csv    | array of dicts    | raw CSV string         |
 | sql    | raw SELECT string | raw SELECT string      |
 
-cute-dbt's JS renderer parses csv at render time so reports look
-identical regardless of which engine compiled the manifest
-(cute-dbt#66 — hand-rolled RFC 4180 parser, unit-tested via
-`tests/headless_csv_parser.rs`). The real engine-divergent surface
-elsewhere is YAML strictness: fusion rejects deprecated test-args
-that core only warns about.
+cute-dbt parses csv in the domain (`table_from_manifest_rows` →
+`parse_csv_rows`) so reports look identical regardless of which engine
+compiled the manifest (cute-dbt#66 — hand-rolled RFC 4180 parser, unit-tested
+via `src/domain/unit_test_table.rs` `g22`–`g26`). Since cute-dbt#138 the
+Current-view table renders the Rust-computed `FixtureTable` POD directly and
+the JS `parseCsvRows` twin is retired — the template is a pure renderer. The
+real engine-divergent surface elsewhere is YAML strictness: fusion rejects
+deprecated test-args that core only warns about.
 
 ## Working rules
 
diff --git a/examples/jaffle-shop-report.html b/examples/jaffle-shop-report.html
index 878494c1..c5e03043 100644
--- a/examples/jaffle-shop-report.html
+++ b/examples/jaffle-shop-report.html
@@ -658,7 +658,7 @@ <h2>Expected</h2>
 
 </div>
 
-<script type="application/json" id="cute-dbt-data">{"baseline":"tests/fixtures/jaffle-shop-baseline.json","models":[{"name":"stg_customers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"renamed","role":"transform"},{"id":"stg_customers","role":"final"}],"edges":[{"from":"source","to":"renamed","edge_type":"from"},{"from":"renamed","to":"stg_customers","edge_type":"from"}]},"compiled_sql":{"renamed":"renamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)","source":"source as (\n    select * from \"jaffle_shop\".\"main\".\"raw_customers\"\n\n)","stg_customers":"select * from renamed"},"raw_sql":"with source as (\n\n    {#-\n    Normally we would select from the table here, but we are using seeds to load\n    our data in this project. Modified body — adds a trim() pass on names so\n    PR 5's StateComparator sees the body checksum change.\n    #}\n    select * from {{ ref('raw_customers') }}\n\n),\n\nrenamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)\n\nselect * from renamed","tests":[{"id":"unit_test.jaffle_shop.stg_customers.test_stg_customers_renames_columns","name":"test_stg_customers_renames_columns","target_model":"stg_customers","changed":true,"description":"stg_customers renames raw_customers.id/first_name/last_name to\ncustomer_id/first_name/last_name and otherwise passes rows through.\nSynthetic fixture rows — no real customer data.\n","tags":["quality","smoke"],"meta":{"owner":"data-eng","priority":"high"},"defined_in":"models/staging/unit_tests.yml","given":[{"input":"ref('raw_customers')","bound_to_node":"source","rows":[{"first_name":"Alice","id":1,"last_name":"A."},{"first_name":"Bob","id":2,"last_name":"B."}],"format":"dict"}],"expected":{"rows":[{"customer_id":1,"first_name":"Alice","last_name":"A."},{"customer_id":2,"first_name":"Bob","last_name":"B."}],"format":"dict"}}],"is_recursive":false}]}</script>
+<script type="application/json" id="cute-dbt-data">{"baseline":"tests/fixtures/jaffle-shop-baseline.json","models":[{"name":"stg_customers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"renamed","role":"transform"},{"id":"stg_customers","role":"final"}],"edges":[{"from":"source","to":"renamed","edge_type":"from"},{"from":"renamed","to":"stg_customers","edge_type":"from"}]},"compiled_sql":{"renamed":"renamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)","source":"source as (\n    select * from \"jaffle_shop\".\"main\".\"raw_customers\"\n\n)","stg_customers":"select * from renamed"},"raw_sql":"with source as (\n\n    {#-\n    Normally we would select from the table here, but we are using seeds to load\n    our data in this project. Modified body — adds a trim() pass on names so\n    PR 5's StateComparator sees the body checksum change.\n    #}\n    select * from {{ ref('raw_customers') }}\n\n),\n\nrenamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)\n\nselect * from renamed","tests":[{"id":"unit_test.jaffle_shop.stg_customers.test_stg_customers_renames_columns","name":"test_stg_customers_renames_columns","target_model":"stg_customers","changed":true,"description":"stg_customers renames raw_customers.id/first_name/last_name to\ncustomer_id/first_name/last_name and otherwise passes rows through.\nSynthetic fixture rows — no real customer data.\n","tags":["quality","smoke"],"meta":{"owner":"data-eng","priority":"high"},"defined_in":"models/staging/unit_tests.yml","given":[{"input":"ref('raw_customers')","bound_to_node":"source","table":{"columns":["first_name","id","last_name"],"rows":[{"cells":[{"display":"Alice","key":{"t":"str","v":"Alice"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"A.","key":{"t":"str","v":"A."}}]},{"cells":[{"display":"Bob","key":{"t":"str","v":"Bob"}},{"display":"2","key":{"t":"number","v":"2"}},{"display":"B.","key":{"t":"str","v":"B."}}]}]},"rows":[{"first_name":"Alice","id":1,"last_name":"A."},{"first_name":"Bob","id":2,"last_name":"B."}],"format":"dict"}],"expected":{"table":{"columns":["customer_id","first_name","last_name"],"rows":[{"cells":[{"display":"1","key":{"t":"number","v":"1"}},{"display":"Alice","key":{"t":"str","v":"Alice"}},{"display":"A.","key":{"t":"str","v":"A."}}]},{"cells":[{"display":"2","key":{"t":"number","v":"2"}},{"display":"Bob","key":{"t":"str","v":"Bob"}},{"display":"B.","key":{"t":"str","v":"B."}}]}]},"rows":[{"customer_id":1,"first_name":"Alice","last_name":"A."},{"customer_id":2,"first_name":"Bob","last_name":"B."}],"format":"dict"}}],"is_recursive":false}]}</script>
 
 <script>
 /*! jQuery v3.7.1 | (c) OpenJS Foundation and other contributors | jquery.org/license */
@@ -4830,16 +4830,17 @@ <h2>Expected</h2>
       $sec.append(buildExternalFixtureAffordance(given.fixture));
       return $sec;
     }
-    var rows = extractRowsFromFixture(given);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = given.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
     if (given.format && given.format !== "dict") {
       $hdr.append($("<span>").addClass("format-badge").text("format: " + given.format));
     }
     $hdr.append($("<span>").addClass("row-count-badge")
-      .text(rows.length + " row" + (rows.length === 1 ? "" : "s")));
+      .text(rowCount + " row" + (rowCount === 1 ? "" : "s")));
     $sec.append($hdr);
-    var cols = collectColumns(rows);
     var diff = givenDataDiff(dataDiff, given.input);
-    $sec.append(buildFixtureView(cols, rows, "given-table", diff));
+    $sec.append(buildFixtureView(table, "given-table", diff));
     return $sec;
   }
 
@@ -4867,14 +4868,15 @@ <h2>Expected</h2>
       $body.append(buildExternalFixtureAffordance(t.expected.fixture));
       return;
     }
-    var rows = extractRowsFromFixture(t.expected);
-    $panel.find(".expected-rowcount").text(rows.length + " row" + (rows.length === 1 ? "" : "s"));
-    var cols = collectColumns(rows);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = t.expected.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
+    $panel.find(".expected-rowcount").text(rowCount + " row" + (rowCount === 1 ? "" : "s"));
     // The expect side's cell diff, when the test's expect fixture carried a
     // real change (cute-dbt#98). `data_diff.expect` already passed
     // `has_real_change()`, so its presence === "default this table to Diff".
     var diff = (t.data_diff && t.data_diff.expect) || undefined;
-    $body.append(buildFixtureView(cols, rows, "expected-table", diff));
+    $body.append(buildFixtureView(table, "expected-table", diff));
     initDataTablesIn($body);
   }
 
@@ -4898,26 +4900,13 @@ <h2>Expected</h2>
     return Boolean(obj && obj.format === "sql" && typeof obj.rows === "string");
   }
 
-  // Return a tabulable row array from a fixture, handling both
-  // array shapes (dict/csv-from-core) and raw-CSV-string shapes
-  // (csv-from-fusion). Returns `[]` for sql fixtures and other
-  // unexpected shapes — the sql path is dispatched separately by the
-  // caller via `isSqlStringFixture`.
-  function extractRowsFromFixture(fixture) {
-    if (!fixture) return [];
-    var rows = fixture.rows;
-    if (rows == null) return [];
-    if (Array.isArray(rows)) return rows;
-    if (typeof rows === "string" && fixture.format === "csv") {
-      return parseCsvRows(rows);
-    }
-    return [];
-  }
-
   // Render a sql-format fixture as a syntax-highlighted code block.
   // Reuses the compiled-SQL drawer's `.sql-block-wrap` shell so visual
   // style stays consistent. No copy button: the source of truth is the
-  // fixture in the manifest, not the rendered DOM.
+  // fixture in the manifest, not the rendered DOM. This is the ONE
+  // remaining non-tabulatable fallback path: a `format: sql` fixture has no
+  // cells, so the domain returns no `table` POD (cute-dbt#138) and the JS
+  // renders the raw SELECT here.
   function buildSqlCodeBlock(sql) {
     var $wrap = $("<div>").addClass("sql-block-wrap fixture-sql-block-wrap");
     var $pre = $("<pre>").addClass("sql-block").append(
@@ -4927,129 +4916,14 @@ <h2>Expected</h2>
     return $wrap;
   }
 
-  // RFC 4180 CSV parser — state machine over the input string.
-  //
-  // dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as
-  // a raw CSV body in the manifest. dbt-core 1.11+ pre-parses csv to
-  // an array of dicts. cute-dbt does the parse at render time so both
-  // engines render identically (cute-dbt#66).
-  //
-  // Behavior summary:
-  //   * First row is headers; subsequent rows are body values.
-  //   * Quoted fields strip the surrounding double-quotes.
-  //   * Double-quote inside a quoted field is the RFC 4180 escape:
-  //     `""` decodes to a literal `"`.
-  //   * Quoted fields may contain commas and newlines verbatim.
-  //   * A single trailing newline (LF or CRLF) is consumed — dbt YAML
-  //     `rows: |` block scalars always end in `\n`; preserving it
-  //     would create a spurious final empty row.
-  //   * An unterminated final row is accepted (graceful — the
-  //     renderer is not a validator).
-  //   * CRLF (`\r\n`) terminates a row as a single unit.
-  //   * Empty input or header-only input returns `[]`.
-  //
-  // Correctness is unit-tested via `tests/headless_csv_parser.rs` —
-  // the parser is exposed on `window.__cuteParseCsv` so that test can
-  // invoke it through the headless Chrome runtime.
-  function parseCsvRows(text) {
-    if (text == null) return [];
-    var t = String(text);
-    if (t.length === 0) return [];
-    // Strip exactly one trailing newline (LF or CRLF).
-    if (t.charAt(t.length - 1) === "\n") {
-      t = t.slice(0, -1);
-      if (t.length > 0 && t.charAt(t.length - 1) === "\r") {
-        t = t.slice(0, -1);
-      }
-    }
-    if (t.length === 0) return [];
-
-    var matrix = [];
-    var fields = [];
-    var field = "";
-    var inQuotes = false;
-    for (var i = 0; i < t.length; i++) {
-      var c = t.charAt(i);
-      if (inQuotes) {
-        if (c === '"') {
-          // RFC 4180 escape: doubled quote inside a quoted field
-          // collapses to a single literal quote.
-          if (i + 1 < t.length && t.charAt(i + 1) === '"') {
-            field += '"';
-            i++;
-            continue;
-          }
-          inQuotes = false;
-          continue;
-        }
-        field += c;
-        continue;
-      }
-      // Outside quotes.
-      if (c === '"' && field === "") {
-        inQuotes = true;
-        continue;
-      }
-      if (c === ',') {
-        fields.push(field);
-        field = "";
-        continue;
-      }
-      if (c === '\n' || c === '\r') {
-        fields.push(field);
-        matrix.push(fields);
-        fields = [];
-        field = "";
-        if (c === '\r' && i + 1 < t.length && t.charAt(i + 1) === '\n') {
-          i++;
-        }
-        continue;
-      }
-      field += c;
-    }
-    // Flush the final field/row. We always run this — the trailing-
-    // newline strip above guarantees at least one character of
-    // content survives to reach this branch in non-empty input.
-    fields.push(field);
-    matrix.push(fields);
-
-    if (matrix.length < 2) return [];
-    var headers = matrix[0];
-    var out = [];
-    for (var ri = 1; ri < matrix.length; ri++) {
-      var row = matrix[ri];
-      var obj = {};
-      for (var hi = 0; hi < headers.length; hi++) {
-        obj[headers[hi]] = hi < row.length ? row[hi] : "";
-      }
-      out.push(obj);
-    }
-    return out;
-  }
-
-  // Expose the parser on `window.__cuteParseCsv` so the headless
-  // integration test at `tests/headless_csv_parser.rs` can invoke it
-  // via `Runtime.Evaluate` and assert RFC 4180 edge cases (quoted
-  // fields, embedded commas, embedded newlines, escaped quotes,
-  // CRLF, empty input, header-only input). The `__` prefix marks it
-  // as a test-only seam; production code reaches it via the closure
-  // binding above.
-  if (typeof window !== "undefined") {
-    window.__cuteParseCsv = parseCsvRows;
-  }
-
-  function collectColumns(rows) {
-    var seen = {};
-    var order = [];
-    rows.forEach(function (row) {
-      if (row && typeof row === "object" && !Array.isArray(row)) {
-        Object.keys(row).forEach(function (k) {
-          if (!seen[k]) { seen[k] = true; order.push(k); }
-        });
-      }
-    });
-    return order;
-  }
+  // cute-dbt#138 — the JS csv/dict table parsers (`parseCsvRows`,
+  // `extractRowsFromFixture`, `collectColumns`) are RETIRED. The Current view
+  // now renders the Rust-computed `FixtureTable` POD that the domain ships on
+  // each given/expect payload (`table`), so csv/dict tabulation — and the
+  // engine-divergence handling above — happens once, in Rust
+  // (`table_from_manifest_rows`), and the template is a pure renderer. CSV
+  // RFC 4180 correctness is now owned by the Rust `parse_csv_rows` unit tests
+  // (the JS `__cuteParseCsv` headless seam retired with the function).
 
   function escapeHtml(s) {
     return String(s)
@@ -5406,7 +5280,7 @@ <h2>Expected</h2>
   function highlightYaml(yaml) { return emitTokens(tokenizeYaml(yaml)); }
 
   // Expose for headless JS unit tests (cute-dbt#69 / cute-dbt#132 — parallels
-  // the `__cuteParseCsv` seam introduced in cute-dbt#66).
+  // the `__cuteRenderYamlDiff` headless seam).
   window.__cuteHighlightYaml = highlightYaml;
   window.__cuteTokenizeSql = tokenizeSql;
   window.__cuteTokenizeYaml = tokenizeYaml;
@@ -5677,7 +5551,16 @@ <h2>Expected</h2>
     done(ok);
   }
 
-  function buildTable(columns, rows, cls) {
+  // The Current-view grid, rendered PURELY from the Rust-computed
+  // `FixtureTable` POD (cute-dbt#138). Each cell is a `{display, key}` Cell:
+  // the text is the authored `display` (so a csv `1.00` shows `1.00`, not the
+  // canonicalized `1`), while the numeric / NULL styling + sort order come
+  // from the canonical `key` (a `Number` key sorts numerically even when its
+  // display is `1.00`). The JS no longer parses csv/dict — the domain's
+  // `table_from_manifest_rows` owns that, and ships this POD on the wire.
+  function buildTable(table, cls) {
+    var columns = (table && table.columns) || [];
+    var rows = (table && table.rows) || [];
     var $wrap = $("<div>").addClass("table-fit");
     var $tbl = $("<table>").addClass(cls);
     var $thead = $("<thead>").appendTo($tbl);
@@ -5689,23 +5572,34 @@ <h2>Expected</h2>
     var $tbody = $("<tbody>").appendTo($tbl);
     rows.forEach(function (r) {
       var $tr = $("<tr>").appendTo($tbody);
-      columns.forEach(function (col) {
-        var cell = (r && typeof r === "object" && !Array.isArray(r)) ? r[col] : undefined;
-        var $td = $("<td>");
-        if (cell === null || cell === undefined) {
-          $td.addClass("cell-null").text("NULL");
-        } else if (typeof cell === "number") {
-          $td.addClass("cell-num").text(cell);
-        } else {
-          $td.text(String(cell));
-        }
-        $tr.append($td);
+      (r.cells || []).forEach(function (cell) {
+        $tr.append(decorateCurrentCell($("<td>"), cell));
       });
     });
     $wrap.append($tbl);
     return $wrap;
   }
 
+  // Render one Current-view Cell (`{display, key}`) into a <td>: the text is
+  // the authored `display`; the styling/sort affordance is read from `key.t`
+  // (the canonical type), never from the display token (cute-dbt#138). Shares
+  // the null/number styling with `decorateValueCell` so Current and Diff read
+  // identically.
+  function decorateCurrentCell($td, cell) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
+      return $td.addClass("cell-null").text("NULL");
+    }
+    if (key.t === "absent") {
+      return $td.addClass("cell-absent").text("");
+    }
+    if (key.t === "number") {
+      // Sort numerically on the canonical value; show the authored display.
+      return $td.addClass("cell-num").attr("data-order", key.v).text(cell.display);
+    }
+    return $td.text(cell.display);
+  }
+
   // cute-dbt#98 ----------------------------------------------------------
   //
   // The structured cell-diff renderer + the per-table Current↔Diff toggle.
@@ -5737,14 +5631,14 @@ <h2>Expected</h2>
   // is present — also build the Diff grid and a Current↔Diff toggle that flips
   // exactly one view visible at a time (the inactive view carries `hidden`).
   // Default-Diff when a diff exists, else Current with no toggle.
-  function buildFixtureView(columns, rows, cls, diff) {
+  function buildFixtureView(table, cls, diff) {
     if (!diff) {
-      // No cell diff for this table → today's behavior, byte-for-byte.
-      return buildTable(columns, rows, cls);
+      // No cell diff for this table → the plain Current grid from the POD.
+      return buildTable(table, cls);
     }
     var $wrap = $("<div>").addClass("fixture-view");
     var $diffView = buildDiffTable(diff, cls);
-    var $currentView = buildTable(columns, rows, cls).prop("hidden", true);
+    var $currentView = buildTable(table, cls).prop("hidden", true);
 
     var $toggle = $("<div>").addClass("cell-diff-toggle");
     var $diffBtn = $("<button>").attr("type", "button")
@@ -5817,67 +5711,56 @@ <h2>Expected</h2>
       return decorateValueCell($td, cell.old);
     }
     if (rowKind === "added" || rowKind === "unchanged" || !cell.changed) {
+      // Not flagged (added/unchanged, or a format-only change where the keys
+      // are equal): show the NEW authored value, no red/green (cute-dbt#138).
       return decorateValueCell($td, cell.new);
     }
-    // Modified + changed → inline old → new. Each side is type-aware via
-    // diffToken: a real null/absent keeps its own neutral styling and never
-    // borrows the value-change red/green (cute-dbt#132).
-    $td.addClass("cell-changed").attr("data-order", cellValueText(cell.new));
+    // Modified + changed → inline old → new, each side showing its authored
+    // display. Each side is type-aware via diffToken: a real null/absent keeps
+    // its own neutral styling and never borrows the value-change red/green
+    // (cute-dbt#132). The sort key is the NEW cell's authored display.
+    $td.addClass("cell-changed").attr("data-order", cellText(cell.new));
     $td.append(diffToken(cell.old, "cell-old"));
     $td.append($("<span>").addClass("cell-arrow").html("&rarr;"));
     $td.append(diffToken(cell.new, "cell-new"));
     return $td;
   }
 
-  // One side of an inline `old → new` changed cell. Type-aware so a real null
-  // or absent cell reads in its own neutral styling rather than the brilliant
-  // red/green reserved for a real value change: null → italic muted-gray
-  // "NULL"; absent → blank; any concrete value (including the string 'null')
-  // → the side's color via `sideClass` (.cell-old = red, .cell-new = green).
-  // This is what keeps a real NULL visually distinct from a string 'null'
-  // even mid-diff (cute-dbt#132).
-  function diffToken(cv, sideClass) {
-    if (!cv || cv.t === "null") {
+  // One side of an inline `old → new` changed cell. `cell` is a `{display,
+  // key}` Cell (cute-dbt#138): the text is the authored `display`; the
+  // neutral-vs-value styling is read from `key.t` so a real null or absent
+  // cell reads in its own neutral styling rather than the brilliant red/green
+  // reserved for a real value change: null → italic muted-gray "NULL"; absent
+  // → blank; any concrete value (including the string 'null') → the side's
+  // color via `sideClass` (.cell-old = red, .cell-new = green). This keeps a
+  // real NULL visually distinct from a string 'null' even mid-diff (#132).
+  function diffToken(cell, sideClass) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
       return $("<span>").addClass("cell-null").text("NULL");
     }
-    if (cv.t === "absent") {
+    if (key.t === "absent") {
       return $("<span>").addClass("cell-absent").text("");
     }
-    return $("<span>").addClass(sideClass).text(cellValueText(cv));
+    return $("<span>").addClass(sideClass).text(cell.display);
   }
 
-  // Render a single CellValue into a plain value cell, matching `buildTable`'s
-  // null/number styling so the Current and Diff views read identically.
-  function decorateValueCell($td, cv) {
-    if (!cv || cv.t === "null") {
-      // A literal null reads as NULL, matching the Current grid's
-      // `cell === null` styling.
-      return $td.addClass("cell-null").text("NULL");
-    }
-    if (cv.t === "absent") {
-      // An absent cell (a unified-diff column the row's source table lacks)
-      // is NOT a null — render it blank so an Absent->Null change stays
-      // visible and added/removed-column cells don't read as literal nulls.
-      return $td.addClass("cell-absent").text("");
-    }
-    if (cv.t === "number") {
-      return $td.addClass("cell-num").attr("data-order", cv.v).text(cv.v);
-    }
-    return $td.text(cellValueText(cv));
-  }
-
-  // The display text of an adjacently-tagged CellValue ({t, v}). `absent`
-  // renders empty; `null` renders the word NULL; bool/number/str verbatim.
-  function cellValueText(cv) {
-    if (!cv) return "";
-    switch (cv.t) {
-      case "null":   return "NULL";
-      case "absent": return "";
-      case "bool":   return cv.v ? "true" : "false";
-      case "number": return String(cv.v);
-      case "str":    return String(cv.v);
-      default:       return cv.v == null ? "" : String(cv.v);
-    }
+  // Render a single `{display, key}` Cell into a plain value cell, matching
+  // `buildTable`'s null/number styling so the Current and Diff views read
+  // identically. Text is the authored `display`; styling/sort from `key`.
+  function decorateValueCell($td, cell) {
+    return decorateCurrentCell($td, cell);
+  }
+
+  // The display text of a `{display, key}` Cell (cute-dbt#138): a `null` key
+  // renders the word NULL, an `absent` key renders empty, otherwise the
+  // authored `display` token verbatim. Used for the changed-cell sort key.
+  function cellText(cell) {
+    var key = cell && cell.key;
+    if (!key) return "";
+    if (key.t === "null") return "NULL";
+    if (key.t === "absent") return "";
+    return cell.display == null ? "" : String(cell.display);
   }
 
   // Re-measure DataTables column widths for any (now-visible) table under
@@ -5892,13 +5775,15 @@ <h2>Expected</h2>
     });
   }
   // Exposed for headless JS unit tests (cute-dbt#98 — parallels the
-  // `__cuteRenderYamlDiff` / `__cuteParseCsv` seams). Returns the outer
-  // wrapper element so a test can assert the toggle + tinted markup.
+  // `__cuteRenderYamlDiff` seam). Takes the Rust-computed `FixtureTable` POD
+  // (cute-dbt#138) + an optional cell diff and returns the outer wrapper
+  // element so a test can assert the authored-display rendering, the toggle,
+  // and the tinted markup.
   if (typeof window !== "undefined") {
-    window.__cuteBuildFixtureView = function (columns, rows, cls, diff) {
-      return buildFixtureView(columns, rows, cls, diff)[0];
+    window.__cuteBuildFixtureView = function (table, cls, diff) {
+      return buildFixtureView(table, cls, diff)[0];
     };
-    window.__cuteCellValueText = cellValueText;
+    window.__cuteCellText = cellText;
   }
 
   function initDataTablesIn($scope) {
diff --git a/examples/playground-pr-diff-report.html b/examples/playground-pr-diff-report.html
index c01905a8..57ad6bf7 100644
--- a/examples/playground-pr-diff-report.html
+++ b/examples/playground-pr-diff-report.html
@@ -667,7 +667,7 @@ <h2>Expected</h2>
 
 </div>
 
-<script type="application/json" id="cute-dbt-data">{"baseline":"","models":[{"name":"fct_provider_metrics","dag":{"nodes":[{"id":"encounters","role":"import"},{"id":"dim_date","role":"import"},{"id":"procedures","role":"import"},{"id":"encounters_with_date","role":"transform"},{"id":"provider_month_agg","role":"transform"},{"id":"final","role":"transform"},{"id":"fct_provider_metrics","role":"final"}],"edges":[{"from":"encounters","to":"encounters_with_date","edge_type":"from"},{"from":"dim_date","to":"encounters_with_date","edge_type":"inner"},{"from":"procedures","to":"encounters_with_date","edge_type":"left"},{"from":"encounters_with_date","to":"provider_month_agg","edge_type":"from"},{"from":"provider_month_agg","to":"final","edge_type":"from"},{"from":"final","to":"fct_provider_metrics","edge_type":"from"}]},"compiled_sql":{"dim_date":"dim_date as (\n    select\n        date_key\n        , year_actual\n        , month_actual\n    from \"memory\".\"main_marts\".\"dim_date\"\n)","encounters":"encounters as (\n    select\n        encounter_key\n        , encounter_id\n        , patient_key\n        , patient_id\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n        , encounter_start_date_key\n        , total_claim_cost\n        , payer_coverage\n        , duration_minutes\n        , patient_age_at_encounter\n    from \"memory\".\"main_marts\".\"fct_encounters\"\n    where provider_key != -1\n)","encounters_with_date":"encounters_with_date as (\n    select\n        e.*\n        , d.year_actual\n        , d.month_actual\n        , cast(d.year_actual as varchar)\n            || '-'\n            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str\n        , coalesce(pr.procedure_count, 0) as procedure_count\n    from encounters as e\n    inner join dim_date as d on e.encounter_start_date_key = d.date_key\n    left join procedures as pr on e.encounter_id = pr.encounter_id\n    where e.encounter_start_date_key != -1\n)","fct_provider_metrics":"select * from final","final":"final as (\n    select\n        -- surrogate key\n        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key\n\n        -- dimension keys\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n\n        -- time grain\n        , year_actual\n        , month_actual\n        , year_month_str\n\n        -- measures: counts\n        , encounter_count\n        , unique_patient_count\n        , total_procedures\n\n        -- measures: averages\n        , round(avg_patient_age, 2) as avg_patient_age\n        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes\n\n        -- measures: costs\n        , coalesce(total_cost, 0) as total_cost\n        , coalesce(total_payer_coverage, 0) as total_payer_coverage\n\n        -- measures: derived ratios\n        , case\n            when encounter_count > 0\n            then round(total_cost / encounter_count, 2)\n            else 0\n        end as avg_cost_per_encounter\n\n        , case\n            when unique_patient_count > 0\n            then round(cast(encounter_count as decimal) / unique_patient_count, 2)\n            else 0\n        end as encounters_per_patient_avg\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from provider_month_agg\n)","procedures":"procedures as (\n    select\n        encounter_id\n        , count(*) as procedure_count\n    from \"memory\".\"main_staging\".\"stg_synthea__procedures\"\n    group by encounter_id\n)","provider_month_agg":"provider_month_agg as (\n    select\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n\n        -- counts\n        , count(*) as encounter_count\n        , count(distinct e.patient_id) as unique_patient_count\n\n        -- patient metrics\n        , avg(e.patient_age_at_encounter) as avg_patient_age\n\n        -- encounter metrics\n        , avg(e.duration_minutes) as avg_encounter_duration_minutes\n\n        -- procedure metrics\n        , sum(e.procedure_count) as total_procedures\n\n        -- cost metrics\n        , sum(e.total_claim_cost) as total_cost\n        , sum(e.payer_coverage) as total_payer_coverage\n\n    from encounters_with_date as e\n    group by\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        unique_key='provider_metrics_key',\n        tags=['analytics', 'marts', 'fact']\n    )\n}}\n\n{#\n    Monthly Provider Metrics Fact\n\n    Purpose: Provider-level utilization and performance metrics\n    Grain: One row per provider per calendar month\n    Use cases: Provider performance dashboards, utilization trending, quality measures\n\n    Aggregates encounters, patients, procedures, and costs by provider and month.\n    Calculates derived metrics like avg cost per encounter and encounters per patient.\n#}\n\nwith encounters as (\n    select\n        encounter_key\n        , encounter_id\n        , patient_key\n        , patient_id\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n        , encounter_start_date_key\n        , total_claim_cost\n        , payer_coverage\n        , duration_minutes\n        , patient_age_at_encounter\n    from {{ ref('fct_encounters') }}\n    where provider_key != -1\n),\n\ndim_date as (\n    select\n        date_key\n        , year_actual\n        , month_actual\n    from {{ ref('dim_date') }}\n),\n\nprocedures as (\n    select\n        encounter_id\n        , count(*) as procedure_count\n    from {{ ref('stg_synthea__procedures') }}\n    group by encounter_id\n),\n\n-- join encounters to date dimension to get year/month\nencounters_with_date as (\n    select\n        e.*\n        , d.year_actual\n        , d.month_actual\n        , cast(d.year_actual as varchar)\n            || '-'\n            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str\n        , coalesce(pr.procedure_count, 0) as procedure_count\n    from encounters as e\n    inner join dim_date as d on e.encounter_start_date_key = d.date_key\n    left join procedures as pr on e.encounter_id = pr.encounter_id\n    where e.encounter_start_date_key != -1\n),\n\n-- aggregate by provider and year-month\nprovider_month_agg as (\n    select\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n\n        -- counts\n        , count(*) as encounter_count\n        , count(distinct e.patient_id) as unique_patient_count\n\n        -- patient metrics\n        , avg(e.patient_age_at_encounter) as avg_patient_age\n\n        -- encounter metrics\n        , avg(e.duration_minutes) as avg_encounter_duration_minutes\n\n        -- procedure metrics\n        , sum(e.procedure_count) as total_procedures\n\n        -- cost metrics\n        , sum(e.total_claim_cost) as total_cost\n        , sum(e.payer_coverage) as total_payer_coverage\n\n    from encounters_with_date as e\n    group by\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n),\n\nfinal as (\n    select\n        -- surrogate key\n        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key\n\n        -- dimension keys\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n\n        -- time grain\n        , year_actual\n        , month_actual\n        , year_month_str\n\n        -- measures: counts\n        , encounter_count\n        , unique_patient_count\n        , total_procedures\n\n        -- measures: averages\n        , round(avg_patient_age, 2) as avg_patient_age\n        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes\n\n        -- measures: costs\n        , coalesce(total_cost, 0) as total_cost\n        , coalesce(total_payer_coverage, 0) as total_payer_coverage\n\n        -- measures: derived ratios\n        , case\n            when encounter_count > 0\n            then round(total_cost / encounter_count, 2)\n            else 0\n        end as avg_cost_per_encounter\n\n        , case\n            when unique_patient_count > 0\n            then round(cast(encounter_count as decimal) / unique_patient_count, 2)\n            else 0\n        end as encounters_per_patient_avg\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from provider_month_agg\n)\n\nselect * from final","sql_diff":{"lines":[{"kind":"context","text":"{{","emphasis":null},{"kind":"context","text":"    config(","emphasis":null},{"kind":"context","text":"        materialized='table',","emphasis":null},{"kind":"context","text":"        unique_key='provider_metrics_key',","emphasis":null},{"kind":"context","text":"        tags=['analytics', 'marts', 'fact']","emphasis":null},{"kind":"context","text":"    )","emphasis":null},{"kind":"context","text":"}}","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"{#","emphasis":null},{"kind":"context","text":"    Monthly Provider Metrics Fact","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    Purpose: Provider-level utilization and performance metrics","emphasis":null},{"kind":"context","text":"    Grain: One row per provider per calendar month","emphasis":null},{"kind":"context","text":"    Use cases: Provider performance dashboards, utilization trending, quality measures","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    Aggregates encounters, patients, procedures, and costs by provider and month.","emphasis":null},{"kind":"context","text":"    Calculates derived metrics like avg cost per encounter and encounters per patient.","emphasis":null},{"kind":"context","text":"#}","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"with encounters as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        encounter_key","emphasis":null},{"kind":"context","text":"        , encounter_id","emphasis":null},{"kind":"context","text":"        , patient_key","emphasis":null},{"kind":"context","text":"        , patient_id","emphasis":null},{"kind":"context","text":"        , provider_key","emphasis":null},{"kind":"context","text":"        , provider_id","emphasis":null},{"kind":"context","text":"        , organization_key","emphasis":null},{"kind":"context","text":"        , organization_id","emphasis":null},{"kind":"context","text":"        , encounter_start_date_key","emphasis":null},{"kind":"context","text":"        , total_claim_cost","emphasis":null},{"kind":"context","text":"        , payer_coverage","emphasis":null},{"kind":"context","text":"        , duration_minutes","emphasis":null},{"kind":"context","text":"        , patient_age_at_encounter","emphasis":null},{"kind":"context","text":"    from {{ ref('fct_encounters') }}","emphasis":null},{"kind":"context","text":"    where provider_key != -1","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"dim_date as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        date_key","emphasis":null},{"kind":"context","text":"        , year_actual","emphasis":null},{"kind":"context","text":"        , month_actual","emphasis":null},{"kind":"context","text":"    from {{ ref('dim_date') }}","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"procedures as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        encounter_id","emphasis":null},{"kind":"removed","text":"        , count(distinct procedure_code) as procedure_count","emphasis":[16,39]},{"kind":"added","text":"        , count(*) as procedure_count","emphasis":[16,17]},{"kind":"context","text":"    from {{ ref('stg_synthea__procedures') }}","emphasis":null},{"kind":"context","text":"    group by encounter_id","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"-- join encounters to date dimension to get year/month","emphasis":null},{"kind":"context","text":"encounters_with_date as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        e.*","emphasis":null},{"kind":"context","text":"        , d.year_actual","emphasis":null},{"kind":"context","text":"        , d.month_actual","emphasis":null},{"kind":"context","text":"        , cast(d.year_actual as varchar)","emphasis":null},{"kind":"context","text":"            || '-'","emphasis":null},{"kind":"context","text":"            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str","emphasis":null},{"kind":"removed","text":"        , pr.procedure_count as procedure_count","emphasis":[10,28]},{"kind":"added","text":"        , coalesce(pr.procedure_count, 0) as procedure_count","emphasis":[10,41]},{"kind":"context","text":"    from encounters as e","emphasis":null},{"kind":"context","text":"    inner join dim_date as d on e.encounter_start_date_key = d.date_key","emphasis":null},{"kind":"context","text":"    left join procedures as pr on e.encounter_id = pr.encounter_id","emphasis":null},{"kind":"context","text":"    where e.encounter_start_date_key != -1","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"-- aggregate by provider and year-month","emphasis":null},{"kind":"context","text":"provider_month_agg as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        e.provider_id","emphasis":null},{"kind":"context","text":"        , e.provider_key","emphasis":null},{"kind":"context","text":"        , e.organization_id","emphasis":null},{"kind":"context","text":"        , e.organization_key","emphasis":null},{"kind":"context","text":"        , e.year_actual","emphasis":null},{"kind":"context","text":"        , e.month_actual","emphasis":null},{"kind":"context","text":"        , e.year_month_str","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- counts","emphasis":null},{"kind":"context","text":"        , count(*) as encounter_count","emphasis":null},{"kind":"context","text":"        , count(distinct e.patient_id) as unique_patient_count","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- patient metrics","emphasis":null},{"kind":"context","text":"        , avg(e.patient_age_at_encounter) as avg_patient_age","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- encounter metrics","emphasis":null},{"kind":"context","text":"        , avg(e.duration_minutes) as avg_encounter_duration_minutes","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- procedure metrics","emphasis":null},{"kind":"removed","text":"        , sum(e.procedure_count) as procedure_total","emphasis":[36,51]},{"kind":"added","text":"        , sum(e.procedure_count) as total_procedures","emphasis":[36,52]},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- cost metrics","emphasis":null},{"kind":"context","text":"        , sum(e.total_claim_cost) as total_cost","emphasis":null},{"kind":"context","text":"        , sum(e.payer_coverage) as total_payer_coverage","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    from encounters_with_date as e","emphasis":null},{"kind":"context","text":"    group by","emphasis":null},{"kind":"context","text":"        e.provider_id","emphasis":null},{"kind":"context","text":"        , e.provider_key","emphasis":null},{"kind":"context","text":"        , e.organization_id","emphasis":null},{"kind":"context","text":"        , e.organization_key","emphasis":null},{"kind":"context","text":"        , e.year_actual","emphasis":null},{"kind":"context","text":"        , e.month_actual","emphasis":null},{"kind":"context","text":"        , e.year_month_str","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"final as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        -- surrogate key","emphasis":null},{"kind":"context","text":"        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- dimension keys","emphasis":null},{"kind":"context","text":"        , provider_key","emphasis":null},{"kind":"context","text":"        , provider_id","emphasis":null},{"kind":"context","text":"        , organization_key","emphasis":null},{"kind":"context","text":"        , organization_id","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- time grain","emphasis":null},{"kind":"context","text":"        , year_actual","emphasis":null},{"kind":"context","text":"        , month_actual","emphasis":null},{"kind":"context","text":"        , year_month_str","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: counts","emphasis":null},{"kind":"context","text":"        , encounter_count","emphasis":null},{"kind":"context","text":"        , unique_patient_count","emphasis":null},{"kind":"context","text":"        , total_procedures","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: averages","emphasis":null},{"kind":"context","text":"        , round(avg_patient_age, 2) as avg_patient_age","emphasis":null},{"kind":"context","text":"        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: costs","emphasis":null},{"kind":"context","text":"        , coalesce(total_cost, 0) as total_cost","emphasis":null},{"kind":"context","text":"        , coalesce(total_payer_coverage, 0) as total_payer_coverage","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: derived ratios","emphasis":null},{"kind":"context","text":"        , case","emphasis":null},{"kind":"context","text":"            when encounter_count > 0","emphasis":null},{"kind":"context","text":"            then round(total_cost / encounter_count, 2)","emphasis":null},{"kind":"context","text":"            else 0","emphasis":null},{"kind":"context","text":"        end as avg_cost_per_encounter","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        , case","emphasis":null},{"kind":"context","text":"            when unique_patient_count > 0","emphasis":null},{"kind":"context","text":"            then round(cast(encounter_count as decimal) / unique_patient_count, 2)","emphasis":null},{"kind":"context","text":"            else 0","emphasis":null},{"kind":"context","text":"        end as encounters_per_patient_avg","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- metadata","emphasis":null},{"kind":"context","text":"        , current_timestamp as _loaded_at","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    from provider_month_agg","emphasis":null},{"kind":"context","text":")","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"select * from final","emphasis":null}]},"tests":[],"is_recursive":false},{"name":"mart_dq_summary","dag":{"nodes":[{"id":"encounter_metrics","role":"import"},{"id":"medication_metrics","role":"import"},{"id":"combined_metrics","role":"transform"},{"id":"mart_dq_summary","role":"final"}],"edges":[{"from":"encounter_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"medication_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"combined_metrics","to":"mart_dq_summary","edge_type":"from"}]},"compiled_sql":{"combined_metrics":"combined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)","encounter_metrics":"encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\n)","mart_dq_summary":"select\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","medication_metrics":"medication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from \"memory\".\"main_staging\".\"stg_synthea__medications\"\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['analytics', 'data_quality']\n    )\n}}\n\nwith encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from {{ ref('stg_synthea__encounters') }}\n),\n\nmedication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from {{ ref('stg_synthea__medications') }}\n),\n\ncombined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)\n\nselect\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_combines_encounter_and_medication_metrics","name":"test_mart_dq_summary_combines_encounter_and_medication_metrics","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\nand stg_synthea__medications. This test mocks two encounter rows\n(one invalid) and two medication rows (both invalid), then asserts\nthe per-entity counts + quarantine rate. Uses csv format for both\nGIVEN and EXPECT — compact tabular form across many boolean\ndq-flag columns.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 1 — the slicer attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test (no blank-line gap below) to this test's authored slice\n  # (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_combines_encounter_and_medication_metrics\n    description: |\n      mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\n      and stg_synthea__medications. This test mocks two encounter rows\n      (one invalid) and two medication rows (both invalid), then asserts\n      the per-entity counts + quarantine rate. Uses csv format for both\n      GIVEN and EXPECT — compact tabular form across many boolean\n      dq-flag columns.\n    model: mart_dq_summary\n    # INSIDE bracket for test 1 — comments at deeper indent than\n    # `- name:` ride along with the body slice as raw text (cute-dbt#69\n    # / #73).\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: csv\n        rows: |\n          is_dq_valid,valid_encounter_timestamps,no_future_encounter_dates,end_after_1900,start_after_1900\n          false,false,true,true,true\n          true,true,true,true,true\n      - input: ref('stg_synthea__medications')\n        format: csv\n        rows: |\n          is_dq_valid,valid_medication_dates,no_future_medication_dates,start_after_1900,end_after_1900_if_present\n          false,false,true,true,true\n          false,true,false,true,true\n    expect:\n      format: csv\n      rows: |\n        entity_type,quarantined_count,total_count,quarantine_rate_pct,failed_timestamp_validations\n        encounters,1,2,50.00,1\n        medications,2,2,100.00,1\n  # TRAILING bracket for test 1 — the comment block butts against the\n  # body above (no blank-line gap) and is terminated below by the blank\n  # line separating the two tests (cute-dbt#69 / #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":[{"end_after_1900":"true","is_dq_valid":"false","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"false"},{"end_after_1900":"true","is_dq_valid":"true","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"true"}],"format":"csv"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":[{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"true","start_after_1900":"true","valid_medication_dates":"false"},{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"false","start_after_1900":"true","valid_medication_dates":"true"}],"format":"csv"}],"expected":{"rows":[{"entity_type":"encounters","failed_timestamp_validations":"1","quarantine_rate_pct":"50.00","quarantined_count":"1","total_count":"2"},{"entity_type":"medications","failed_timestamp_validations":"1","quarantine_rate_pct":"100.00","quarantined_count":"2","total_count":"2"}],"format":"csv"}},{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_zero_quarantined_when_all_valid","name":"test_mart_dq_summary_zero_quarantined_when_all_valid","target_model":"mart_dq_summary","changed":true,"description":"mart_dq_summary should report zero quarantined_count and 0.00\nquarantine_rate_pct when every upstream row passes data-quality\nvalidations. Uses sql format for the GIVEN mocks and dict format\nfor the EXPECT — dict expects only compare listed columns, so\nnon-deterministic outputs (`_generated_at` from\n`current_timestamp`) are excluded from comparison.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 2 — sits between the blank-line\n  # terminator above and `- name:` below with no further gap, so the\n  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_zero_quarantined_when_all_valid\n    description: |\n      mart_dq_summary should report zero quarantined_count and 0.00\n      quarantine_rate_pct when every upstream row passes data-quality\n      validations. Uses sql format for the GIVEN mocks and dict format\n      for the EXPECT — dict expects only compare listed columns, so\n      non-deterministic outputs (`_generated_at` from\n      `current_timestamp`) are excluded from comparison.\n    model: mart_dq_summary\n    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /\n    # #73). Demonstrates that the slicer ride-along applies to every\n    # unit_test, not just the first in the file.\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_encounter_timestamps\n            , true as no_future_encounter_dates\n            , true as end_after_1900\n            , true as start_after_1900\n      - input: ref('stg_synthea__medications')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_medication_dates\n            , true as no_future_medication_dates\n            , true as start_after_1900\n            , true as end_after_1900_if_present\n    expect:\n      format: dict\n      rows:\n        - entity_type: 'encounters'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n        - entity_type: 'medications'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /\n  # #73).","yaml_diff":{"lines":[{"kind":"context","text":"  # LEADING bracket for test 2 — sits between the blank-line","emphasis":null},{"kind":"context","text":"  # terminator above and `- name:` below with no further gap, so the","emphasis":null},{"kind":"context","text":"  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).","emphasis":null},{"kind":"context","text":"  - name: test_mart_dq_summary_zero_quarantined_when_all_valid","emphasis":null},{"kind":"context","text":"    description: |","emphasis":null},{"kind":"context","text":"      mart_dq_summary should report zero quarantined_count and 0.00","emphasis":null},{"kind":"context","text":"      quarantine_rate_pct when every upstream row passes data-quality","emphasis":null},{"kind":"context","text":"      validations. Uses sql format for the GIVEN mocks and dict format","emphasis":null},{"kind":"context","text":"      for the EXPECT — dict expects only compare listed columns, so","emphasis":null},{"kind":"context","text":"      non-deterministic outputs (`_generated_at` from","emphasis":null},{"kind":"context","text":"      `current_timestamp`) are excluded from comparison.","emphasis":null},{"kind":"context","text":"    model: mart_dq_summary","emphasis":null},{"kind":"context","text":"    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /","emphasis":null},{"kind":"context","text":"    # #73). Demonstrates that the slicer ride-along applies to every","emphasis":null},{"kind":"context","text":"    # unit_test, not just the first in the file.","emphasis":null},{"kind":"context","text":"    given:","emphasis":null},{"kind":"context","text":"      - input: ref('stg_synthea__encounters')","emphasis":null},{"kind":"context","text":"        format: sql","emphasis":null},{"kind":"context","text":"        rows: |","emphasis":null},{"kind":"context","text":"          select","emphasis":null},{"kind":"context","text":"            true as is_dq_valid","emphasis":null},{"kind":"context","text":"            , true as valid_encounter_timestamps","emphasis":null},{"kind":"context","text":"            , true as no_future_encounter_dates","emphasis":null},{"kind":"context","text":"            , true as end_after_1900","emphasis":null},{"kind":"context","text":"            , true as start_after_1900","emphasis":null},{"kind":"context","text":"      - input: ref('stg_synthea__medications')","emphasis":null},{"kind":"context","text":"        format: sql","emphasis":null},{"kind":"context","text":"        rows: |","emphasis":null},{"kind":"context","text":"          select","emphasis":null},{"kind":"context","text":"            true as is_dq_valid","emphasis":null},{"kind":"context","text":"            , true as valid_medication_dates","emphasis":null},{"kind":"context","text":"            , true as no_future_medication_dates","emphasis":null},{"kind":"context","text":"            , true as start_after_1900","emphasis":null},{"kind":"context","text":"            , true as end_after_1900_if_present","emphasis":null},{"kind":"context","text":"    expect:","emphasis":null},{"kind":"context","text":"      format: dict","emphasis":null},{"kind":"context","text":"      rows:","emphasis":null},{"kind":"context","text":"        - entity_type: 'encounters'","emphasis":null},{"kind":"removed","text":"          quarantined_count: null","emphasis":[29,33]},{"kind":"removed","text":"          total_count: 2","emphasis":[23,24]},{"kind":"added","text":"          quarantined_count: 0","emphasis":[29,30]},{"kind":"added","text":"          total_count: 1","emphasis":[23,24]},{"kind":"context","text":"          quarantine_rate_pct: 0.00","emphasis":null},{"kind":"context","text":"        - entity_type: 'medications'","emphasis":null},{"kind":"removed","text":"          quarantined_count: 1","emphasis":[29,30]},{"kind":"removed","text":"          total_count: 2","emphasis":[23,24]},{"kind":"added","text":"          quarantined_count: 0","emphasis":[29,30]},{"kind":"added","text":"          total_count: 1","emphasis":[23,24]},{"kind":"context","text":"          quarantine_rate_pct: 0.00","emphasis":null},{"kind":"context","text":"  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /","emphasis":null},{"kind":"context","text":"  # #73).","emphasis":null}]},"data_diff":{"given":[],"expect":{"columns":[{"name":"entity_type","status":"present"},{"name":"quarantine_rate_pct","status":"present"},{"name":"quarantined_count","status":"present"},{"name":"total_count","status":"present"}],"rows":[{"kind":"modified","cells":[{"old":{"t":"str","v":"encounters"},"new":{"t":"str","v":"encounters"},"changed":false},{"old":{"t":"number","v":"0"},"new":{"t":"number","v":"0"},"changed":false},{"old":{"t":"null"},"new":{"t":"number","v":"0"},"changed":true},{"old":{"t":"number","v":"2"},"new":{"t":"number","v":"1"},"changed":true}]},{"kind":"modified","cells":[{"old":{"t":"str","v":"medications"},"new":{"t":"str","v":"medications"},"changed":false},{"old":{"t":"number","v":"0"},"new":{"t":"number","v":"0"},"changed":false},{"old":{"t":"number","v":"1"},"new":{"t":"number","v":"0"},"changed":true},{"old":{"t":"number","v":"2"},"new":{"t":"number","v":"1"},"changed":true}]}]}},"given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_encounter_timestamps\n  , true as no_future_encounter_dates\n  , true as end_after_1900\n  , true as start_after_1900","format":"sql"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_medication_dates\n  , true as no_future_medication_dates\n  , true as start_after_1900\n  , true as end_after_1900_if_present","format":"sql"}],"expected":{"rows":[{"entity_type":"encounters","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1},{"entity_type":"medications","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1}],"format":"dict"}}],"is_recursive":false}]}</script>
+<script type="application/json" id="cute-dbt-data">{"baseline":"","models":[{"name":"fct_provider_metrics","dag":{"nodes":[{"id":"encounters","role":"import"},{"id":"dim_date","role":"import"},{"id":"procedures","role":"import"},{"id":"encounters_with_date","role":"transform"},{"id":"provider_month_agg","role":"transform"},{"id":"final","role":"transform"},{"id":"fct_provider_metrics","role":"final"}],"edges":[{"from":"encounters","to":"encounters_with_date","edge_type":"from"},{"from":"dim_date","to":"encounters_with_date","edge_type":"inner"},{"from":"procedures","to":"encounters_with_date","edge_type":"left"},{"from":"encounters_with_date","to":"provider_month_agg","edge_type":"from"},{"from":"provider_month_agg","to":"final","edge_type":"from"},{"from":"final","to":"fct_provider_metrics","edge_type":"from"}]},"compiled_sql":{"dim_date":"dim_date as (\n    select\n        date_key\n        , year_actual\n        , month_actual\n    from \"memory\".\"main_marts\".\"dim_date\"\n)","encounters":"encounters as (\n    select\n        encounter_key\n        , encounter_id\n        , patient_key\n        , patient_id\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n        , encounter_start_date_key\n        , total_claim_cost\n        , payer_coverage\n        , duration_minutes\n        , patient_age_at_encounter\n    from \"memory\".\"main_marts\".\"fct_encounters\"\n    where provider_key != -1\n)","encounters_with_date":"encounters_with_date as (\n    select\n        e.*\n        , d.year_actual\n        , d.month_actual\n        , cast(d.year_actual as varchar)\n            || '-'\n            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str\n        , coalesce(pr.procedure_count, 0) as procedure_count\n    from encounters as e\n    inner join dim_date as d on e.encounter_start_date_key = d.date_key\n    left join procedures as pr on e.encounter_id = pr.encounter_id\n    where e.encounter_start_date_key != -1\n)","fct_provider_metrics":"select * from final","final":"final as (\n    select\n        -- surrogate key\n        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key\n\n        -- dimension keys\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n\n        -- time grain\n        , year_actual\n        , month_actual\n        , year_month_str\n\n        -- measures: counts\n        , encounter_count\n        , unique_patient_count\n        , total_procedures\n\n        -- measures: averages\n        , round(avg_patient_age, 2) as avg_patient_age\n        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes\n\n        -- measures: costs\n        , coalesce(total_cost, 0) as total_cost\n        , coalesce(total_payer_coverage, 0) as total_payer_coverage\n\n        -- measures: derived ratios\n        , case\n            when encounter_count > 0\n            then round(total_cost / encounter_count, 2)\n            else 0\n        end as avg_cost_per_encounter\n\n        , case\n            when unique_patient_count > 0\n            then round(cast(encounter_count as decimal) / unique_patient_count, 2)\n            else 0\n        end as encounters_per_patient_avg\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from provider_month_agg\n)","procedures":"procedures as (\n    select\n        encounter_id\n        , count(*) as procedure_count\n    from \"memory\".\"main_staging\".\"stg_synthea__procedures\"\n    group by encounter_id\n)","provider_month_agg":"provider_month_agg as (\n    select\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n\n        -- counts\n        , count(*) as encounter_count\n        , count(distinct e.patient_id) as unique_patient_count\n\n        -- patient metrics\n        , avg(e.patient_age_at_encounter) as avg_patient_age\n\n        -- encounter metrics\n        , avg(e.duration_minutes) as avg_encounter_duration_minutes\n\n        -- procedure metrics\n        , sum(e.procedure_count) as total_procedures\n\n        -- cost metrics\n        , sum(e.total_claim_cost) as total_cost\n        , sum(e.payer_coverage) as total_payer_coverage\n\n    from encounters_with_date as e\n    group by\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        unique_key='provider_metrics_key',\n        tags=['analytics', 'marts', 'fact']\n    )\n}}\n\n{#\n    Monthly Provider Metrics Fact\n\n    Purpose: Provider-level utilization and performance metrics\n    Grain: One row per provider per calendar month\n    Use cases: Provider performance dashboards, utilization trending, quality measures\n\n    Aggregates encounters, patients, procedures, and costs by provider and month.\n    Calculates derived metrics like avg cost per encounter and encounters per patient.\n#}\n\nwith encounters as (\n    select\n        encounter_key\n        , encounter_id\n        , patient_key\n        , patient_id\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n        , encounter_start_date_key\n        , total_claim_cost\n        , payer_coverage\n        , duration_minutes\n        , patient_age_at_encounter\n    from {{ ref('fct_encounters') }}\n    where provider_key != -1\n),\n\ndim_date as (\n    select\n        date_key\n        , year_actual\n        , month_actual\n    from {{ ref('dim_date') }}\n),\n\nprocedures as (\n    select\n        encounter_id\n        , count(*) as procedure_count\n    from {{ ref('stg_synthea__procedures') }}\n    group by encounter_id\n),\n\n-- join encounters to date dimension to get year/month\nencounters_with_date as (\n    select\n        e.*\n        , d.year_actual\n        , d.month_actual\n        , cast(d.year_actual as varchar)\n            || '-'\n            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str\n        , coalesce(pr.procedure_count, 0) as procedure_count\n    from encounters as e\n    inner join dim_date as d on e.encounter_start_date_key = d.date_key\n    left join procedures as pr on e.encounter_id = pr.encounter_id\n    where e.encounter_start_date_key != -1\n),\n\n-- aggregate by provider and year-month\nprovider_month_agg as (\n    select\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n\n        -- counts\n        , count(*) as encounter_count\n        , count(distinct e.patient_id) as unique_patient_count\n\n        -- patient metrics\n        , avg(e.patient_age_at_encounter) as avg_patient_age\n\n        -- encounter metrics\n        , avg(e.duration_minutes) as avg_encounter_duration_minutes\n\n        -- procedure metrics\n        , sum(e.procedure_count) as total_procedures\n\n        -- cost metrics\n        , sum(e.total_claim_cost) as total_cost\n        , sum(e.payer_coverage) as total_payer_coverage\n\n    from encounters_with_date as e\n    group by\n        e.provider_id\n        , e.provider_key\n        , e.organization_id\n        , e.organization_key\n        , e.year_actual\n        , e.month_actual\n        , e.year_month_str\n),\n\nfinal as (\n    select\n        -- surrogate key\n        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key\n\n        -- dimension keys\n        , provider_key\n        , provider_id\n        , organization_key\n        , organization_id\n\n        -- time grain\n        , year_actual\n        , month_actual\n        , year_month_str\n\n        -- measures: counts\n        , encounter_count\n        , unique_patient_count\n        , total_procedures\n\n        -- measures: averages\n        , round(avg_patient_age, 2) as avg_patient_age\n        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes\n\n        -- measures: costs\n        , coalesce(total_cost, 0) as total_cost\n        , coalesce(total_payer_coverage, 0) as total_payer_coverage\n\n        -- measures: derived ratios\n        , case\n            when encounter_count > 0\n            then round(total_cost / encounter_count, 2)\n            else 0\n        end as avg_cost_per_encounter\n\n        , case\n            when unique_patient_count > 0\n            then round(cast(encounter_count as decimal) / unique_patient_count, 2)\n            else 0\n        end as encounters_per_patient_avg\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from provider_month_agg\n)\n\nselect * from final","sql_diff":{"lines":[{"kind":"context","text":"{{","emphasis":null},{"kind":"context","text":"    config(","emphasis":null},{"kind":"context","text":"        materialized='table',","emphasis":null},{"kind":"context","text":"        unique_key='provider_metrics_key',","emphasis":null},{"kind":"context","text":"        tags=['analytics', 'marts', 'fact']","emphasis":null},{"kind":"context","text":"    )","emphasis":null},{"kind":"context","text":"}}","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"{#","emphasis":null},{"kind":"context","text":"    Monthly Provider Metrics Fact","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    Purpose: Provider-level utilization and performance metrics","emphasis":null},{"kind":"context","text":"    Grain: One row per provider per calendar month","emphasis":null},{"kind":"context","text":"    Use cases: Provider performance dashboards, utilization trending, quality measures","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    Aggregates encounters, patients, procedures, and costs by provider and month.","emphasis":null},{"kind":"context","text":"    Calculates derived metrics like avg cost per encounter and encounters per patient.","emphasis":null},{"kind":"context","text":"#}","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"with encounters as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        encounter_key","emphasis":null},{"kind":"context","text":"        , encounter_id","emphasis":null},{"kind":"context","text":"        , patient_key","emphasis":null},{"kind":"context","text":"        , patient_id","emphasis":null},{"kind":"context","text":"        , provider_key","emphasis":null},{"kind":"context","text":"        , provider_id","emphasis":null},{"kind":"context","text":"        , organization_key","emphasis":null},{"kind":"context","text":"        , organization_id","emphasis":null},{"kind":"context","text":"        , encounter_start_date_key","emphasis":null},{"kind":"context","text":"        , total_claim_cost","emphasis":null},{"kind":"context","text":"        , payer_coverage","emphasis":null},{"kind":"context","text":"        , duration_minutes","emphasis":null},{"kind":"context","text":"        , patient_age_at_encounter","emphasis":null},{"kind":"context","text":"    from {{ ref('fct_encounters') }}","emphasis":null},{"kind":"context","text":"    where provider_key != -1","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"dim_date as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        date_key","emphasis":null},{"kind":"context","text":"        , year_actual","emphasis":null},{"kind":"context","text":"        , month_actual","emphasis":null},{"kind":"context","text":"    from {{ ref('dim_date') }}","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"procedures as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        encounter_id","emphasis":null},{"kind":"removed","text":"        , count(distinct procedure_code) as procedure_count","emphasis":[16,39]},{"kind":"added","text":"        , count(*) as procedure_count","emphasis":[16,17]},{"kind":"context","text":"    from {{ ref('stg_synthea__procedures') }}","emphasis":null},{"kind":"context","text":"    group by encounter_id","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"-- join encounters to date dimension to get year/month","emphasis":null},{"kind":"context","text":"encounters_with_date as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        e.*","emphasis":null},{"kind":"context","text":"        , d.year_actual","emphasis":null},{"kind":"context","text":"        , d.month_actual","emphasis":null},{"kind":"context","text":"        , cast(d.year_actual as varchar)","emphasis":null},{"kind":"context","text":"            || '-'","emphasis":null},{"kind":"context","text":"            || lpad(cast(d.month_actual as varchar), 2, '0') as year_month_str","emphasis":null},{"kind":"removed","text":"        , pr.procedure_count as procedure_count","emphasis":[10,28]},{"kind":"added","text":"        , coalesce(pr.procedure_count, 0) as procedure_count","emphasis":[10,41]},{"kind":"context","text":"    from encounters as e","emphasis":null},{"kind":"context","text":"    inner join dim_date as d on e.encounter_start_date_key = d.date_key","emphasis":null},{"kind":"context","text":"    left join procedures as pr on e.encounter_id = pr.encounter_id","emphasis":null},{"kind":"context","text":"    where e.encounter_start_date_key != -1","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"-- aggregate by provider and year-month","emphasis":null},{"kind":"context","text":"provider_month_agg as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        e.provider_id","emphasis":null},{"kind":"context","text":"        , e.provider_key","emphasis":null},{"kind":"context","text":"        , e.organization_id","emphasis":null},{"kind":"context","text":"        , e.organization_key","emphasis":null},{"kind":"context","text":"        , e.year_actual","emphasis":null},{"kind":"context","text":"        , e.month_actual","emphasis":null},{"kind":"context","text":"        , e.year_month_str","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- counts","emphasis":null},{"kind":"context","text":"        , count(*) as encounter_count","emphasis":null},{"kind":"context","text":"        , count(distinct e.patient_id) as unique_patient_count","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- patient metrics","emphasis":null},{"kind":"context","text":"        , avg(e.patient_age_at_encounter) as avg_patient_age","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- encounter metrics","emphasis":null},{"kind":"context","text":"        , avg(e.duration_minutes) as avg_encounter_duration_minutes","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- procedure metrics","emphasis":null},{"kind":"removed","text":"        , sum(e.procedure_count) as procedure_total","emphasis":[36,51]},{"kind":"added","text":"        , sum(e.procedure_count) as total_procedures","emphasis":[36,52]},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- cost metrics","emphasis":null},{"kind":"context","text":"        , sum(e.total_claim_cost) as total_cost","emphasis":null},{"kind":"context","text":"        , sum(e.payer_coverage) as total_payer_coverage","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    from encounters_with_date as e","emphasis":null},{"kind":"context","text":"    group by","emphasis":null},{"kind":"context","text":"        e.provider_id","emphasis":null},{"kind":"context","text":"        , e.provider_key","emphasis":null},{"kind":"context","text":"        , e.organization_id","emphasis":null},{"kind":"context","text":"        , e.organization_key","emphasis":null},{"kind":"context","text":"        , e.year_actual","emphasis":null},{"kind":"context","text":"        , e.month_actual","emphasis":null},{"kind":"context","text":"        , e.year_month_str","emphasis":null},{"kind":"context","text":"),","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"final as (","emphasis":null},{"kind":"context","text":"    select","emphasis":null},{"kind":"context","text":"        -- surrogate key","emphasis":null},{"kind":"context","text":"        md5(cast(provider_id as varchar) || '-' || year_month_str) as provider_metrics_key","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- dimension keys","emphasis":null},{"kind":"context","text":"        , provider_key","emphasis":null},{"kind":"context","text":"        , provider_id","emphasis":null},{"kind":"context","text":"        , organization_key","emphasis":null},{"kind":"context","text":"        , organization_id","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- time grain","emphasis":null},{"kind":"context","text":"        , year_actual","emphasis":null},{"kind":"context","text":"        , month_actual","emphasis":null},{"kind":"context","text":"        , year_month_str","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: counts","emphasis":null},{"kind":"context","text":"        , encounter_count","emphasis":null},{"kind":"context","text":"        , unique_patient_count","emphasis":null},{"kind":"context","text":"        , total_procedures","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: averages","emphasis":null},{"kind":"context","text":"        , round(avg_patient_age, 2) as avg_patient_age","emphasis":null},{"kind":"context","text":"        , round(avg_encounter_duration_minutes, 2) as avg_encounter_duration_minutes","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: costs","emphasis":null},{"kind":"context","text":"        , coalesce(total_cost, 0) as total_cost","emphasis":null},{"kind":"context","text":"        , coalesce(total_payer_coverage, 0) as total_payer_coverage","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- measures: derived ratios","emphasis":null},{"kind":"context","text":"        , case","emphasis":null},{"kind":"context","text":"            when encounter_count > 0","emphasis":null},{"kind":"context","text":"            then round(total_cost / encounter_count, 2)","emphasis":null},{"kind":"context","text":"            else 0","emphasis":null},{"kind":"context","text":"        end as avg_cost_per_encounter","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        , case","emphasis":null},{"kind":"context","text":"            when unique_patient_count > 0","emphasis":null},{"kind":"context","text":"            then round(cast(encounter_count as decimal) / unique_patient_count, 2)","emphasis":null},{"kind":"context","text":"            else 0","emphasis":null},{"kind":"context","text":"        end as encounters_per_patient_avg","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"        -- metadata","emphasis":null},{"kind":"context","text":"        , current_timestamp as _loaded_at","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"    from provider_month_agg","emphasis":null},{"kind":"context","text":")","emphasis":null},{"kind":"context","text":"","emphasis":null},{"kind":"context","text":"select * from final","emphasis":null}]},"tests":[],"is_recursive":false},{"name":"mart_dq_summary","dag":{"nodes":[{"id":"encounter_metrics","role":"import"},{"id":"medication_metrics","role":"import"},{"id":"combined_metrics","role":"transform"},{"id":"mart_dq_summary","role":"final"}],"edges":[{"from":"encounter_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"medication_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"combined_metrics","to":"mart_dq_summary","edge_type":"from"}]},"compiled_sql":{"combined_metrics":"combined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)","encounter_metrics":"encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\n)","mart_dq_summary":"select\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","medication_metrics":"medication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from \"memory\".\"main_staging\".\"stg_synthea__medications\"\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['analytics', 'data_quality']\n    )\n}}\n\nwith encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from {{ ref('stg_synthea__encounters') }}\n),\n\nmedication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from {{ ref('stg_synthea__medications') }}\n),\n\ncombined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)\n\nselect\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_combines_encounter_and_medication_metrics","name":"test_mart_dq_summary_combines_encounter_and_medication_metrics","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\nand stg_synthea__medications. This test mocks two encounter rows\n(one invalid) and two medication rows (both invalid), then asserts\nthe per-entity counts + quarantine rate. Uses csv format for both\nGIVEN and EXPECT — compact tabular form across many boolean\ndq-flag columns.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 1 — the slicer attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test (no blank-line gap below) to this test's authored slice\n  # (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_combines_encounter_and_medication_metrics\n    description: |\n      mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\n      and stg_synthea__medications. This test mocks two encounter rows\n      (one invalid) and two medication rows (both invalid), then asserts\n      the per-entity counts + quarantine rate. Uses csv format for both\n      GIVEN and EXPECT — compact tabular form across many boolean\n      dq-flag columns.\n    model: mart_dq_summary\n    # INSIDE bracket for test 1 — comments at deeper indent than\n    # `- name:` ride along with the body slice as raw text (cute-dbt#69\n    # / #73).\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: csv\n        rows: |\n          is_dq_valid,valid_encounter_timestamps,no_future_encounter_dates,end_after_1900,start_after_1900\n          false,false,true,true,true\n          true,true,true,true,true\n      - input: ref('stg_synthea__medications')\n        format: csv\n        rows: |\n          is_dq_valid,valid_medication_dates,no_future_medication_dates,start_after_1900,end_after_1900_if_present\n          false,false,true,true,true\n          false,true,false,true,true\n    expect:\n      format: csv\n      rows: |\n        entity_type,quarantined_count,total_count,quarantine_rate_pct,failed_timestamp_validations\n        encounters,1,2,50.00,1\n        medications,2,2,100.00,1\n  # TRAILING bracket for test 1 — the comment block butts against the\n  # body above (no blank-line gap) and is terminated below by the blank\n  # line separating the two tests (cute-dbt#69 / #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","table":{"columns":["end_after_1900","is_dq_valid","no_future_encounter_dates","start_after_1900","valid_encounter_timestamps"],"rows":[{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}}]},{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}}]}]},"rows":[{"end_after_1900":"true","is_dq_valid":"false","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"false"},{"end_after_1900":"true","is_dq_valid":"true","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"true"}],"format":"csv"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","table":{"columns":["end_after_1900_if_present","is_dq_valid","no_future_medication_dates","start_after_1900","valid_medication_dates"],"rows":[{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}}]},{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}}]}]},"rows":[{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"true","start_after_1900":"true","valid_medication_dates":"false"},{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"false","start_after_1900":"true","valid_medication_dates":"true"}],"format":"csv"}],"expected":{"table":{"columns":["entity_type","failed_timestamp_validations","quarantine_rate_pct","quarantined_count","total_count"],"rows":[{"cells":[{"display":"encounters","key":{"t":"str","v":"encounters"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"50.00","key":{"t":"number","v":"50"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"2","key":{"t":"number","v":"2"}}]},{"cells":[{"display":"medications","key":{"t":"str","v":"medications"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"100.00","key":{"t":"number","v":"100"}},{"display":"2","key":{"t":"number","v":"2"}},{"display":"2","key":{"t":"number","v":"2"}}]}]},"rows":[{"entity_type":"encounters","failed_timestamp_validations":"1","quarantine_rate_pct":"50.00","quarantined_count":"1","total_count":"2"},{"entity_type":"medications","failed_timestamp_validations":"1","quarantine_rate_pct":"100.00","quarantined_count":"2","total_count":"2"}],"format":"csv"}},{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_zero_quarantined_when_all_valid","name":"test_mart_dq_summary_zero_quarantined_when_all_valid","target_model":"mart_dq_summary","changed":true,"description":"mart_dq_summary should report zero quarantined_count and 0.00\nquarantine_rate_pct when every upstream row passes data-quality\nvalidations. Uses sql format for the GIVEN mocks and dict format\nfor the EXPECT — dict expects only compare listed columns, so\nnon-deterministic outputs (`_generated_at` from\n`current_timestamp`) are excluded from comparison.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 2 — sits between the blank-line\n  # terminator above and `- name:` below with no further gap, so the\n  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_zero_quarantined_when_all_valid\n    description: |\n      mart_dq_summary should report zero quarantined_count and 0.00\n      quarantine_rate_pct when every upstream row passes data-quality\n      validations. Uses sql format for the GIVEN mocks and dict format\n      for the EXPECT — dict expects only compare listed columns, so\n      non-deterministic outputs (`_generated_at` from\n      `current_timestamp`) are excluded from comparison.\n    model: mart_dq_summary\n    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /\n    # #73). Demonstrates that the slicer ride-along applies to every\n    # unit_test, not just the first in the file.\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_encounter_timestamps\n            , true as no_future_encounter_dates\n            , true as end_after_1900\n            , true as start_after_1900\n      - input: ref('stg_synthea__medications')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_medication_dates\n            , true as no_future_medication_dates\n            , true as start_after_1900\n            , true as end_after_1900_if_present\n    expect:\n      format: dict\n      rows:\n        - entity_type: 'encounters'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n        - entity_type: 'medications'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /\n  # #73).","yaml_diff":{"lines":[{"kind":"context","text":"  # LEADING bracket for test 2 — sits between the blank-line","emphasis":null},{"kind":"context","text":"  # terminator above and `- name:` below with no further gap, so the","emphasis":null},{"kind":"context","text":"  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).","emphasis":null},{"kind":"context","text":"  - name: test_mart_dq_summary_zero_quarantined_when_all_valid","emphasis":null},{"kind":"context","text":"    description: |","emphasis":null},{"kind":"context","text":"      mart_dq_summary should report zero quarantined_count and 0.00","emphasis":null},{"kind":"context","text":"      quarantine_rate_pct when every upstream row passes data-quality","emphasis":null},{"kind":"context","text":"      validations. Uses sql format for the GIVEN mocks and dict format","emphasis":null},{"kind":"context","text":"      for the EXPECT — dict expects only compare listed columns, so","emphasis":null},{"kind":"context","text":"      non-deterministic outputs (`_generated_at` from","emphasis":null},{"kind":"context","text":"      `current_timestamp`) are excluded from comparison.","emphasis":null},{"kind":"context","text":"    model: mart_dq_summary","emphasis":null},{"kind":"context","text":"    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /","emphasis":null},{"kind":"context","text":"    # #73). Demonstrates that the slicer ride-along applies to every","emphasis":null},{"kind":"context","text":"    # unit_test, not just the first in the file.","emphasis":null},{"kind":"context","text":"    given:","emphasis":null},{"kind":"context","text":"      - input: ref('stg_synthea__encounters')","emphasis":null},{"kind":"context","text":"        format: sql","emphasis":null},{"kind":"context","text":"        rows: |","emphasis":null},{"kind":"context","text":"          select","emphasis":null},{"kind":"context","text":"            true as is_dq_valid","emphasis":null},{"kind":"context","text":"            , true as valid_encounter_timestamps","emphasis":null},{"kind":"context","text":"            , true as no_future_encounter_dates","emphasis":null},{"kind":"context","text":"            , true as end_after_1900","emphasis":null},{"kind":"context","text":"            , true as start_after_1900","emphasis":null},{"kind":"context","text":"      - input: ref('stg_synthea__medications')","emphasis":null},{"kind":"context","text":"        format: sql","emphasis":null},{"kind":"context","text":"        rows: |","emphasis":null},{"kind":"context","text":"          select","emphasis":null},{"kind":"context","text":"            true as is_dq_valid","emphasis":null},{"kind":"context","text":"            , true as valid_medication_dates","emphasis":null},{"kind":"context","text":"            , true as no_future_medication_dates","emphasis":null},{"kind":"context","text":"            , true as start_after_1900","emphasis":null},{"kind":"context","text":"            , true as end_after_1900_if_present","emphasis":null},{"kind":"context","text":"    expect:","emphasis":null},{"kind":"context","text":"      format: dict","emphasis":null},{"kind":"context","text":"      rows:","emphasis":null},{"kind":"context","text":"        - entity_type: 'encounters'","emphasis":null},{"kind":"removed","text":"          quarantined_count: null","emphasis":[29,33]},{"kind":"removed","text":"          total_count: 2","emphasis":[23,24]},{"kind":"added","text":"          quarantined_count: 0","emphasis":[29,30]},{"kind":"added","text":"          total_count: 1","emphasis":[23,24]},{"kind":"context","text":"          quarantine_rate_pct: 0.00","emphasis":null},{"kind":"context","text":"        - entity_type: 'medications'","emphasis":null},{"kind":"removed","text":"          quarantined_count: 1","emphasis":[29,30]},{"kind":"removed","text":"          total_count: 2","emphasis":[23,24]},{"kind":"added","text":"          quarantined_count: 0","emphasis":[29,30]},{"kind":"added","text":"          total_count: 1","emphasis":[23,24]},{"kind":"context","text":"          quarantine_rate_pct: 0.00","emphasis":null},{"kind":"context","text":"  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /","emphasis":null},{"kind":"context","text":"  # #73).","emphasis":null}]},"data_diff":{"given":[],"expect":{"columns":[{"name":"entity_type","status":"present"},{"name":"quarantine_rate_pct","status":"present"},{"name":"quarantined_count","status":"present"},{"name":"total_count","status":"present"}],"rows":[{"kind":"modified","cells":[{"old":{"display":"encounters","key":{"t":"str","v":"encounters"}},"new":{"display":"encounters","key":{"t":"str","v":"encounters"}},"changed":false},{"old":{"display":"0.00","key":{"t":"number","v":"0"}},"new":{"display":"0.0","key":{"t":"number","v":"0"}},"changed":false},{"old":{"display":"","key":{"t":"null"}},"new":{"display":"0","key":{"t":"number","v":"0"}},"changed":true},{"old":{"display":"2","key":{"t":"number","v":"2"}},"new":{"display":"1","key":{"t":"number","v":"1"}},"changed":true}]},{"kind":"modified","cells":[{"old":{"display":"medications","key":{"t":"str","v":"medications"}},"new":{"display":"medications","key":{"t":"str","v":"medications"}},"changed":false},{"old":{"display":"0.00","key":{"t":"number","v":"0"}},"new":{"display":"0.0","key":{"t":"number","v":"0"}},"changed":false},{"old":{"display":"1","key":{"t":"number","v":"1"}},"new":{"display":"0","key":{"t":"number","v":"0"}},"changed":true},{"old":{"display":"2","key":{"t":"number","v":"2"}},"new":{"display":"1","key":{"t":"number","v":"1"}},"changed":true}]}]}},"given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_encounter_timestamps\n  , true as no_future_encounter_dates\n  , true as end_after_1900\n  , true as start_after_1900","format":"sql"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_medication_dates\n  , true as no_future_medication_dates\n  , true as start_after_1900\n  , true as end_after_1900_if_present","format":"sql"}],"expected":{"table":{"columns":["entity_type","quarantine_rate_pct","quarantined_count","total_count"],"rows":[{"cells":[{"display":"encounters","key":{"t":"str","v":"encounters"}},{"display":"0.0","key":{"t":"number","v":"0"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"1","key":{"t":"number","v":"1"}}]},{"cells":[{"display":"medications","key":{"t":"str","v":"medications"}},{"display":"0.0","key":{"t":"number","v":"0"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"1","key":{"t":"number","v":"1"}}]}]},"rows":[{"entity_type":"encounters","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1},{"entity_type":"medications","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1}],"format":"dict"}}],"is_recursive":false}]}</script>
 
 <script>
 /*! jQuery v3.7.1 | (c) OpenJS Foundation and other contributors | jquery.org/license */
@@ -4839,16 +4839,17 @@ <h2>Expected</h2>
       $sec.append(buildExternalFixtureAffordance(given.fixture));
       return $sec;
     }
-    var rows = extractRowsFromFixture(given);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = given.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
     if (given.format && given.format !== "dict") {
       $hdr.append($("<span>").addClass("format-badge").text("format: " + given.format));
     }
     $hdr.append($("<span>").addClass("row-count-badge")
-      .text(rows.length + " row" + (rows.length === 1 ? "" : "s")));
+      .text(rowCount + " row" + (rowCount === 1 ? "" : "s")));
     $sec.append($hdr);
-    var cols = collectColumns(rows);
     var diff = givenDataDiff(dataDiff, given.input);
-    $sec.append(buildFixtureView(cols, rows, "given-table", diff));
+    $sec.append(buildFixtureView(table, "given-table", diff));
     return $sec;
   }
 
@@ -4876,14 +4877,15 @@ <h2>Expected</h2>
       $body.append(buildExternalFixtureAffordance(t.expected.fixture));
       return;
     }
-    var rows = extractRowsFromFixture(t.expected);
-    $panel.find(".expected-rowcount").text(rows.length + " row" + (rows.length === 1 ? "" : "s"));
-    var cols = collectColumns(rows);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = t.expected.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
+    $panel.find(".expected-rowcount").text(rowCount + " row" + (rowCount === 1 ? "" : "s"));
     // The expect side's cell diff, when the test's expect fixture carried a
     // real change (cute-dbt#98). `data_diff.expect` already passed
     // `has_real_change()`, so its presence === "default this table to Diff".
     var diff = (t.data_diff && t.data_diff.expect) || undefined;
-    $body.append(buildFixtureView(cols, rows, "expected-table", diff));
+    $body.append(buildFixtureView(table, "expected-table", diff));
     initDataTablesIn($body);
   }
 
@@ -4907,26 +4909,13 @@ <h2>Expected</h2>
     return Boolean(obj && obj.format === "sql" && typeof obj.rows === "string");
   }
 
-  // Return a tabulable row array from a fixture, handling both
-  // array shapes (dict/csv-from-core) and raw-CSV-string shapes
-  // (csv-from-fusion). Returns `[]` for sql fixtures and other
-  // unexpected shapes — the sql path is dispatched separately by the
-  // caller via `isSqlStringFixture`.
-  function extractRowsFromFixture(fixture) {
-    if (!fixture) return [];
-    var rows = fixture.rows;
-    if (rows == null) return [];
-    if (Array.isArray(rows)) return rows;
-    if (typeof rows === "string" && fixture.format === "csv") {
-      return parseCsvRows(rows);
-    }
-    return [];
-  }
-
   // Render a sql-format fixture as a syntax-highlighted code block.
   // Reuses the compiled-SQL drawer's `.sql-block-wrap` shell so visual
   // style stays consistent. No copy button: the source of truth is the
-  // fixture in the manifest, not the rendered DOM.
+  // fixture in the manifest, not the rendered DOM. This is the ONE
+  // remaining non-tabulatable fallback path: a `format: sql` fixture has no
+  // cells, so the domain returns no `table` POD (cute-dbt#138) and the JS
+  // renders the raw SELECT here.
   function buildSqlCodeBlock(sql) {
     var $wrap = $("<div>").addClass("sql-block-wrap fixture-sql-block-wrap");
     var $pre = $("<pre>").addClass("sql-block").append(
@@ -4936,129 +4925,14 @@ <h2>Expected</h2>
     return $wrap;
   }
 
-  // RFC 4180 CSV parser — state machine over the input string.
-  //
-  // dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as
-  // a raw CSV body in the manifest. dbt-core 1.11+ pre-parses csv to
-  // an array of dicts. cute-dbt does the parse at render time so both
-  // engines render identically (cute-dbt#66).
-  //
-  // Behavior summary:
-  //   * First row is headers; subsequent rows are body values.
-  //   * Quoted fields strip the surrounding double-quotes.
-  //   * Double-quote inside a quoted field is the RFC 4180 escape:
-  //     `""` decodes to a literal `"`.
-  //   * Quoted fields may contain commas and newlines verbatim.
-  //   * A single trailing newline (LF or CRLF) is consumed — dbt YAML
-  //     `rows: |` block scalars always end in `\n`; preserving it
-  //     would create a spurious final empty row.
-  //   * An unterminated final row is accepted (graceful — the
-  //     renderer is not a validator).
-  //   * CRLF (`\r\n`) terminates a row as a single unit.
-  //   * Empty input or header-only input returns `[]`.
-  //
-  // Correctness is unit-tested via `tests/headless_csv_parser.rs` —
-  // the parser is exposed on `window.__cuteParseCsv` so that test can
-  // invoke it through the headless Chrome runtime.
-  function parseCsvRows(text) {
-    if (text == null) return [];
-    var t = String(text);
-    if (t.length === 0) return [];
-    // Strip exactly one trailing newline (LF or CRLF).
-    if (t.charAt(t.length - 1) === "\n") {
-      t = t.slice(0, -1);
-      if (t.length > 0 && t.charAt(t.length - 1) === "\r") {
-        t = t.slice(0, -1);
-      }
-    }
-    if (t.length === 0) return [];
-
-    var matrix = [];
-    var fields = [];
-    var field = "";
-    var inQuotes = false;
-    for (var i = 0; i < t.length; i++) {
-      var c = t.charAt(i);
-      if (inQuotes) {
-        if (c === '"') {
-          // RFC 4180 escape: doubled quote inside a quoted field
-          // collapses to a single literal quote.
-          if (i + 1 < t.length && t.charAt(i + 1) === '"') {
-            field += '"';
-            i++;
-            continue;
-          }
-          inQuotes = false;
-          continue;
-        }
-        field += c;
-        continue;
-      }
-      // Outside quotes.
-      if (c === '"' && field === "") {
-        inQuotes = true;
-        continue;
-      }
-      if (c === ',') {
-        fields.push(field);
-        field = "";
-        continue;
-      }
-      if (c === '\n' || c === '\r') {
-        fields.push(field);
-        matrix.push(fields);
-        fields = [];
-        field = "";
-        if (c === '\r' && i + 1 < t.length && t.charAt(i + 1) === '\n') {
-          i++;
-        }
-        continue;
-      }
-      field += c;
-    }
-    // Flush the final field/row. We always run this — the trailing-
-    // newline strip above guarantees at least one character of
-    // content survives to reach this branch in non-empty input.
-    fields.push(field);
-    matrix.push(fields);
-
-    if (matrix.length < 2) return [];
-    var headers = matrix[0];
-    var out = [];
-    for (var ri = 1; ri < matrix.length; ri++) {
-      var row = matrix[ri];
-      var obj = {};
-      for (var hi = 0; hi < headers.length; hi++) {
-        obj[headers[hi]] = hi < row.length ? row[hi] : "";
-      }
-      out.push(obj);
-    }
-    return out;
-  }
-
-  // Expose the parser on `window.__cuteParseCsv` so the headless
-  // integration test at `tests/headless_csv_parser.rs` can invoke it
-  // via `Runtime.Evaluate` and assert RFC 4180 edge cases (quoted
-  // fields, embedded commas, embedded newlines, escaped quotes,
-  // CRLF, empty input, header-only input). The `__` prefix marks it
-  // as a test-only seam; production code reaches it via the closure
-  // binding above.
-  if (typeof window !== "undefined") {
-    window.__cuteParseCsv = parseCsvRows;
-  }
-
-  function collectColumns(rows) {
-    var seen = {};
-    var order = [];
-    rows.forEach(function (row) {
-      if (row && typeof row === "object" && !Array.isArray(row)) {
-        Object.keys(row).forEach(function (k) {
-          if (!seen[k]) { seen[k] = true; order.push(k); }
-        });
-      }
-    });
-    return order;
-  }
+  // cute-dbt#138 — the JS csv/dict table parsers (`parseCsvRows`,
+  // `extractRowsFromFixture`, `collectColumns`) are RETIRED. The Current view
+  // now renders the Rust-computed `FixtureTable` POD that the domain ships on
+  // each given/expect payload (`table`), so csv/dict tabulation — and the
+  // engine-divergence handling above — happens once, in Rust
+  // (`table_from_manifest_rows`), and the template is a pure renderer. CSV
+  // RFC 4180 correctness is now owned by the Rust `parse_csv_rows` unit tests
+  // (the JS `__cuteParseCsv` headless seam retired with the function).
 
   function escapeHtml(s) {
     return String(s)
@@ -5415,7 +5289,7 @@ <h2>Expected</h2>
   function highlightYaml(yaml) { return emitTokens(tokenizeYaml(yaml)); }
 
   // Expose for headless JS unit tests (cute-dbt#69 / cute-dbt#132 — parallels
-  // the `__cuteParseCsv` seam introduced in cute-dbt#66).
+  // the `__cuteRenderYamlDiff` headless seam).
   window.__cuteHighlightYaml = highlightYaml;
   window.__cuteTokenizeSql = tokenizeSql;
   window.__cuteTokenizeYaml = tokenizeYaml;
@@ -5686,7 +5560,16 @@ <h2>Expected</h2>
     done(ok);
   }
 
-  function buildTable(columns, rows, cls) {
+  // The Current-view grid, rendered PURELY from the Rust-computed
+  // `FixtureTable` POD (cute-dbt#138). Each cell is a `{display, key}` Cell:
+  // the text is the authored `display` (so a csv `1.00` shows `1.00`, not the
+  // canonicalized `1`), while the numeric / NULL styling + sort order come
+  // from the canonical `key` (a `Number` key sorts numerically even when its
+  // display is `1.00`). The JS no longer parses csv/dict — the domain's
+  // `table_from_manifest_rows` owns that, and ships this POD on the wire.
+  function buildTable(table, cls) {
+    var columns = (table && table.columns) || [];
+    var rows = (table && table.rows) || [];
     var $wrap = $("<div>").addClass("table-fit");
     var $tbl = $("<table>").addClass(cls);
     var $thead = $("<thead>").appendTo($tbl);
@@ -5698,23 +5581,34 @@ <h2>Expected</h2>
     var $tbody = $("<tbody>").appendTo($tbl);
     rows.forEach(function (r) {
       var $tr = $("<tr>").appendTo($tbody);
-      columns.forEach(function (col) {
-        var cell = (r && typeof r === "object" && !Array.isArray(r)) ? r[col] : undefined;
-        var $td = $("<td>");
-        if (cell === null || cell === undefined) {
-          $td.addClass("cell-null").text("NULL");
-        } else if (typeof cell === "number") {
-          $td.addClass("cell-num").text(cell);
-        } else {
-          $td.text(String(cell));
-        }
-        $tr.append($td);
+      (r.cells || []).forEach(function (cell) {
+        $tr.append(decorateCurrentCell($("<td>"), cell));
       });
     });
     $wrap.append($tbl);
     return $wrap;
   }
 
+  // Render one Current-view Cell (`{display, key}`) into a <td>: the text is
+  // the authored `display`; the styling/sort affordance is read from `key.t`
+  // (the canonical type), never from the display token (cute-dbt#138). Shares
+  // the null/number styling with `decorateValueCell` so Current and Diff read
+  // identically.
+  function decorateCurrentCell($td, cell) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
+      return $td.addClass("cell-null").text("NULL");
+    }
+    if (key.t === "absent") {
+      return $td.addClass("cell-absent").text("");
+    }
+    if (key.t === "number") {
+      // Sort numerically on the canonical value; show the authored display.
+      return $td.addClass("cell-num").attr("data-order", key.v).text(cell.display);
+    }
+    return $td.text(cell.display);
+  }
+
   // cute-dbt#98 ----------------------------------------------------------
   //
   // The structured cell-diff renderer + the per-table Current↔Diff toggle.
@@ -5746,14 +5640,14 @@ <h2>Expected</h2>
   // is present — also build the Diff grid and a Current↔Diff toggle that flips
   // exactly one view visible at a time (the inactive view carries `hidden`).
   // Default-Diff when a diff exists, else Current with no toggle.
-  function buildFixtureView(columns, rows, cls, diff) {
+  function buildFixtureView(table, cls, diff) {
     if (!diff) {
-      // No cell diff for this table → today's behavior, byte-for-byte.
-      return buildTable(columns, rows, cls);
+      // No cell diff for this table → the plain Current grid from the POD.
+      return buildTable(table, cls);
     }
     var $wrap = $("<div>").addClass("fixture-view");
     var $diffView = buildDiffTable(diff, cls);
-    var $currentView = buildTable(columns, rows, cls).prop("hidden", true);
+    var $currentView = buildTable(table, cls).prop("hidden", true);
 
     var $toggle = $("<div>").addClass("cell-diff-toggle");
     var $diffBtn = $("<button>").attr("type", "button")
@@ -5826,67 +5720,56 @@ <h2>Expected</h2>
       return decorateValueCell($td, cell.old);
     }
     if (rowKind === "added" || rowKind === "unchanged" || !cell.changed) {
+      // Not flagged (added/unchanged, or a format-only change where the keys
+      // are equal): show the NEW authored value, no red/green (cute-dbt#138).
       return decorateValueCell($td, cell.new);
     }
-    // Modified + changed → inline old → new. Each side is type-aware via
-    // diffToken: a real null/absent keeps its own neutral styling and never
-    // borrows the value-change red/green (cute-dbt#132).
-    $td.addClass("cell-changed").attr("data-order", cellValueText(cell.new));
+    // Modified + changed → inline old → new, each side showing its authored
+    // display. Each side is type-aware via diffToken: a real null/absent keeps
+    // its own neutral styling and never borrows the value-change red/green
+    // (cute-dbt#132). The sort key is the NEW cell's authored display.
+    $td.addClass("cell-changed").attr("data-order", cellText(cell.new));
     $td.append(diffToken(cell.old, "cell-old"));
     $td.append($("<span>").addClass("cell-arrow").html("&rarr;"));
     $td.append(diffToken(cell.new, "cell-new"));
     return $td;
   }
 
-  // One side of an inline `old → new` changed cell. Type-aware so a real null
-  // or absent cell reads in its own neutral styling rather than the brilliant
-  // red/green reserved for a real value change: null → italic muted-gray
-  // "NULL"; absent → blank; any concrete value (including the string 'null')
-  // → the side's color via `sideClass` (.cell-old = red, .cell-new = green).
-  // This is what keeps a real NULL visually distinct from a string 'null'
-  // even mid-diff (cute-dbt#132).
-  function diffToken(cv, sideClass) {
-    if (!cv || cv.t === "null") {
+  // One side of an inline `old → new` changed cell. `cell` is a `{display,
+  // key}` Cell (cute-dbt#138): the text is the authored `display`; the
+  // neutral-vs-value styling is read from `key.t` so a real null or absent
+  // cell reads in its own neutral styling rather than the brilliant red/green
+  // reserved for a real value change: null → italic muted-gray "NULL"; absent
+  // → blank; any concrete value (including the string 'null') → the side's
+  // color via `sideClass` (.cell-old = red, .cell-new = green). This keeps a
+  // real NULL visually distinct from a string 'null' even mid-diff (#132).
+  function diffToken(cell, sideClass) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
       return $("<span>").addClass("cell-null").text("NULL");
     }
-    if (cv.t === "absent") {
+    if (key.t === "absent") {
       return $("<span>").addClass("cell-absent").text("");
     }
-    return $("<span>").addClass(sideClass).text(cellValueText(cv));
+    return $("<span>").addClass(sideClass).text(cell.display);
   }
 
-  // Render a single CellValue into a plain value cell, matching `buildTable`'s
-  // null/number styling so the Current and Diff views read identically.
-  function decorateValueCell($td, cv) {
-    if (!cv || cv.t === "null") {
-      // A literal null reads as NULL, matching the Current grid's
-      // `cell === null` styling.
-      return $td.addClass("cell-null").text("NULL");
-    }
-    if (cv.t === "absent") {
-      // An absent cell (a unified-diff column the row's source table lacks)
-      // is NOT a null — render it blank so an Absent->Null change stays
-      // visible and added/removed-column cells don't read as literal nulls.
-      return $td.addClass("cell-absent").text("");
-    }
-    if (cv.t === "number") {
-      return $td.addClass("cell-num").attr("data-order", cv.v).text(cv.v);
-    }
-    return $td.text(cellValueText(cv));
-  }
-
-  // The display text of an adjacently-tagged CellValue ({t, v}). `absent`
-  // renders empty; `null` renders the word NULL; bool/number/str verbatim.
-  function cellValueText(cv) {
-    if (!cv) return "";
-    switch (cv.t) {
-      case "null":   return "NULL";
-      case "absent": return "";
-      case "bool":   return cv.v ? "true" : "false";
-      case "number": return String(cv.v);
-      case "str":    return String(cv.v);
-      default:       return cv.v == null ? "" : String(cv.v);
-    }
+  // Render a single `{display, key}` Cell into a plain value cell, matching
+  // `buildTable`'s null/number styling so the Current and Diff views read
+  // identically. Text is the authored `display`; styling/sort from `key`.
+  function decorateValueCell($td, cell) {
+    return decorateCurrentCell($td, cell);
+  }
+
+  // The display text of a `{display, key}` Cell (cute-dbt#138): a `null` key
+  // renders the word NULL, an `absent` key renders empty, otherwise the
+  // authored `display` token verbatim. Used for the changed-cell sort key.
+  function cellText(cell) {
+    var key = cell && cell.key;
+    if (!key) return "";
+    if (key.t === "null") return "NULL";
+    if (key.t === "absent") return "";
+    return cell.display == null ? "" : String(cell.display);
   }
 
   // Re-measure DataTables column widths for any (now-visible) table under
@@ -5901,13 +5784,15 @@ <h2>Expected</h2>
     });
   }
   // Exposed for headless JS unit tests (cute-dbt#98 — parallels the
-  // `__cuteRenderYamlDiff` / `__cuteParseCsv` seams). Returns the outer
-  // wrapper element so a test can assert the toggle + tinted markup.
+  // `__cuteRenderYamlDiff` seam). Takes the Rust-computed `FixtureTable` POD
+  // (cute-dbt#138) + an optional cell diff and returns the outer wrapper
+  // element so a test can assert the authored-display rendering, the toggle,
+  // and the tinted markup.
   if (typeof window !== "undefined") {
-    window.__cuteBuildFixtureView = function (columns, rows, cls, diff) {
-      return buildFixtureView(columns, rows, cls, diff)[0];
+    window.__cuteBuildFixtureView = function (table, cls, diff) {
+      return buildFixtureView(table, cls, diff)[0];
     };
-    window.__cuteCellValueText = cellValueText;
+    window.__cuteCellText = cellText;
   }
 
   function initDataTablesIn($scope) {
diff --git a/examples/playground-report.html b/examples/playground-report.html
index c24f952a..77ad544d 100644
--- a/examples/playground-report.html
+++ b/examples/playground-report.html
@@ -658,7 +658,7 @@ <h2>Expected</h2>
 
 </div>
 
-<script type="application/json" id="cute-dbt-data">{"baseline":"tests/fixtures/playground-baseline.json","models":[{"name":"dim_payers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"sequenced","role":"transform"},{"id":"unknown_member","role":"transform"},{"id":"final","role":"transform"},{"id":"dim_payers","role":"final"}],"edges":[{"from":"source","to":"sequenced","edge_type":"from"},{"from":"unknown_member","to":"final","edge_type":"union_all"},{"from":"sequenced","to":"final","edge_type":"union_all"},{"from":"final","to":"dim_payers","edge_type":"from"}]},"compiled_sql":{"dim_payers":"select * from final\n\n-- cute-dbt fixture: synthetic body mod","final":"final as (\n    select * from unknown_member\n    union all\n    select * from sequenced\n)","sequenced":"sequenced as (\n    select\n        -- surrogate key (starting from 1, reserving -1 for unknown)\n        row_number() over (order by payer_id) as payer_key\n\n        -- natural key\n        , payer_id\n\n        -- attributes\n        , payer_name\n        , address\n        , city\n        , state\n        , zip_code\n        , phone\n\n        -- financial metrics\n        , amount_covered\n        , amount_uncovered\n        , revenue\n\n        -- utilization metrics\n        , covered_encounters\n        , uncovered_encounters\n        , covered_medications\n        , uncovered_medications\n        , covered_procedures\n        , uncovered_procedures\n        , covered_immunizations\n        , uncovered_immunizations\n\n        -- member metrics\n        , unique_customers\n        , average_quality_of_life_score\n        , member_months\n\n        -- derived metrics\n        , case\n            when (amount_covered + amount_uncovered) > 0\n            then amount_covered / (amount_covered + amount_uncovered)\n            else 0\n        end as coverage_rate\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from source\n)","source":"source as (\n    select * from \"memory\".\"main_staging\".\"stg_synthea__payers\"\n)","unknown_member":"unknown_member as (\n    select\n        -1 as payer_key\n        , 'UNKNOWN' as payer_id\n        , 'Unknown Payer' as payer_name\n        , cast(null as varchar) as address\n        , cast(null as varchar) as city\n        , cast(null as varchar) as state\n        , cast(null as varchar) as zip_code\n        , cast(null as varchar) as phone\n        , cast(null as double) as amount_covered\n        , cast(null as double) as amount_uncovered\n        , cast(null as double) as revenue\n        , cast(null as bigint) as covered_encounters\n        , cast(null as bigint) as uncovered_encounters\n        , cast(null as bigint) as covered_medications\n        , cast(null as bigint) as uncovered_medications\n        , cast(null as bigint) as covered_procedures\n        , cast(null as bigint) as uncovered_procedures\n        , cast(null as bigint) as covered_immunizations\n        , cast(null as bigint) as uncovered_immunizations\n        , cast(null as bigint) as unique_customers\n        , cast(null as double) as average_quality_of_life_score\n        , cast(null as bigint) as member_months\n        , cast(null as double) as coverage_rate\n        , current_timestamp as _loaded_at\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        unique_key='payer_key'\n    )\n}}\n\nwith source as (\n    select * from {{ ref('stg_synthea__payers') }}\n),\n\nsequenced as (\n    select\n        -- surrogate key (starting from 1, reserving -1 for unknown)\n        row_number() over (order by payer_id) as payer_key\n\n        -- natural key\n        , payer_id\n\n        -- attributes\n        , payer_name\n        , address\n        , city\n        , state\n        , zip_code\n        , phone\n\n        -- financial metrics\n        , amount_covered\n        , amount_uncovered\n        , revenue\n\n        -- utilization metrics\n        , covered_encounters\n        , uncovered_encounters\n        , covered_medications\n        , uncovered_medications\n        , covered_procedures\n        , uncovered_procedures\n        , covered_immunizations\n        , uncovered_immunizations\n\n        -- member metrics\n        , unique_customers\n        , average_quality_of_life_score\n        , member_months\n\n        -- derived metrics\n        , case\n            when (amount_covered + amount_uncovered) > 0\n            then amount_covered / (amount_covered + amount_uncovered)\n            else 0\n        end as coverage_rate\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from source\n),\n\n-- unknown member for handling missing dimension lookups\nunknown_member as (\n    select\n        -1 as payer_key\n        , 'UNKNOWN' as payer_id\n        , 'Unknown Payer' as payer_name\n        , cast(null as varchar) as address\n        , cast(null as varchar) as city\n        , cast(null as varchar) as state\n        , cast(null as varchar) as zip_code\n        , cast(null as varchar) as phone\n        , cast(null as double) as amount_covered\n        , cast(null as double) as amount_uncovered\n        , cast(null as double) as revenue\n        , cast(null as bigint) as covered_encounters\n        , cast(null as bigint) as uncovered_encounters\n        , cast(null as bigint) as covered_medications\n        , cast(null as bigint) as uncovered_medications\n        , cast(null as bigint) as covered_procedures\n        , cast(null as bigint) as uncovered_procedures\n        , cast(null as bigint) as covered_immunizations\n        , cast(null as bigint) as uncovered_immunizations\n        , cast(null as bigint) as unique_customers\n        , cast(null as double) as average_quality_of_life_score\n        , cast(null as bigint) as member_months\n        , cast(null as double) as coverage_rate\n        , current_timestamp as _loaded_at\n),\n\nfinal as (\n    select * from unknown_member\n    union all\n    select * from sequenced\n)\n\nselect * from final\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.dim_payers.test_dim_payers_injects_unknown_sentinel","name":"test_dim_payers_injects_unknown_sentinel","target_model":"dim_payers","changed":false,"description":"dim_payers UNION-ALLs an unknown-member sentinel row (payer_key = -1)\nwith the sequenced source rows. This test verifies the sentinel\nsurvives the union and the sequenced rows are surrogate-keyed\nstarting at 1. Uses dbt's dict format for both GIVEN and EXPECT —\ndict format only compares listed columns, so non-deterministic\noutputs (`_loaded_at` via `current_timestamp`) are excluded from\nassertion automatically.\n","tags":[],"meta":{},"defined_in":"models/marts/core/_core__models.yml","authoring_yaml":"  # cute-dbt's authoring-YAML drawer (since #69) attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test to the test's authored slice. This block demonstrates the\n  # LEADING bracket — useful for stating the test's intent before the\n  # reader sees the assertion (cute-dbt#73).\n  - name: test_dim_payers_injects_unknown_sentinel\n    description: |\n      dim_payers UNION-ALLs an unknown-member sentinel row (payer_key = -1)\n      with the sequenced source rows. This test verifies the sentinel\n      survives the union and the sequenced rows are surrogate-keyed\n      starting at 1. Uses dbt's dict format for both GIVEN and EXPECT —\n      dict format only compares listed columns, so non-deterministic\n      outputs (`_loaded_at` via `current_timestamp`) are excluded from\n      assertion automatically.\n    model: dim_payers\n    # The INSIDE bracket — comments at deeper indent than `- name:` ride\n    # along with the body slice as raw text. Cute-dbt preserves them\n    # verbatim in the Authoring YAML drawer (cute-dbt#73).\n    given:\n      - input: ref('stg_synthea__payers')\n        format: dict\n        rows:\n          - payer_id: 'a00000aa-0000-0000-0000-000000000001'\n            payer_name: 'Acme Health'\n            address: '1 Acme Way'\n            city: 'Boston'\n            state: 'MA'\n            zip_code: '02101'\n            phone: '555-0100'\n            amount_covered: 1000.0\n            amount_uncovered: 250.0\n            revenue: 1250.0\n            covered_encounters: 10\n            uncovered_encounters: 2\n            covered_medications: 5\n            uncovered_medications: 1\n            covered_procedures: 3\n            uncovered_procedures: 0\n            covered_immunizations: 4\n            uncovered_immunizations: 0\n            unique_customers: 8\n            average_quality_of_life_score: 0.85\n            member_months: 96\n    expect:\n      format: dict\n      rows:\n        - payer_key: -1\n          payer_id: 'UNKNOWN'\n          payer_name: 'Unknown Payer'\n        - payer_key: 1\n          payer_id: 'a00000aa-0000-0000-0000-000000000001'\n          payer_name: 'Acme Health'\n  # The TRAILING bracket — contiguous `#`-comment lines at the SAME\n  # indent as `- name:`, immediately following the test body and\n  # terminated by EOF or a blank line, are attached to this test's\n  # slice. Useful for cross-references and follow-up notes (cute-dbt#73).","given":[{"input":"ref('stg_synthea__payers')","bound_to_node":"source","rows":[{"address":"1 Acme Way","amount_covered":1000.0,"amount_uncovered":250.0,"average_quality_of_life_score":0.85,"city":"Boston","covered_encounters":10,"covered_immunizations":4,"covered_medications":5,"covered_procedures":3,"member_months":96,"payer_id":"a00000aa-0000-0000-0000-000000000001","payer_name":"Acme Health","phone":"555-0100","revenue":1250.0,"state":"MA","uncovered_encounters":2,"uncovered_immunizations":0,"uncovered_medications":1,"uncovered_procedures":0,"unique_customers":8,"zip_code":"02101"}],"format":"dict"}],"expected":{"rows":[{"payer_id":"UNKNOWN","payer_key":-1,"payer_name":"Unknown Payer"},{"payer_id":"a00000aa-0000-0000-0000-000000000001","payer_key":1,"payer_name":"Acme Health"}],"format":"dict"}}],"is_recursive":false},{"name":"int_dq_quarantine__encounters","dag":{"nodes":[],"edges":[]},"compiled_sql":{"int_dq_quarantine__encounters":"\n\n\n\n-- Encounters quarantined due to data quality violations\n\nselect * from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\nwhere is_dq_valid = false\n\n\n-- cute-dbt fixture: synthetic body mod"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['intermediate', 'quarantine', 'data_quality']\n    )\n}}\n\n{{ generate_quarantine_model(\n    source_model='stg_synthea__encounters',\n    description='Encounters quarantined due to data quality violations'\n) }}\n\n-- cute-dbt fixture: synthetic body mod","tests":[],"is_recursive":false},{"name":"mart_dq_summary","dag":{"nodes":[{"id":"encounter_metrics","role":"import"},{"id":"medication_metrics","role":"import"},{"id":"combined_metrics","role":"transform"},{"id":"mart_dq_summary","role":"final"}],"edges":[{"from":"encounter_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"medication_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"combined_metrics","to":"mart_dq_summary","edge_type":"from"}]},"compiled_sql":{"combined_metrics":"combined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)","encounter_metrics":"encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\n)","mart_dq_summary":"select\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","medication_metrics":"medication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from \"memory\".\"main_staging\".\"stg_synthea__medications\"\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['analytics', 'data_quality']\n    )\n}}\n\nwith encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from {{ ref('stg_synthea__encounters') }}\n),\n\nmedication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from {{ ref('stg_synthea__medications') }}\n),\n\ncombined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)\n\nselect\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_combines_encounter_and_medication_metrics","name":"test_mart_dq_summary_combines_encounter_and_medication_metrics","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\nand stg_synthea__medications. This test mocks two encounter rows\n(one invalid) and two medication rows (both invalid), then asserts\nthe per-entity counts + quarantine rate. Uses csv format for both\nGIVEN and EXPECT — compact tabular form across many boolean\ndq-flag columns.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 1 — the slicer attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test (no blank-line gap below) to this test's authored slice\n  # (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_combines_encounter_and_medication_metrics\n    description: |\n      mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\n      and stg_synthea__medications. This test mocks two encounter rows\n      (one invalid) and two medication rows (both invalid), then asserts\n      the per-entity counts + quarantine rate. Uses csv format for both\n      GIVEN and EXPECT — compact tabular form across many boolean\n      dq-flag columns.\n    model: mart_dq_summary\n    # INSIDE bracket for test 1 — comments at deeper indent than\n    # `- name:` ride along with the body slice as raw text (cute-dbt#69\n    # / #73).\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: csv\n        rows: |\n          is_dq_valid,valid_encounter_timestamps,no_future_encounter_dates,end_after_1900,start_after_1900\n          false,false,true,true,true\n          true,true,true,true,true\n      - input: ref('stg_synthea__medications')\n        format: csv\n        rows: |\n          is_dq_valid,valid_medication_dates,no_future_medication_dates,start_after_1900,end_after_1900_if_present\n          false,false,true,true,true\n          false,true,false,true,true\n    expect:\n      format: csv\n      rows: |\n        entity_type,quarantined_count,total_count,quarantine_rate_pct,failed_timestamp_validations\n        encounters,1,2,50.00,1\n        medications,2,2,100.00,1\n  # TRAILING bracket for test 1 — the comment block butts against the\n  # body above (no blank-line gap) and is terminated below by the blank\n  # line separating the two tests (cute-dbt#69 / #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":[{"end_after_1900":"true","is_dq_valid":"false","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"false"},{"end_after_1900":"true","is_dq_valid":"true","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"true"}],"format":"csv"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":[{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"true","start_after_1900":"true","valid_medication_dates":"false"},{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"false","start_after_1900":"true","valid_medication_dates":"true"}],"format":"csv"}],"expected":{"rows":[{"entity_type":"encounters","failed_timestamp_validations":"1","quarantine_rate_pct":"50.00","quarantined_count":"1","total_count":"2"},{"entity_type":"medications","failed_timestamp_validations":"1","quarantine_rate_pct":"100.00","quarantined_count":"2","total_count":"2"}],"format":"csv"}},{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_zero_quarantined_when_all_valid","name":"test_mart_dq_summary_zero_quarantined_when_all_valid","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary should report zero quarantined_count and 0.00\nquarantine_rate_pct when every upstream row passes data-quality\nvalidations. Uses sql format for the GIVEN mocks and dict format\nfor the EXPECT — dict expects only compare listed columns, so\nnon-deterministic outputs (`_generated_at` from\n`current_timestamp`) are excluded from comparison.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 2 — sits between the blank-line\n  # terminator above and `- name:` below with no further gap, so the\n  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_zero_quarantined_when_all_valid\n    description: |\n      mart_dq_summary should report zero quarantined_count and 0.00\n      quarantine_rate_pct when every upstream row passes data-quality\n      validations. Uses sql format for the GIVEN mocks and dict format\n      for the EXPECT — dict expects only compare listed columns, so\n      non-deterministic outputs (`_generated_at` from\n      `current_timestamp`) are excluded from comparison.\n    model: mart_dq_summary\n    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /\n    # #73). Demonstrates that the slicer ride-along applies to every\n    # unit_test, not just the first in the file.\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_encounter_timestamps\n            , true as no_future_encounter_dates\n            , true as end_after_1900\n            , true as start_after_1900\n      - input: ref('stg_synthea__medications')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_medication_dates\n            , true as no_future_medication_dates\n            , true as start_after_1900\n            , true as end_after_1900_if_present\n    expect:\n      format: dict\n      rows:\n        - entity_type: 'encounters'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n        - entity_type: 'medications'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /\n  # #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_encounter_timestamps\n  , true as no_future_encounter_dates\n  , true as end_after_1900\n  , true as start_after_1900","format":"sql"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_medication_dates\n  , true as no_future_medication_dates\n  , true as start_after_1900\n  , true as end_after_1900_if_present","format":"sql"}],"expected":{"rows":[{"entity_type":"encounters","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1},{"entity_type":"medications","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1}],"format":"dict"}}],"is_recursive":false}]}</script>
+<script type="application/json" id="cute-dbt-data">{"baseline":"tests/fixtures/playground-baseline.json","models":[{"name":"dim_payers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"sequenced","role":"transform"},{"id":"unknown_member","role":"transform"},{"id":"final","role":"transform"},{"id":"dim_payers","role":"final"}],"edges":[{"from":"source","to":"sequenced","edge_type":"from"},{"from":"unknown_member","to":"final","edge_type":"union_all"},{"from":"sequenced","to":"final","edge_type":"union_all"},{"from":"final","to":"dim_payers","edge_type":"from"}]},"compiled_sql":{"dim_payers":"select * from final\n\n-- cute-dbt fixture: synthetic body mod","final":"final as (\n    select * from unknown_member\n    union all\n    select * from sequenced\n)","sequenced":"sequenced as (\n    select\n        -- surrogate key (starting from 1, reserving -1 for unknown)\n        row_number() over (order by payer_id) as payer_key\n\n        -- natural key\n        , payer_id\n\n        -- attributes\n        , payer_name\n        , address\n        , city\n        , state\n        , zip_code\n        , phone\n\n        -- financial metrics\n        , amount_covered\n        , amount_uncovered\n        , revenue\n\n        -- utilization metrics\n        , covered_encounters\n        , uncovered_encounters\n        , covered_medications\n        , uncovered_medications\n        , covered_procedures\n        , uncovered_procedures\n        , covered_immunizations\n        , uncovered_immunizations\n\n        -- member metrics\n        , unique_customers\n        , average_quality_of_life_score\n        , member_months\n\n        -- derived metrics\n        , case\n            when (amount_covered + amount_uncovered) > 0\n            then amount_covered / (amount_covered + amount_uncovered)\n            else 0\n        end as coverage_rate\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from source\n)","source":"source as (\n    select * from \"memory\".\"main_staging\".\"stg_synthea__payers\"\n)","unknown_member":"unknown_member as (\n    select\n        -1 as payer_key\n        , 'UNKNOWN' as payer_id\n        , 'Unknown Payer' as payer_name\n        , cast(null as varchar) as address\n        , cast(null as varchar) as city\n        , cast(null as varchar) as state\n        , cast(null as varchar) as zip_code\n        , cast(null as varchar) as phone\n        , cast(null as double) as amount_covered\n        , cast(null as double) as amount_uncovered\n        , cast(null as double) as revenue\n        , cast(null as bigint) as covered_encounters\n        , cast(null as bigint) as uncovered_encounters\n        , cast(null as bigint) as covered_medications\n        , cast(null as bigint) as uncovered_medications\n        , cast(null as bigint) as covered_procedures\n        , cast(null as bigint) as uncovered_procedures\n        , cast(null as bigint) as covered_immunizations\n        , cast(null as bigint) as uncovered_immunizations\n        , cast(null as bigint) as unique_customers\n        , cast(null as double) as average_quality_of_life_score\n        , cast(null as bigint) as member_months\n        , cast(null as double) as coverage_rate\n        , current_timestamp as _loaded_at\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        unique_key='payer_key'\n    )\n}}\n\nwith source as (\n    select * from {{ ref('stg_synthea__payers') }}\n),\n\nsequenced as (\n    select\n        -- surrogate key (starting from 1, reserving -1 for unknown)\n        row_number() over (order by payer_id) as payer_key\n\n        -- natural key\n        , payer_id\n\n        -- attributes\n        , payer_name\n        , address\n        , city\n        , state\n        , zip_code\n        , phone\n\n        -- financial metrics\n        , amount_covered\n        , amount_uncovered\n        , revenue\n\n        -- utilization metrics\n        , covered_encounters\n        , uncovered_encounters\n        , covered_medications\n        , uncovered_medications\n        , covered_procedures\n        , uncovered_procedures\n        , covered_immunizations\n        , uncovered_immunizations\n\n        -- member metrics\n        , unique_customers\n        , average_quality_of_life_score\n        , member_months\n\n        -- derived metrics\n        , case\n            when (amount_covered + amount_uncovered) > 0\n            then amount_covered / (amount_covered + amount_uncovered)\n            else 0\n        end as coverage_rate\n\n        -- metadata\n        , current_timestamp as _loaded_at\n\n    from source\n),\n\n-- unknown member for handling missing dimension lookups\nunknown_member as (\n    select\n        -1 as payer_key\n        , 'UNKNOWN' as payer_id\n        , 'Unknown Payer' as payer_name\n        , cast(null as varchar) as address\n        , cast(null as varchar) as city\n        , cast(null as varchar) as state\n        , cast(null as varchar) as zip_code\n        , cast(null as varchar) as phone\n        , cast(null as double) as amount_covered\n        , cast(null as double) as amount_uncovered\n        , cast(null as double) as revenue\n        , cast(null as bigint) as covered_encounters\n        , cast(null as bigint) as uncovered_encounters\n        , cast(null as bigint) as covered_medications\n        , cast(null as bigint) as uncovered_medications\n        , cast(null as bigint) as covered_procedures\n        , cast(null as bigint) as uncovered_procedures\n        , cast(null as bigint) as covered_immunizations\n        , cast(null as bigint) as uncovered_immunizations\n        , cast(null as bigint) as unique_customers\n        , cast(null as double) as average_quality_of_life_score\n        , cast(null as bigint) as member_months\n        , cast(null as double) as coverage_rate\n        , current_timestamp as _loaded_at\n),\n\nfinal as (\n    select * from unknown_member\n    union all\n    select * from sequenced\n)\n\nselect * from final\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.dim_payers.test_dim_payers_injects_unknown_sentinel","name":"test_dim_payers_injects_unknown_sentinel","target_model":"dim_payers","changed":false,"description":"dim_payers UNION-ALLs an unknown-member sentinel row (payer_key = -1)\nwith the sequenced source rows. This test verifies the sentinel\nsurvives the union and the sequenced rows are surrogate-keyed\nstarting at 1. Uses dbt's dict format for both GIVEN and EXPECT —\ndict format only compares listed columns, so non-deterministic\noutputs (`_loaded_at` via `current_timestamp`) are excluded from\nassertion automatically.\n","tags":[],"meta":{},"defined_in":"models/marts/core/_core__models.yml","authoring_yaml":"  # cute-dbt's authoring-YAML drawer (since #69) attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test to the test's authored slice. This block demonstrates the\n  # LEADING bracket — useful for stating the test's intent before the\n  # reader sees the assertion (cute-dbt#73).\n  - name: test_dim_payers_injects_unknown_sentinel\n    description: |\n      dim_payers UNION-ALLs an unknown-member sentinel row (payer_key = -1)\n      with the sequenced source rows. This test verifies the sentinel\n      survives the union and the sequenced rows are surrogate-keyed\n      starting at 1. Uses dbt's dict format for both GIVEN and EXPECT —\n      dict format only compares listed columns, so non-deterministic\n      outputs (`_loaded_at` via `current_timestamp`) are excluded from\n      assertion automatically.\n    model: dim_payers\n    # The INSIDE bracket — comments at deeper indent than `- name:` ride\n    # along with the body slice as raw text. Cute-dbt preserves them\n    # verbatim in the Authoring YAML drawer (cute-dbt#73).\n    given:\n      - input: ref('stg_synthea__payers')\n        format: dict\n        rows:\n          - payer_id: 'a00000aa-0000-0000-0000-000000000001'\n            payer_name: 'Acme Health'\n            address: '1 Acme Way'\n            city: 'Boston'\n            state: 'MA'\n            zip_code: '02101'\n            phone: '555-0100'\n            amount_covered: 1000.0\n            amount_uncovered: 250.0\n            revenue: 1250.0\n            covered_encounters: 10\n            uncovered_encounters: 2\n            covered_medications: 5\n            uncovered_medications: 1\n            covered_procedures: 3\n            uncovered_procedures: 0\n            covered_immunizations: 4\n            uncovered_immunizations: 0\n            unique_customers: 8\n            average_quality_of_life_score: 0.85\n            member_months: 96\n    expect:\n      format: dict\n      rows:\n        - payer_key: -1\n          payer_id: 'UNKNOWN'\n          payer_name: 'Unknown Payer'\n        - payer_key: 1\n          payer_id: 'a00000aa-0000-0000-0000-000000000001'\n          payer_name: 'Acme Health'\n  # The TRAILING bracket — contiguous `#`-comment lines at the SAME\n  # indent as `- name:`, immediately following the test body and\n  # terminated by EOF or a blank line, are attached to this test's\n  # slice. Useful for cross-references and follow-up notes (cute-dbt#73).","given":[{"input":"ref('stg_synthea__payers')","bound_to_node":"source","table":{"columns":["address","amount_covered","amount_uncovered","average_quality_of_life_score","city","covered_encounters","covered_immunizations","covered_medications","covered_procedures","member_months","payer_id","payer_name","phone","revenue","state","uncovered_encounters","uncovered_immunizations","uncovered_medications","uncovered_procedures","unique_customers","zip_code"],"rows":[{"cells":[{"display":"1 Acme Way","key":{"t":"str","v":"1 Acme Way"}},{"display":"1000.0","key":{"t":"number","v":"1000"}},{"display":"250.0","key":{"t":"number","v":"250"}},{"display":"0.85","key":{"t":"number","v":"0.85"}},{"display":"Boston","key":{"t":"str","v":"Boston"}},{"display":"10","key":{"t":"number","v":"10"}},{"display":"4","key":{"t":"number","v":"4"}},{"display":"5","key":{"t":"number","v":"5"}},{"display":"3","key":{"t":"number","v":"3"}},{"display":"96","key":{"t":"number","v":"96"}},{"display":"a00000aa-0000-0000-0000-000000000001","key":{"t":"str","v":"a00000aa-0000-0000-0000-000000000001"}},{"display":"Acme Health","key":{"t":"str","v":"Acme Health"}},{"display":"555-0100","key":{"t":"str","v":"555-0100"}},{"display":"1250.0","key":{"t":"number","v":"1250"}},{"display":"MA","key":{"t":"str","v":"MA"}},{"display":"2","key":{"t":"number","v":"2"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"8","key":{"t":"number","v":"8"}},{"display":"02101","key":{"t":"str","v":"02101"}}]}]},"rows":[{"address":"1 Acme Way","amount_covered":1000.0,"amount_uncovered":250.0,"average_quality_of_life_score":0.85,"city":"Boston","covered_encounters":10,"covered_immunizations":4,"covered_medications":5,"covered_procedures":3,"member_months":96,"payer_id":"a00000aa-0000-0000-0000-000000000001","payer_name":"Acme Health","phone":"555-0100","revenue":1250.0,"state":"MA","uncovered_encounters":2,"uncovered_immunizations":0,"uncovered_medications":1,"uncovered_procedures":0,"unique_customers":8,"zip_code":"02101"}],"format":"dict"}],"expected":{"table":{"columns":["payer_id","payer_key","payer_name"],"rows":[{"cells":[{"display":"UNKNOWN","key":{"t":"str","v":"UNKNOWN"}},{"display":"-1","key":{"t":"number","v":"-1"}},{"display":"Unknown Payer","key":{"t":"str","v":"Unknown Payer"}}]},{"cells":[{"display":"a00000aa-0000-0000-0000-000000000001","key":{"t":"str","v":"a00000aa-0000-0000-0000-000000000001"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"Acme Health","key":{"t":"str","v":"Acme Health"}}]}]},"rows":[{"payer_id":"UNKNOWN","payer_key":-1,"payer_name":"Unknown Payer"},{"payer_id":"a00000aa-0000-0000-0000-000000000001","payer_key":1,"payer_name":"Acme Health"}],"format":"dict"}}],"is_recursive":false},{"name":"int_dq_quarantine__encounters","dag":{"nodes":[],"edges":[]},"compiled_sql":{"int_dq_quarantine__encounters":"\n\n\n\n-- Encounters quarantined due to data quality violations\n\nselect * from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\nwhere is_dq_valid = false\n\n\n-- cute-dbt fixture: synthetic body mod"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['intermediate', 'quarantine', 'data_quality']\n    )\n}}\n\n{{ generate_quarantine_model(\n    source_model='stg_synthea__encounters',\n    description='Encounters quarantined due to data quality violations'\n) }}\n\n-- cute-dbt fixture: synthetic body mod","tests":[],"is_recursive":false},{"name":"mart_dq_summary","dag":{"nodes":[{"id":"encounter_metrics","role":"import"},{"id":"medication_metrics","role":"import"},{"id":"combined_metrics","role":"transform"},{"id":"mart_dq_summary","role":"final"}],"edges":[{"from":"encounter_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"medication_metrics","to":"combined_metrics","edge_type":"union_all"},{"from":"combined_metrics","to":"mart_dq_summary","edge_type":"from"}]},"compiled_sql":{"combined_metrics":"combined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)","encounter_metrics":"encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from \"memory\".\"main_staging\".\"stg_synthea__encounters\"\n)","mart_dq_summary":"select\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","medication_metrics":"medication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from \"memory\".\"main_staging\".\"stg_synthea__medications\"\n)"},"raw_sql":"{{\n    config(\n        materialized='table',\n        tags=['analytics', 'data_quality']\n    )\n}}\n\nwith encounter_metrics as (\n    select\n        'encounters' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_encounter_timestamps) as failed_valid_encounter_timestamps\n        , count(*) filter (where not no_future_encounter_dates) as failed_no_future_encounter_dates\n        , count(*) filter (where not end_after_1900) as failed_end_after_1900\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n    from {{ ref('stg_synthea__encounters') }}\n),\n\nmedication_metrics as (\n    select\n        'medications' as entity_type\n        , count(*) filter (where is_dq_valid = false) as quarantined_count\n        , count(*) as total_count\n        , round(100.0 * count(*) filter (where is_dq_valid = false) / count(*), 2) as quarantine_rate_pct\n        , count(*) filter (where not valid_medication_dates) as failed_valid_medication_dates\n        , count(*) filter (where not no_future_medication_dates) as failed_no_future_medication_dates\n        , count(*) filter (where not start_after_1900) as failed_start_after_1900\n        , count(*) filter (where not end_after_1900_if_present) as failed_end_after_1900_if_present\n    from {{ ref('stg_synthea__medications') }}\n),\n\ncombined_metrics as (\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_encounter_timestamps as failed_timestamp_validations\n        , failed_no_future_encounter_dates as failed_future_date_validations\n        , failed_end_after_1900 as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from encounter_metrics\n\n    union all\n\n    select\n        entity_type\n        , quarantined_count\n        , total_count\n        , quarantine_rate_pct\n        , failed_valid_medication_dates as failed_timestamp_validations\n        , failed_no_future_medication_dates as failed_future_date_validations\n        , failed_end_after_1900_if_present as failed_historical_end_validations\n        , failed_start_after_1900 as failed_historical_start_validations\n    from medication_metrics\n)\n\nselect\n    entity_type\n    , quarantined_count\n    , total_count\n    , quarantine_rate_pct\n    , failed_timestamp_validations\n    , failed_future_date_validations\n    , failed_historical_end_validations\n    , failed_historical_start_validations\n    , current_timestamp as _generated_at\nfrom combined_metrics\norder by entity_type\n\n-- cute-dbt fixture: synthetic body mod","tests":[{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_combines_encounter_and_medication_metrics","name":"test_mart_dq_summary_combines_encounter_and_medication_metrics","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\nand stg_synthea__medications. This test mocks two encounter rows\n(one invalid) and two medication rows (both invalid), then asserts\nthe per-entity counts + quarantine rate. Uses csv format for both\nGIVEN and EXPECT — compact tabular form across many boolean\ndq-flag columns.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 1 — the slicer attaches contiguous\n  # `#`-comment lines at the SAME indent as `- name:` immediately above\n  # the test (no blank-line gap below) to this test's authored slice\n  # (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_combines_encounter_and_medication_metrics\n    description: |\n      mart_dq_summary UNION-ALLs metrics from stg_synthea__encounters\n      and stg_synthea__medications. This test mocks two encounter rows\n      (one invalid) and two medication rows (both invalid), then asserts\n      the per-entity counts + quarantine rate. Uses csv format for both\n      GIVEN and EXPECT — compact tabular form across many boolean\n      dq-flag columns.\n    model: mart_dq_summary\n    # INSIDE bracket for test 1 — comments at deeper indent than\n    # `- name:` ride along with the body slice as raw text (cute-dbt#69\n    # / #73).\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: csv\n        rows: |\n          is_dq_valid,valid_encounter_timestamps,no_future_encounter_dates,end_after_1900,start_after_1900\n          false,false,true,true,true\n          true,true,true,true,true\n      - input: ref('stg_synthea__medications')\n        format: csv\n        rows: |\n          is_dq_valid,valid_medication_dates,no_future_medication_dates,start_after_1900,end_after_1900_if_present\n          false,false,true,true,true\n          false,true,false,true,true\n    expect:\n      format: csv\n      rows: |\n        entity_type,quarantined_count,total_count,quarantine_rate_pct,failed_timestamp_validations\n        encounters,1,2,50.00,1\n        medications,2,2,100.00,1\n  # TRAILING bracket for test 1 — the comment block butts against the\n  # body above (no blank-line gap) and is terminated below by the blank\n  # line separating the two tests (cute-dbt#69 / #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","table":{"columns":["end_after_1900","is_dq_valid","no_future_encounter_dates","start_after_1900","valid_encounter_timestamps"],"rows":[{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}}]},{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}}]}]},"rows":[{"end_after_1900":"true","is_dq_valid":"false","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"false"},{"end_after_1900":"true","is_dq_valid":"true","no_future_encounter_dates":"true","start_after_1900":"true","valid_encounter_timestamps":"true"}],"format":"csv"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","table":{"columns":["end_after_1900_if_present","is_dq_valid","no_future_medication_dates","start_after_1900","valid_medication_dates"],"rows":[{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}}]},{"cells":[{"display":"true","key":{"t":"bool","v":true}},{"display":"false","key":{"t":"bool","v":false}},{"display":"false","key":{"t":"bool","v":false}},{"display":"true","key":{"t":"bool","v":true}},{"display":"true","key":{"t":"bool","v":true}}]}]},"rows":[{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"true","start_after_1900":"true","valid_medication_dates":"false"},{"end_after_1900_if_present":"true","is_dq_valid":"false","no_future_medication_dates":"false","start_after_1900":"true","valid_medication_dates":"true"}],"format":"csv"}],"expected":{"table":{"columns":["entity_type","failed_timestamp_validations","quarantine_rate_pct","quarantined_count","total_count"],"rows":[{"cells":[{"display":"encounters","key":{"t":"str","v":"encounters"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"50.00","key":{"t":"number","v":"50"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"2","key":{"t":"number","v":"2"}}]},{"cells":[{"display":"medications","key":{"t":"str","v":"medications"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"100.00","key":{"t":"number","v":"100"}},{"display":"2","key":{"t":"number","v":"2"}},{"display":"2","key":{"t":"number","v":"2"}}]}]},"rows":[{"entity_type":"encounters","failed_timestamp_validations":"1","quarantine_rate_pct":"50.00","quarantined_count":"1","total_count":"2"},{"entity_type":"medications","failed_timestamp_validations":"1","quarantine_rate_pct":"100.00","quarantined_count":"2","total_count":"2"}],"format":"csv"}},{"id":"unit_test.healthcare_analytics.mart_dq_summary.test_mart_dq_summary_zero_quarantined_when_all_valid","name":"test_mart_dq_summary_zero_quarantined_when_all_valid","target_model":"mart_dq_summary","changed":false,"description":"mart_dq_summary should report zero quarantined_count and 0.00\nquarantine_rate_pct when every upstream row passes data-quality\nvalidations. Uses sql format for the GIVEN mocks and dict format\nfor the EXPECT — dict expects only compare listed columns, so\nnon-deterministic outputs (`_generated_at` from\n`current_timestamp`) are excluded from comparison.\n","tags":[],"meta":{},"defined_in":"models/marts/analytics/_analytics__models.yml","authoring_yaml":"  # LEADING bracket for test 2 — sits between the blank-line\n  # terminator above and `- name:` below with no further gap, so the\n  # leading-wins tiebreaker is moot here (cute-dbt#69 / #73).\n  - name: test_mart_dq_summary_zero_quarantined_when_all_valid\n    description: |\n      mart_dq_summary should report zero quarantined_count and 0.00\n      quarantine_rate_pct when every upstream row passes data-quality\n      validations. Uses sql format for the GIVEN mocks and dict format\n      for the EXPECT — dict expects only compare listed columns, so\n      non-deterministic outputs (`_generated_at` from\n      `current_timestamp`) are excluded from comparison.\n    model: mart_dq_summary\n    # INSIDE bracket for test 2 — same rule as test 1 (cute-dbt#69 /\n    # #73). Demonstrates that the slicer ride-along applies to every\n    # unit_test, not just the first in the file.\n    given:\n      - input: ref('stg_synthea__encounters')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_encounter_timestamps\n            , true as no_future_encounter_dates\n            , true as end_after_1900\n            , true as start_after_1900\n      - input: ref('stg_synthea__medications')\n        format: sql\n        rows: |\n          select\n            true as is_dq_valid\n            , true as valid_medication_dates\n            , true as no_future_medication_dates\n            , true as start_after_1900\n            , true as end_after_1900_if_present\n    expect:\n      format: dict\n      rows:\n        - entity_type: 'encounters'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n        - entity_type: 'medications'\n          quarantined_count: 0\n          total_count: 1\n          quarantine_rate_pct: 0.00\n  # TRAILING bracket for test 2 — terminated by EOF (cute-dbt#69 /\n  # #73).","given":[{"input":"ref('stg_synthea__encounters')","bound_to_node":"encounter_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_encounter_timestamps\n  , true as no_future_encounter_dates\n  , true as end_after_1900\n  , true as start_after_1900","format":"sql"},{"input":"ref('stg_synthea__medications')","bound_to_node":"medication_metrics","rows":"select\n  true as is_dq_valid\n  , true as valid_medication_dates\n  , true as no_future_medication_dates\n  , true as start_after_1900\n  , true as end_after_1900_if_present","format":"sql"}],"expected":{"table":{"columns":["entity_type","quarantine_rate_pct","quarantined_count","total_count"],"rows":[{"cells":[{"display":"encounters","key":{"t":"str","v":"encounters"}},{"display":"0.0","key":{"t":"number","v":"0"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"1","key":{"t":"number","v":"1"}}]},{"cells":[{"display":"medications","key":{"t":"str","v":"medications"}},{"display":"0.0","key":{"t":"number","v":"0"}},{"display":"0","key":{"t":"number","v":"0"}},{"display":"1","key":{"t":"number","v":"1"}}]}]},"rows":[{"entity_type":"encounters","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1},{"entity_type":"medications","quarantine_rate_pct":0.0,"quarantined_count":0,"total_count":1}],"format":"dict"}}],"is_recursive":false}]}</script>
 
 <script>
 /*! jQuery v3.7.1 | (c) OpenJS Foundation and other contributors | jquery.org/license */
@@ -4830,16 +4830,17 @@ <h2>Expected</h2>
       $sec.append(buildExternalFixtureAffordance(given.fixture));
       return $sec;
     }
-    var rows = extractRowsFromFixture(given);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = given.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
     if (given.format && given.format !== "dict") {
       $hdr.append($("<span>").addClass("format-badge").text("format: " + given.format));
     }
     $hdr.append($("<span>").addClass("row-count-badge")
-      .text(rows.length + " row" + (rows.length === 1 ? "" : "s")));
+      .text(rowCount + " row" + (rowCount === 1 ? "" : "s")));
     $sec.append($hdr);
-    var cols = collectColumns(rows);
     var diff = givenDataDiff(dataDiff, given.input);
-    $sec.append(buildFixtureView(cols, rows, "given-table", diff));
+    $sec.append(buildFixtureView(table, "given-table", diff));
     return $sec;
   }
 
@@ -4867,14 +4868,15 @@ <h2>Expected</h2>
       $body.append(buildExternalFixtureAffordance(t.expected.fixture));
       return;
     }
-    var rows = extractRowsFromFixture(t.expected);
-    $panel.find(".expected-rowcount").text(rows.length + " row" + (rows.length === 1 ? "" : "s"));
-    var cols = collectColumns(rows);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = t.expected.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
+    $panel.find(".expected-rowcount").text(rowCount + " row" + (rowCount === 1 ? "" : "s"));
     // The expect side's cell diff, when the test's expect fixture carried a
     // real change (cute-dbt#98). `data_diff.expect` already passed
     // `has_real_change()`, so its presence === "default this table to Diff".
     var diff = (t.data_diff && t.data_diff.expect) || undefined;
-    $body.append(buildFixtureView(cols, rows, "expected-table", diff));
+    $body.append(buildFixtureView(table, "expected-table", diff));
     initDataTablesIn($body);
   }
 
@@ -4898,26 +4900,13 @@ <h2>Expected</h2>
     return Boolean(obj && obj.format === "sql" && typeof obj.rows === "string");
   }
 
-  // Return a tabulable row array from a fixture, handling both
-  // array shapes (dict/csv-from-core) and raw-CSV-string shapes
-  // (csv-from-fusion). Returns `[]` for sql fixtures and other
-  // unexpected shapes — the sql path is dispatched separately by the
-  // caller via `isSqlStringFixture`.
-  function extractRowsFromFixture(fixture) {
-    if (!fixture) return [];
-    var rows = fixture.rows;
-    if (rows == null) return [];
-    if (Array.isArray(rows)) return rows;
-    if (typeof rows === "string" && fixture.format === "csv") {
-      return parseCsvRows(rows);
-    }
-    return [];
-  }
-
   // Render a sql-format fixture as a syntax-highlighted code block.
   // Reuses the compiled-SQL drawer's `.sql-block-wrap` shell so visual
   // style stays consistent. No copy button: the source of truth is the
-  // fixture in the manifest, not the rendered DOM.
+  // fixture in the manifest, not the rendered DOM. This is the ONE
+  // remaining non-tabulatable fallback path: a `format: sql` fixture has no
+  // cells, so the domain returns no `table` POD (cute-dbt#138) and the JS
+  // renders the raw SELECT here.
   function buildSqlCodeBlock(sql) {
     var $wrap = $("<div>").addClass("sql-block-wrap fixture-sql-block-wrap");
     var $pre = $("<pre>").addClass("sql-block").append(
@@ -4927,129 +4916,14 @@ <h2>Expected</h2>
     return $wrap;
   }
 
-  // RFC 4180 CSV parser — state machine over the input string.
-  //
-  // dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as
-  // a raw CSV body in the manifest. dbt-core 1.11+ pre-parses csv to
-  // an array of dicts. cute-dbt does the parse at render time so both
-  // engines render identically (cute-dbt#66).
-  //
-  // Behavior summary:
-  //   * First row is headers; subsequent rows are body values.
-  //   * Quoted fields strip the surrounding double-quotes.
-  //   * Double-quote inside a quoted field is the RFC 4180 escape:
-  //     `""` decodes to a literal `"`.
-  //   * Quoted fields may contain commas and newlines verbatim.
-  //   * A single trailing newline (LF or CRLF) is consumed — dbt YAML
-  //     `rows: |` block scalars always end in `\n`; preserving it
-  //     would create a spurious final empty row.
-  //   * An unterminated final row is accepted (graceful — the
-  //     renderer is not a validator).
-  //   * CRLF (`\r\n`) terminates a row as a single unit.
-  //   * Empty input or header-only input returns `[]`.
-  //
-  // Correctness is unit-tested via `tests/headless_csv_parser.rs` —
-  // the parser is exposed on `window.__cuteParseCsv` so that test can
-  // invoke it through the headless Chrome runtime.
-  function parseCsvRows(text) {
-    if (text == null) return [];
-    var t = String(text);
-    if (t.length === 0) return [];
-    // Strip exactly one trailing newline (LF or CRLF).
-    if (t.charAt(t.length - 1) === "\n") {
-      t = t.slice(0, -1);
-      if (t.length > 0 && t.charAt(t.length - 1) === "\r") {
-        t = t.slice(0, -1);
-      }
-    }
-    if (t.length === 0) return [];
-
-    var matrix = [];
-    var fields = [];
-    var field = "";
-    var inQuotes = false;
-    for (var i = 0; i < t.length; i++) {
-      var c = t.charAt(i);
-      if (inQuotes) {
-        if (c === '"') {
-          // RFC 4180 escape: doubled quote inside a quoted field
-          // collapses to a single literal quote.
-          if (i + 1 < t.length && t.charAt(i + 1) === '"') {
-            field += '"';
-            i++;
-            continue;
-          }
-          inQuotes = false;
-          continue;
-        }
-        field += c;
-        continue;
-      }
-      // Outside quotes.
-      if (c === '"' && field === "") {
-        inQuotes = true;
-        continue;
-      }
-      if (c === ',') {
-        fields.push(field);
-        field = "";
-        continue;
-      }
-      if (c === '\n' || c === '\r') {
-        fields.push(field);
-        matrix.push(fields);
-        fields = [];
-        field = "";
-        if (c === '\r' && i + 1 < t.length && t.charAt(i + 1) === '\n') {
-          i++;
-        }
-        continue;
-      }
-      field += c;
-    }
-    // Flush the final field/row. We always run this — the trailing-
-    // newline strip above guarantees at least one character of
-    // content survives to reach this branch in non-empty input.
-    fields.push(field);
-    matrix.push(fields);
-
-    if (matrix.length < 2) return [];
-    var headers = matrix[0];
-    var out = [];
-    for (var ri = 1; ri < matrix.length; ri++) {
-      var row = matrix[ri];
-      var obj = {};
-      for (var hi = 0; hi < headers.length; hi++) {
-        obj[headers[hi]] = hi < row.length ? row[hi] : "";
-      }
-      out.push(obj);
-    }
-    return out;
-  }
-
-  // Expose the parser on `window.__cuteParseCsv` so the headless
-  // integration test at `tests/headless_csv_parser.rs` can invoke it
-  // via `Runtime.Evaluate` and assert RFC 4180 edge cases (quoted
-  // fields, embedded commas, embedded newlines, escaped quotes,
-  // CRLF, empty input, header-only input). The `__` prefix marks it
-  // as a test-only seam; production code reaches it via the closure
-  // binding above.
-  if (typeof window !== "undefined") {
-    window.__cuteParseCsv = parseCsvRows;
-  }
-
-  function collectColumns(rows) {
-    var seen = {};
-    var order = [];
-    rows.forEach(function (row) {
-      if (row && typeof row === "object" && !Array.isArray(row)) {
-        Object.keys(row).forEach(function (k) {
-          if (!seen[k]) { seen[k] = true; order.push(k); }
-        });
-      }
-    });
-    return order;
-  }
+  // cute-dbt#138 — the JS csv/dict table parsers (`parseCsvRows`,
+  // `extractRowsFromFixture`, `collectColumns`) are RETIRED. The Current view
+  // now renders the Rust-computed `FixtureTable` POD that the domain ships on
+  // each given/expect payload (`table`), so csv/dict tabulation — and the
+  // engine-divergence handling above — happens once, in Rust
+  // (`table_from_manifest_rows`), and the template is a pure renderer. CSV
+  // RFC 4180 correctness is now owned by the Rust `parse_csv_rows` unit tests
+  // (the JS `__cuteParseCsv` headless seam retired with the function).
 
   function escapeHtml(s) {
     return String(s)
@@ -5406,7 +5280,7 @@ <h2>Expected</h2>
   function highlightYaml(yaml) { return emitTokens(tokenizeYaml(yaml)); }
 
   // Expose for headless JS unit tests (cute-dbt#69 / cute-dbt#132 — parallels
-  // the `__cuteParseCsv` seam introduced in cute-dbt#66).
+  // the `__cuteRenderYamlDiff` headless seam).
   window.__cuteHighlightYaml = highlightYaml;
   window.__cuteTokenizeSql = tokenizeSql;
   window.__cuteTokenizeYaml = tokenizeYaml;
@@ -5677,7 +5551,16 @@ <h2>Expected</h2>
     done(ok);
   }
 
-  function buildTable(columns, rows, cls) {
+  // The Current-view grid, rendered PURELY from the Rust-computed
+  // `FixtureTable` POD (cute-dbt#138). Each cell is a `{display, key}` Cell:
+  // the text is the authored `display` (so a csv `1.00` shows `1.00`, not the
+  // canonicalized `1`), while the numeric / NULL styling + sort order come
+  // from the canonical `key` (a `Number` key sorts numerically even when its
+  // display is `1.00`). The JS no longer parses csv/dict — the domain's
+  // `table_from_manifest_rows` owns that, and ships this POD on the wire.
+  function buildTable(table, cls) {
+    var columns = (table && table.columns) || [];
+    var rows = (table && table.rows) || [];
     var $wrap = $("<div>").addClass("table-fit");
     var $tbl = $("<table>").addClass(cls);
     var $thead = $("<thead>").appendTo($tbl);
@@ -5689,23 +5572,34 @@ <h2>Expected</h2>
     var $tbody = $("<tbody>").appendTo($tbl);
     rows.forEach(function (r) {
       var $tr = $("<tr>").appendTo($tbody);
-      columns.forEach(function (col) {
-        var cell = (r && typeof r === "object" && !Array.isArray(r)) ? r[col] : undefined;
-        var $td = $("<td>");
-        if (cell === null || cell === undefined) {
-          $td.addClass("cell-null").text("NULL");
-        } else if (typeof cell === "number") {
-          $td.addClass("cell-num").text(cell);
-        } else {
-          $td.text(String(cell));
-        }
-        $tr.append($td);
+      (r.cells || []).forEach(function (cell) {
+        $tr.append(decorateCurrentCell($("<td>"), cell));
       });
     });
     $wrap.append($tbl);
     return $wrap;
   }
 
+  // Render one Current-view Cell (`{display, key}`) into a <td>: the text is
+  // the authored `display`; the styling/sort affordance is read from `key.t`
+  // (the canonical type), never from the display token (cute-dbt#138). Shares
+  // the null/number styling with `decorateValueCell` so Current and Diff read
+  // identically.
+  function decorateCurrentCell($td, cell) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
+      return $td.addClass("cell-null").text("NULL");
+    }
+    if (key.t === "absent") {
+      return $td.addClass("cell-absent").text("");
+    }
+    if (key.t === "number") {
+      // Sort numerically on the canonical value; show the authored display.
+      return $td.addClass("cell-num").attr("data-order", key.v).text(cell.display);
+    }
+    return $td.text(cell.display);
+  }
+
   // cute-dbt#98 ----------------------------------------------------------
   //
   // The structured cell-diff renderer + the per-table Current↔Diff toggle.
@@ -5737,14 +5631,14 @@ <h2>Expected</h2>
   // is present — also build the Diff grid and a Current↔Diff toggle that flips
   // exactly one view visible at a time (the inactive view carries `hidden`).
   // Default-Diff when a diff exists, else Current with no toggle.
-  function buildFixtureView(columns, rows, cls, diff) {
+  function buildFixtureView(table, cls, diff) {
     if (!diff) {
-      // No cell diff for this table → today's behavior, byte-for-byte.
-      return buildTable(columns, rows, cls);
+      // No cell diff for this table → the plain Current grid from the POD.
+      return buildTable(table, cls);
     }
     var $wrap = $("<div>").addClass("fixture-view");
     var $diffView = buildDiffTable(diff, cls);
-    var $currentView = buildTable(columns, rows, cls).prop("hidden", true);
+    var $currentView = buildTable(table, cls).prop("hidden", true);
 
     var $toggle = $("<div>").addClass("cell-diff-toggle");
     var $diffBtn = $("<button>").attr("type", "button")
@@ -5817,67 +5711,56 @@ <h2>Expected</h2>
       return decorateValueCell($td, cell.old);
     }
     if (rowKind === "added" || rowKind === "unchanged" || !cell.changed) {
+      // Not flagged (added/unchanged, or a format-only change where the keys
+      // are equal): show the NEW authored value, no red/green (cute-dbt#138).
       return decorateValueCell($td, cell.new);
     }
-    // Modified + changed → inline old → new. Each side is type-aware via
-    // diffToken: a real null/absent keeps its own neutral styling and never
-    // borrows the value-change red/green (cute-dbt#132).
-    $td.addClass("cell-changed").attr("data-order", cellValueText(cell.new));
+    // Modified + changed → inline old → new, each side showing its authored
+    // display. Each side is type-aware via diffToken: a real null/absent keeps
+    // its own neutral styling and never borrows the value-change red/green
+    // (cute-dbt#132). The sort key is the NEW cell's authored display.
+    $td.addClass("cell-changed").attr("data-order", cellText(cell.new));
     $td.append(diffToken(cell.old, "cell-old"));
     $td.append($("<span>").addClass("cell-arrow").html("&rarr;"));
     $td.append(diffToken(cell.new, "cell-new"));
     return $td;
   }
 
-  // One side of an inline `old → new` changed cell. Type-aware so a real null
-  // or absent cell reads in its own neutral styling rather than the brilliant
-  // red/green reserved for a real value change: null → italic muted-gray
-  // "NULL"; absent → blank; any concrete value (including the string 'null')
-  // → the side's color via `sideClass` (.cell-old = red, .cell-new = green).
-  // This is what keeps a real NULL visually distinct from a string 'null'
-  // even mid-diff (cute-dbt#132).
-  function diffToken(cv, sideClass) {
-    if (!cv || cv.t === "null") {
+  // One side of an inline `old → new` changed cell. `cell` is a `{display,
+  // key}` Cell (cute-dbt#138): the text is the authored `display`; the
+  // neutral-vs-value styling is read from `key.t` so a real null or absent
+  // cell reads in its own neutral styling rather than the brilliant red/green
+  // reserved for a real value change: null → italic muted-gray "NULL"; absent
+  // → blank; any concrete value (including the string 'null') → the side's
+  // color via `sideClass` (.cell-old = red, .cell-new = green). This keeps a
+  // real NULL visually distinct from a string 'null' even mid-diff (#132).
+  function diffToken(cell, sideClass) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
       return $("<span>").addClass("cell-null").text("NULL");
     }
-    if (cv.t === "absent") {
+    if (key.t === "absent") {
       return $("<span>").addClass("cell-absent").text("");
     }
-    return $("<span>").addClass(sideClass).text(cellValueText(cv));
+    return $("<span>").addClass(sideClass).text(cell.display);
   }
 
-  // Render a single CellValue into a plain value cell, matching `buildTable`'s
-  // null/number styling so the Current and Diff views read identically.
-  function decorateValueCell($td, cv) {
-    if (!cv || cv.t === "null") {
-      // A literal null reads as NULL, matching the Current grid's
-      // `cell === null` styling.
-      return $td.addClass("cell-null").text("NULL");
-    }
-    if (cv.t === "absent") {
-      // An absent cell (a unified-diff column the row's source table lacks)
-      // is NOT a null — render it blank so an Absent->Null change stays
-      // visible and added/removed-column cells don't read as literal nulls.
-      return $td.addClass("cell-absent").text("");
-    }
-    if (cv.t === "number") {
-      return $td.addClass("cell-num").attr("data-order", cv.v).text(cv.v);
-    }
-    return $td.text(cellValueText(cv));
-  }
-
-  // The display text of an adjacently-tagged CellValue ({t, v}). `absent`
-  // renders empty; `null` renders the word NULL; bool/number/str verbatim.
-  function cellValueText(cv) {
-    if (!cv) return "";
-    switch (cv.t) {
-      case "null":   return "NULL";
-      case "absent": return "";
-      case "bool":   return cv.v ? "true" : "false";
-      case "number": return String(cv.v);
-      case "str":    return String(cv.v);
-      default:       return cv.v == null ? "" : String(cv.v);
-    }
+  // Render a single `{display, key}` Cell into a plain value cell, matching
+  // `buildTable`'s null/number styling so the Current and Diff views read
+  // identically. Text is the authored `display`; styling/sort from `key`.
+  function decorateValueCell($td, cell) {
+    return decorateCurrentCell($td, cell);
+  }
+
+  // The display text of a `{display, key}` Cell (cute-dbt#138): a `null` key
+  // renders the word NULL, an `absent` key renders empty, otherwise the
+  // authored `display` token verbatim. Used for the changed-cell sort key.
+  function cellText(cell) {
+    var key = cell && cell.key;
+    if (!key) return "";
+    if (key.t === "null") return "NULL";
+    if (key.t === "absent") return "";
+    return cell.display == null ? "" : String(cell.display);
   }
 
   // Re-measure DataTables column widths for any (now-visible) table under
@@ -5892,13 +5775,15 @@ <h2>Expected</h2>
     });
   }
   // Exposed for headless JS unit tests (cute-dbt#98 — parallels the
-  // `__cuteRenderYamlDiff` / `__cuteParseCsv` seams). Returns the outer
-  // wrapper element so a test can assert the toggle + tinted markup.
+  // `__cuteRenderYamlDiff` seam). Takes the Rust-computed `FixtureTable` POD
+  // (cute-dbt#138) + an optional cell diff and returns the outer wrapper
+  // element so a test can assert the authored-display rendering, the toggle,
+  // and the tinted markup.
   if (typeof window !== "undefined") {
-    window.__cuteBuildFixtureView = function (columns, rows, cls, diff) {
-      return buildFixtureView(columns, rows, cls, diff)[0];
+    window.__cuteBuildFixtureView = function (table, cls, diff) {
+      return buildFixtureView(table, cls, diff)[0];
     };
-    window.__cuteCellValueText = cellValueText;
+    window.__cuteCellText = cellText;
   }
 
   function initDataTablesIn($scope) {
diff --git a/features/unit_test_format_coverage.feature b/features/unit_test_format_coverage.feature
index 2792544f..a234405b 100644
--- a/features/unit_test_format_coverage.feature
+++ b/features/unit_test_format_coverage.feature
@@ -6,9 +6,10 @@
 #
 #   - `dict` rows arrive as an array of dicts (both engines).
 #   - `csv` rows arrive as an array of dicts (dbt-core 1.11+) OR as a
-#     raw CSV string (dbt-fusion 2.0-preview). cute-dbt's JS renderer
-#     contains a hand-rolled RFC 4180 parser for the fusion path
-#     (cute-dbt#66; tested via `tests/headless_csv_parser.rs`).
+#     raw CSV string (dbt-fusion 2.0-preview). cute-dbt parses the fusion
+#     path with a hand-rolled RFC 4180 parser in the domain
+#     (cute-dbt#66; `parse_csv_rows`, unit-tested by `g22`-`g26` in
+#     src/domain/unit_test_table.rs — the JS twin retired in cute-dbt#138).
 #   - `sql` rows arrive as a raw SELECT string (both engines) — cannot
 #     be tabulated without execution; rendered as a syntax-highlighted
 #     code block.
diff --git a/src/adapters/render.rs b/src/adapters/render.rs
index 837714b2..8246520c 100644
--- a/src/adapters/render.rs
+++ b/src/adapters/render.rs
@@ -66,8 +66,9 @@ use crate::adapters::asset_embed::{
 };
 use crate::adapters::cte_engine::{TERMINAL_NODE_NAME, parse_cte_graph};
 use crate::domain::{
-    BANNER_EMPTY_SCOPE, BlockDiff, CteGraph, EdgeType, InScopeSet, Manifest, ModelInScopeSet, Node,
-    NodeId, UnitTest, UnitTestDataDiff, UnitTestGiven, UnitTestYamlBlock, resolve_target_model,
+    BANNER_EMPTY_SCOPE, BlockDiff, CteGraph, EdgeType, FixtureTable, InScopeSet, Manifest,
+    ModelInScopeSet, Node, NodeId, UnitTest, UnitTestDataDiff, UnitTestGiven, UnitTestYamlBlock,
+    resolve_target_model, table_from_manifest_rows,
 };
 
 /// Snake-case wire key for an [`EdgeType`] — the exact JSON-serde string
@@ -274,13 +275,23 @@ pub struct GivenPayload {
     /// copy on the bound node.
     #[serde(skip_serializing_if = "Option::is_none")]
     pub bound_to_node: Option<String>,
+    /// The Rust-computed Current-view [`FixtureTable`] — the authoritative
+    /// tabulated cells (authored `display` + canonical `key` per cell),
+    /// computed in the domain via `table_from_manifest_rows` (cute-dbt#138).
+    /// `None` for a non-tabulatable fixture (sql/opaque, or external-`fixture:`
+    /// rows not in the manifest); the JS then renders the sql code block or the
+    /// external-fixture affordance from [`rows`](Self::rows) /
+    /// [`format`](Self::format) / [`fixture`](Self::fixture). When `Some`, the
+    /// template is a PURE renderer of this POD — it no longer parses
+    /// csv/dict in JS.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub table: Option<FixtureTable>,
     /// Tabular fixture rows lifted verbatim from the unit-test
-    /// `given[i].rows` field (kept as serde `Value` to accept the
-    /// heterogeneous shapes dbt allows). `Value::Null` when this given's
-    /// data lives in an external [`fixture`](Self::fixture) file (the rows
-    /// are not in the manifest) — the render layer then shows the
-    /// external-fixture affordance + YAML fallback rather than an empty
-    /// grid (cute-dbt#98, #126).
+    /// `given[i].rows` field (kept as serde `Value`). Retained for the
+    /// non-tabulatable fallback only: the sql code block, and external-fixture
+    /// detection (`rows == null`). When [`table`](Self::table) is `Some` the
+    /// renderer ignores this field (cute-dbt#138). `Value::Null` when this
+    /// given's data lives in an external [`fixture`](Self::fixture) file.
     pub rows: Value,
     /// Fixture format hint (`csv`, `yaml`, `dict`, etc.) when the
     /// manifest specified one; absent for inline-rows fixtures.
@@ -367,10 +378,19 @@ pub struct TestPayload {
 /// `expect` block lifted into payload shape.
 #[derive(Debug, Clone, Serialize)]
 pub struct ExpectedPayload {
+    /// The Rust-computed Current-view [`FixtureTable`] for the `expect`
+    /// fixture (authored `display` + canonical `key` per cell, cute-dbt#138).
+    /// `None` for a non-tabulatable fixture (sql/opaque or external);
+    /// otherwise the template renders this POD directly. Same contract as
+    /// [`GivenPayload::table`].
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub table: Option<FixtureTable>,
     /// Expected tabular rows lifted verbatim from the unit-test
-    /// `expect.rows` field (serde `Value` accepts the heterogeneous
-    /// shapes dbt allows). `Value::Null` when the data lives in an
-    /// external [`fixture`](Self::fixture) file (cute-dbt#98, #126).
+    /// `expect.rows` field (serde `Value`). Retained for the non-tabulatable
+    /// fallback only (sql code block; external-fixture detection). When
+    /// [`table`](Self::table) is `Some` the renderer ignores this field
+    /// (cute-dbt#138). `Value::Null` when the data lives in an external
+    /// [`fixture`](Self::fixture) file.
     pub rows: Value,
     /// Expected-block format hint (`csv`, `yaml`, `dict`, etc.) when
     /// the manifest specified one; absent for inline-rows fixtures.
@@ -787,6 +807,7 @@ fn build_test_payload(
             GivenPayload {
                 input: g.input().to_owned(),
                 bound_to_node,
+                table: current_view_table(g.rows(), g.format(), g.fixture()),
                 rows: g.rows().clone(),
                 format: g.format().map(str::to_owned),
                 fixture: g.fixture().map(str::to_owned),
@@ -807,6 +828,11 @@ fn build_test_payload(
         data_diff: data_diff.cloned(),
         given,
         expected: ExpectedPayload {
+            table: current_view_table(
+                unit_test.expect().rows(),
+                unit_test.expect().format(),
+                unit_test.expect().fixture(),
+            ),
             rows: unit_test.expect().rows().clone(),
             format: unit_test.expect().format().map(str::to_owned),
             fixture: unit_test.expect().fixture().map(str::to_owned),
@@ -814,6 +840,32 @@ fn build_test_payload(
     }
 }
 
+/// Compute the Current-view [`FixtureTable`] POD for a fixture's `rows` +
+/// `format` (cute-dbt#138): the authoritative tabulated cells the template
+/// renders directly, so the JS never parses csv/dict.
+///
+/// Returns `None` for the two non-tabulatable cases so the JS falls back to
+/// its sql / external-fixture affordances:
+///
+/// 1. **External fixture** — `fixture:` is set AND `rows` is `Value::Null`
+///    (the data is not in the manifest). A `Null` `rows` would otherwise
+///    normalize to the *empty* table, hiding the "data in external file"
+///    affordance behind a silently-empty grid (cute-dbt#126). We gate on the
+///    `fixture:`+`Null` pair (NOT on `rows == Null` alone — a genuinely empty
+///    inline fixture still tabulates to the empty grid).
+/// 2. **sql / opaque** — [`table_from_manifest_rows`] returns `None` (a raw
+///    `SELECT` string has no cells); the JS renders the sql code block.
+fn current_view_table(
+    rows: &Value,
+    format: Option<&str>,
+    fixture: Option<&str>,
+) -> Option<FixtureTable> {
+    if fixture.is_some() && rows.is_null() {
+        return None; // external fixture → JS affordance, not an empty grid
+    }
+    table_from_manifest_rows(rows, format)
+}
+
 /// Locate the leaf CTE node that binds to `ref_name`.
 ///
 /// Two-pass match — both case-insensitive:
@@ -2570,19 +2622,20 @@ mod tests {
     // ===== cute-dbt#98 — data_diff + fixture wire shape =====
 
     use crate::domain::{
-        CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
+        Cell, CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
         RowChange, RowChangeKind, UnitTestDataDiff,
     };
 
     /// A `TestPayload` carrying a `data_diff` serializes the cell-diff with the
-    /// EXACT JS contract the renderer will branch on: each `CellValue` is
-    /// adjacently tagged `{"t": <type>, "v": <value>}`, the `Absent` unit
-    /// variant is `{"t":"absent"}` (no `"v"`), and the row/column enums are
-    /// lowercase tokens. This pins the wire shape independently of the
-    /// domain's own round-trip test (which proves Rust↔Rust; this proves the
-    /// JSON the template consumes).
-    #[test]
-    fn data_diff_payload_wire_shape_has_adjacent_cellvalue_tags() {
+    /// EXACT JS contract the renderer will branch on (cute-dbt#138): each cell
+    /// side is a `Cell` carrying BOTH `display` (the authored token) AND `key`
+    /// (an adjacently-tagged `{"t": <type>, "v": <value>}` `CellValue`). The
+    /// `Absent` key is `{"t":"absent"}` (no `"v"`), and the row/column enums
+    /// are lowercase tokens. This pins the dual-axis wire shape independently
+    /// of the domain's own round-trip test (which proves Rust↔Rust; this proves
+    /// the JSON the template consumes) — cute-dbt#139 reads both axes here.
+    #[test]
+    fn data_diff_payload_wire_ships_both_display_and_key_axes() {
         let data_diff = UnitTestDataDiff {
             given: vec![NamedTableDiff {
                 input: "ref('a')".into(),
@@ -2600,14 +2653,22 @@ mod tests {
                     rows: vec![RowChange {
                         kind: RowChangeKind::Modified,
                         cells: vec![
+                            // A format-only cell: authored "1.00" on the NEW
+                            // side, "1" on the OLD side, but BOTH key to
+                            // Number("1") → changed: false (the headline #138
+                            // case: the diff shows the authored value yet does
+                            // NOT flag).
                             CellChange {
-                                old: CellValue::Number("1".into()),
-                                new: CellValue::Number("1".into()),
+                                old: Cell::with_display("1".into(), CellValue::Number("1".into())),
+                                new: Cell::with_display(
+                                    "1.00".into(),
+                                    CellValue::Number("1".into()),
+                                ),
                                 changed: false,
                             },
                             CellChange {
-                                old: CellValue::Absent,
-                                new: CellValue::Str("NYC".into()),
+                                old: Cell::new(CellValue::Absent),
+                                new: Cell::new(CellValue::Str("NYC".into())),
                                 changed: true,
                             },
                         ],
@@ -2618,25 +2679,37 @@ mod tests {
         };
         let json = serde_json::to_value(&data_diff).expect("data_diff serializes");
         let cells = &json["given"][0]["diff"]["rows"][0]["cells"];
-        // Adjacently-tagged number: {"t":"number","v":"1"}.
-        assert_eq!(cells[0]["old"]["t"], "number");
-        assert_eq!(cells[0]["old"]["v"], "1");
-        // Str cell.
-        assert_eq!(cells[1]["new"]["t"], "str");
-        assert_eq!(cells[1]["new"]["v"], "NYC");
-        // Absent is a bare unit variant — tag only, NO "v" key.
-        assert_eq!(cells[1]["old"]["t"], "absent");
+        // BOTH axes ship: `display` is the authored token; `key` is the
+        // adjacently-tagged CellValue {"t":"number","v":"1"}.
+        assert_eq!(
+            cells[0]["new"]["display"], "1.00",
+            "authored token survives"
+        );
+        assert_eq!(cells[0]["new"]["key"]["t"], "number");
+        assert_eq!(cells[0]["new"]["key"]["v"], "1", "key is canonical");
+        assert_eq!(cells[0]["old"]["display"], "1");
+        assert_eq!(cells[0]["old"]["key"]["v"], "1");
+        // A format-only change is NOT flagged (keys equal) yet shows the
+        // authored NEW display.
+        assert_eq!(cells[0]["changed"], false);
+        // Str cell: display mirrors the key string verbatim.
+        assert_eq!(cells[1]["new"]["display"], "NYC");
+        assert_eq!(cells[1]["new"]["key"]["t"], "str");
+        assert_eq!(cells[1]["new"]["key"]["v"], "NYC");
+        // Absent key is a bare unit variant — tag only, NO "v" key; its
+        // display is the empty string.
+        assert_eq!(cells[1]["old"]["key"]["t"], "absent");
         assert!(
-            cells[1]["old"].get("v").is_none(),
-            "Absent serializes with no \"v\" key; got {}",
-            cells[1]["old"],
+            cells[1]["old"]["key"].get("v").is_none(),
+            "Absent key serializes with no \"v\" key; got {}",
+            cells[1]["old"]["key"],
         );
+        assert_eq!(cells[1]["old"]["display"], "");
         // Row-kind + column-status enums are lowercase tokens.
         assert_eq!(json["given"][0]["diff"]["rows"][0]["kind"], "modified");
         assert_eq!(json["given"][0]["diff"]["columns"][0]["status"], "present");
         assert_eq!(json["given"][0]["diff"]["columns"][1]["status"], "added");
         // The precomputed `changed` verdict rides on each cell.
-        assert_eq!(cells[0]["changed"], false);
         assert_eq!(cells[1]["changed"], true);
     }
 
@@ -2676,8 +2749,8 @@ mod tests {
                 rows: vec![RowChange {
                     kind: RowChangeKind::Added,
                     cells: vec![CellChange {
-                        old: CellValue::Absent,
-                        new: CellValue::Number("9".into()),
+                        old: Cell::new(CellValue::Absent),
+                        new: Cell::new(CellValue::Number("9".into())),
                         changed: true,
                     }],
                 }],
diff --git a/src/domain/cell_diff.rs b/src/domain/cell_diff.rs
index 69c81530..592d5551 100644
--- a/src/domain/cell_diff.rs
+++ b/src/domain/cell_diff.rs
@@ -72,7 +72,7 @@ use crate::domain::pr_diff::{
 use crate::domain::state::InScopeSet;
 use crate::domain::unit_test::UnitTest;
 use crate::domain::unit_test_table::{
-    CellValue, FixtureTable, TableRow, table_from_manifest_rows, table_from_yaml_fragment,
+    Cell, CellValue, FixtureTable, TableRow, table_from_manifest_rows, table_from_yaml_fragment,
 };
 use crate::domain::unit_test_yaml::UnitTestYamlBlock;
 
@@ -194,19 +194,24 @@ pub enum RowChangeKind {
     Removed,
 }
 
-/// One cell's before/after value plus the precomputed semantic verdict.
+/// One cell's before/after [`Cell`]s (each carrying its authored `display` +
+/// canonical `key`) plus the precomputed semantic verdict (cute-dbt#138).
 ///
-/// `changed` is `old != new` — the ONE equality oracle, computed once here
-/// over the canonical [`CellValue`]s so the render layer never re-derives
-/// it (a `Str "1"` vs `Number "1"` never collide; a format-only difference
-/// is `changed: false`).
+/// `changed` is `old.key != new.key` — the ONE equality oracle, computed once
+/// here over the canonical [`CellValue`] **keys** (NEVER the display token) so
+/// a format-only reformat (`1` → `1.00`, both keying to `Number("1")`) is
+/// `changed: false` while a real value change is `changed: true`. The render
+/// layer renders each side's `display` (authored truth) but takes the flag
+/// verdict from here; shipping both axes lets cute-dbt#139 re-flag client-side
+/// without a Rust round-trip.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct CellChange {
-    /// The OLD cell value ([`CellValue::Absent`] for an added row/column).
-    pub old: CellValue,
-    /// The NEW cell value ([`CellValue::Absent`] for a removed row/column).
-    pub new: CellValue,
-    /// `old != new` — precomputed semantic verdict.
+    /// The OLD cell (key [`CellValue::Absent`] for an added row/column).
+    pub old: Cell,
+    /// The NEW cell (key [`CellValue::Absent`] for a removed row/column).
+    pub new: Cell,
+    /// `old.key != new.key` — precomputed semantic verdict on the equality
+    /// axis only (the display token never participates).
     pub changed: bool,
 }
 
@@ -240,8 +245,8 @@ pub fn diff_fixture_tables(old: &FixtureTable, new: &FixtureTable) -> FixtureTab
     let columns = unify_columns(old, new);
     let col_names: Vec<&str> = columns.iter().map(|c| c.name.as_str()).collect();
 
-    let old_rows: Vec<Vec<CellValue>> = project_rows(old, &col_names);
-    let new_rows: Vec<Vec<CellValue>> = project_rows(new, &col_names);
+    let old_rows: Vec<Vec<Cell>> = project_rows(old, &col_names);
+    let new_rows: Vec<Vec<Cell>> = project_rows(new, &col_names);
 
     let rows = align_rows(&old_rows, &new_rows);
 
@@ -251,7 +256,7 @@ pub fn diff_fixture_tables(old: &FixtureTable, new: &FixtureTable) -> FixtureTab
 /// Align the projected OLD/NEW rows: multiset-match equal keys as
 /// `Unchanged`, then pair the residual positionally, emitting in NEW order
 /// with trailing `Removed` appended.
-fn align_rows(old_rows: &[Vec<CellValue>], new_rows: &[Vec<CellValue>]) -> Vec<RowChange> {
+fn align_rows(old_rows: &[Vec<Cell>], new_rows: &[Vec<Cell>]) -> Vec<RowChange> {
     let (new_matched, old_consumed) = match_multiset(old_rows, new_rows);
 
     // Residual rows, still in source order.
@@ -291,10 +296,7 @@ fn align_rows(old_rows: &[Vec<CellValue>], new_rows: &[Vec<CellValue>]) -> Vec<R
 /// the same key, which consumes that OLD row. Returns `(new_matched,
 /// old_consumed)`. Duplicate keys match duplicate-for-duplicate (the
 /// LCS-on-repeats property without LCS).
-fn match_multiset(
-    old_rows: &[Vec<CellValue>],
-    new_rows: &[Vec<CellValue>],
-) -> (Vec<bool>, Vec<bool>) {
+fn match_multiset(old_rows: &[Vec<Cell>], new_rows: &[Vec<Cell>]) -> (Vec<bool>, Vec<bool>) {
     let old_keys: Vec<String> = old_rows.iter().map(|r| row_key(r)).collect();
     let mut consumed = vec![false; old_rows.len()];
     let mut matched = vec![false; new_rows.len()];
@@ -325,8 +327,8 @@ fn first_unconsumed(old_keys: &[String], consumed: &[bool], nkey: &str) -> Optio
 fn pair_residual(
     residual_old: &[usize],
     residual_new: &[usize],
-    old_rows: &[Vec<CellValue>],
-    new_rows: &[Vec<CellValue>],
+    old_rows: &[Vec<Cell>],
+    new_rows: &[Vec<Cell>],
     by_new: &mut HashMap<usize, RowChange>,
     trailing_removed: &mut Vec<RowChange>,
 ) {
@@ -376,8 +378,9 @@ fn unify_columns(old: &FixtureTable, new: &FixtureTable) -> Vec<DiffColumn> {
 }
 
 /// Project each of `table`'s rows over the unified `col_names`, filling a
-/// column the row lacks with [`CellValue::Absent`].
-fn project_rows(table: &FixtureTable, col_names: &[&str]) -> Vec<Vec<CellValue>> {
+/// column the row lacks with an absent [`Cell`]. Each projected cell keeps its
+/// authored display alongside the canonical key (cute-dbt#138).
+fn project_rows(table: &FixtureTable, col_names: &[&str]) -> Vec<Vec<Cell>> {
     table
         .rows
         .iter()
@@ -390,15 +393,16 @@ fn project_rows(table: &FixtureTable, col_names: &[&str]) -> Vec<Vec<CellValue>>
         .collect()
 }
 
-/// The [`CellValue`] of `row` for unified column `col`, or
-/// [`CellValue::Absent`] when this table has no such column.
-fn cell_for_column(table: &FixtureTable, row: &TableRow, col: &str) -> CellValue {
+/// The [`Cell`] of `row` for unified column `col`, or an absent cell (key
+/// [`CellValue::Absent`], empty display) when this table has no such column.
+fn cell_for_column(table: &FixtureTable, row: &TableRow, col: &str) -> Cell {
     table
         .columns
         .iter()
         .position(|c| c == col)
         .and_then(|idx| row.cells.get(idx))
-        .map_or(CellValue::Absent, |cell| cell.value.clone())
+        .cloned()
+        .unwrap_or_else(|| Cell::new(CellValue::Absent))
 }
 
 /// The canonical row key — each cell's tag+value, length-prefixed and
@@ -414,10 +418,13 @@ fn cell_for_column(table: &FixtureTable, row: &TableRow, col: &str) -> CellValue
 /// `["a|str:x", "b"]` both flatten to `str:a|str:x|str:b` — which would let
 /// [`match_multiset`] mark unrelated rows `Unchanged` and silently drop a
 /// real cell diff.
-fn row_key(cells: &[CellValue]) -> String {
+fn row_key(cells: &[Cell]) -> String {
     let mut key = String::new();
     for cell in cells {
-        let token = cell_key_token(cell);
+        // Keyed on the canonical equality axis ONLY — the authored `display`
+        // never participates, so the display lens (cute-dbt#139) cannot
+        // re-pair rows (cute-dbt#138).
+        let token = cell_key_token(&cell.key);
         // `<byte-len>:<token>` — the length prefix makes concatenation
         // unambiguous regardless of what bytes the token carries.
         key.push_str(&token.len().to_string());
@@ -438,73 +445,85 @@ fn cell_key_token(v: &CellValue) -> String {
     }
 }
 
-/// An `Unchanged` row: `old`/`new` echo the projected cells, every
-/// `changed` is `false`.
-fn unchanged_row(cells: &[CellValue]) -> RowChange {
+/// An absent [`Cell`] (key [`CellValue::Absent`], empty display) — the
+/// add/remove placeholder for the side a row/column is missing from.
+fn absent_cell() -> Cell {
+    Cell::new(CellValue::Absent)
+}
+
+/// An `Unchanged` row: `old`/`new` echo the projected cells (display + key),
+/// every `changed` is `false`.
+fn unchanged_row(cells: &[Cell]) -> RowChange {
     RowChange {
         kind: RowChangeKind::Unchanged,
         cells: cells
             .iter()
-            .map(|v| CellChange {
-                old: v.clone(),
-                new: v.clone(),
+            .map(|c| CellChange {
+                old: c.clone(),
+                new: c.clone(),
                 changed: false,
             })
             .collect(),
     }
 }
 
-/// An `Added` row: `old` is `Absent`, `new` is the projected cell.
-fn added_row(cells: &[CellValue]) -> RowChange {
+/// An `Added` row: `old` is absent, `new` is the projected cell. Honors the
+/// `changed == old.key != new.key` contract — an `Absent` projected cell (a
+/// column this row lacks) stays NOT changed (cute-dbt#138).
+fn added_row(cells: &[Cell]) -> RowChange {
     RowChange {
         kind: RowChangeKind::Added,
         cells: cells
             .iter()
-            .map(|v| CellChange {
-                old: CellValue::Absent,
-                new: v.clone(),
-                changed: true,
+            .map(|c| CellChange {
+                old: absent_cell(),
+                new: c.clone(),
+                changed: c.key != CellValue::Absent,
             })
             .collect(),
     }
 }
 
-/// A `Removed` row: `new` is `Absent`, `old` is the projected cell.
-fn removed_row(cells: &[CellValue]) -> RowChange {
+/// A `Removed` row: `new` is absent, `old` is the projected cell. Honors the
+/// `changed == old.key != new.key` contract — an `Absent` projected cell (a
+/// column this row lacks) stays NOT changed (cute-dbt#138).
+fn removed_row(cells: &[Cell]) -> RowChange {
     RowChange {
         kind: RowChangeKind::Removed,
         cells: cells
             .iter()
-            .map(|v| CellChange {
-                old: v.clone(),
-                new: CellValue::Absent,
-                changed: true,
+            .map(|c| CellChange {
+                old: c.clone(),
+                new: absent_cell(),
+                changed: c.key != CellValue::Absent,
             })
             .collect(),
     }
 }
 
 /// Whether two projected rows share at least one column position where
-/// BOTH cells are present (non-`Absent`) — the heuristic that distinguishes
-/// an in-place edit (`Modified`) from a wholesale replacement (a separate
-/// `Removed` + `Added`).
-fn rows_share_present_cell(old_cells: &[CellValue], new_cells: &[CellValue]) -> bool {
+/// BOTH cells are present (non-`Absent` key) — the heuristic that
+/// distinguishes an in-place edit (`Modified`) from a wholesale replacement
+/// (a separate `Removed` + `Added`).
+fn rows_share_present_cell(old_cells: &[Cell], new_cells: &[Cell]) -> bool {
     old_cells
         .iter()
         .zip(new_cells.iter())
-        .any(|(o, n)| *o != CellValue::Absent && *n != CellValue::Absent)
+        .any(|(o, n)| o.key != CellValue::Absent && n.key != CellValue::Absent)
 }
 
 /// Fuse a residual OLD row's cells with a residual NEW row's cells into one
-/// `Modified` row, computing each cell's `changed` verdict (`old != new`).
-fn modified_row(old_cells: &[CellValue], new_cells: &[CellValue]) -> RowChange {
+/// `Modified` row, computing each cell's `changed` verdict on the **key**
+/// axis only (`old.key != new.key`) — a format-only display change is not a
+/// change (cute-dbt#138).
+fn modified_row(old_cells: &[Cell], new_cells: &[Cell]) -> RowChange {
     let cells = old_cells
         .iter()
         .zip(new_cells.iter())
         .map(|(o, n)| CellChange {
             old: o.clone(),
             new: n.clone(),
-            changed: o != n,
+            changed: o.key != n.key,
         })
         .collect();
     RowChange {
@@ -828,8 +847,8 @@ mod tests {
         // id unchanged, qty changed.
         assert!(!diff.rows[0].cells[0].changed, "id cell unchanged");
         assert!(diff.rows[0].cells[1].changed, "qty cell changed");
-        assert_eq!(diff.rows[0].cells[1].old, n("100"));
-        assert_eq!(diff.rows[0].cells[1].new, n("200"));
+        assert_eq!(diff.rows[0].cells[1].old.key, n("100"));
+        assert_eq!(diff.rows[0].cells[1].new.key, n("200"));
     }
 
     #[test]
@@ -838,8 +857,16 @@ mod tests {
         let old = table(&["v"], vec![vec![n("10")]]);
         let new = table(&["v"], vec![vec![n("20")]]);
         let diff = diff_fixture_tables(&old, &new);
-        assert_eq!(diff.rows[0].cells[0].old, n("10"), "old side must be 10");
-        assert_eq!(diff.rows[0].cells[0].new, n("20"), "new side must be 20");
+        assert_eq!(
+            diff.rows[0].cells[0].old.key,
+            n("10"),
+            "old side must be 10"
+        );
+        assert_eq!(
+            diff.rows[0].cells[0].new.key,
+            n("20"),
+            "new side must be 20"
+        );
     }
 
     // ----- D. row add / remove / pairing boundary -----
@@ -852,8 +879,8 @@ mod tests {
         assert_eq!(diff.rows.len(), 2);
         assert_eq!(diff.rows[0].kind, RowChangeKind::Unchanged);
         assert_eq!(diff.rows[1].kind, RowChangeKind::Added);
-        assert_eq!(diff.rows[1].cells[0].old, CellValue::Absent);
-        assert_eq!(diff.rows[1].cells[0].new, n("2"));
+        assert_eq!(diff.rows[1].cells[0].old.key, CellValue::Absent);
+        assert_eq!(diff.rows[1].cells[0].new.key, n("2"));
     }
 
     #[test]
@@ -864,8 +891,8 @@ mod tests {
         assert_eq!(diff.rows.len(), 2);
         assert_eq!(diff.rows[0].kind, RowChangeKind::Unchanged);
         assert_eq!(diff.rows[1].kind, RowChangeKind::Removed);
-        assert_eq!(diff.rows[1].cells[0].old, n("2"));
-        assert_eq!(diff.rows[1].cells[0].new, CellValue::Absent);
+        assert_eq!(diff.rows[1].cells[0].old.key, n("2"));
+        assert_eq!(diff.rows[1].cells[0].new.key, CellValue::Absent);
     }
 
     #[test]
@@ -887,10 +914,10 @@ mod tests {
         // The one row is Modified: c1 a -> a|str:x, c2 x|str:b -> b.
         assert_eq!(diff.rows.len(), 1);
         assert_eq!(diff.rows[0].kind, RowChangeKind::Modified);
-        assert_eq!(diff.rows[0].cells[0].old, s("a"));
-        assert_eq!(diff.rows[0].cells[0].new, s("a|str:x"));
-        assert_eq!(diff.rows[0].cells[1].old, s("x|str:b"));
-        assert_eq!(diff.rows[0].cells[1].new, s("b"));
+        assert_eq!(diff.rows[0].cells[0].old.key, s("a"));
+        assert_eq!(diff.rows[0].cells[0].new.key, s("a|str:x"));
+        assert_eq!(diff.rows[0].cells[1].old.key, s("x|str:b"));
+        assert_eq!(diff.rows[0].cells[1].new.key, s("b"));
     }
 
     #[test]
@@ -953,12 +980,12 @@ mod tests {
         );
         assert!(diff.rows.iter().all(|r| r.kind == RowChangeKind::Modified));
         // Row 0 pairs id=1: old v=10, new v=11. Row 1 pairs id=2: old 20, new 21.
-        assert_eq!(diff.rows[0].cells[0].new, n("1"));
-        assert_eq!(diff.rows[0].cells[1].old, n("10"));
-        assert_eq!(diff.rows[0].cells[1].new, n("11"));
-        assert_eq!(diff.rows[1].cells[0].new, n("2"));
-        assert_eq!(diff.rows[1].cells[1].old, n("20"));
-        assert_eq!(diff.rows[1].cells[1].new, n("21"));
+        assert_eq!(diff.rows[0].cells[0].new.key, n("1"));
+        assert_eq!(diff.rows[0].cells[1].old.key, n("10"));
+        assert_eq!(diff.rows[0].cells[1].new.key, n("11"));
+        assert_eq!(diff.rows[1].cells[0].new.key, n("2"));
+        assert_eq!(diff.rows[1].cells[1].old.key, n("20"));
+        assert_eq!(diff.rows[1].cells[1].new.key, n("21"));
     }
 
     // ----- E. row reorder (the riskiest seam) -----
@@ -1006,8 +1033,8 @@ mod tests {
             .collect();
         // Only the id=1 row is a genuine edit; the moved rows are Unchanged.
         assert_eq!(modified.len(), 1, "exactly one row genuinely edited");
-        assert_eq!(modified[0].cells[1].old, n("10"));
-        assert_eq!(modified[0].cells[1].new, n("99"));
+        assert_eq!(modified[0].cells[1].old.key, n("10"));
+        assert_eq!(modified[0].cells[1].new.key, n("99"));
     }
 
     #[test]
@@ -1098,7 +1125,7 @@ mod tests {
         assert!(diff.has_real_change(), "a column add is a real change");
         // Each row's new city cell is present, old is Absent, changed.
         for r in &diff.rows {
-            assert_eq!(r.cells[1].old, CellValue::Absent);
+            assert_eq!(r.cells[1].old.key, CellValue::Absent);
             assert!(r.cells[1].changed);
         }
     }
@@ -1121,7 +1148,79 @@ mod tests {
             .iter()
             .position(|c| c.name == "legacy")
             .unwrap();
-        assert_eq!(diff.rows[0].cells[legacy_idx].new, CellValue::Absent);
+        assert_eq!(diff.rows[0].cells[legacy_idx].new.key, CellValue::Absent);
+    }
+
+    #[test]
+    fn f_added_removed_rows_keep_absent_cells_unchanged() {
+        // cute-dbt#138 (CodeRabbit #140): when a row add/remove coincides with
+        // an unrelated column add/remove, the unified axis projects an
+        // `Absent` cell into the added/removed row. Per the
+        // `changed == old.key != new.key` contract that `modified_row` honors,
+        // an `Absent -> Absent` cell is NOT a change — the builders must not
+        // hardcode `changed: true`. Pins both `&&`/comparison mutants.
+        let absent = Cell::new(CellValue::Absent);
+        let added = added_row(&[absent.clone(), Cell::new(n("2"))]);
+        assert!(
+            !added.cells[0].changed,
+            "Absent projected cell in an Added row is not a change",
+        );
+        assert!(
+            added.cells[1].changed,
+            "a present projected cell in an Added row is a change",
+        );
+        let removed = removed_row(&[Cell::new(n("2")), absent.clone()]);
+        assert!(
+            removed.cells[0].changed,
+            "a present projected cell in a Removed row is a change",
+        );
+        assert!(
+            !removed.cells[1].changed,
+            "Absent projected cell in a Removed row is not a change",
+        );
+    }
+
+    #[test]
+    fn g_added_row_absent_overlap_cell_is_unchanged_end_to_end() {
+        // cute-dbt#138 (CI stage review): the integration counterpart of
+        // `f_added_removed_rows_keep_absent_cells_unchanged`. A row add that
+        // COINCIDES with a column add makes the unified column axis project an
+        // `Absent` cell into the Added row. Building the tables through the
+        // production normalizer (not the builders directly) proves the
+        // end-to-end path leaves that `Absent -> Absent` cell unflagged.
+        use crate::domain::unit_test_table::table_from_manifest_rows;
+        let old = table_from_manifest_rows(&serde_json::json!([{"a": 1}]), Some("dict")).unwrap();
+        let new = table_from_manifest_rows(&serde_json::json!([{"a": 1}, {"b": 2}]), Some("dict"))
+            .unwrap();
+        let diff = diff_fixture_tables(&old, &new);
+        let a = diff
+            .columns
+            .iter()
+            .position(|c| c.name == "a")
+            .expect("col a");
+        let b = diff
+            .columns
+            .iter()
+            .position(|c| c.name == "b")
+            .expect("col b");
+        let added = diff
+            .rows
+            .iter()
+            .find(|r| r.kind == RowChangeKind::Added)
+            .expect("{b:2} is wholly added — shares no present cell with {a:1}");
+        assert_eq!(
+            added.cells[a].new.key,
+            CellValue::Absent,
+            "the added row lacks column a, so its a-cell is Absent",
+        );
+        assert!(
+            !added.cells[a].changed,
+            "Absent->Absent in the Added row stays unchanged (row-add overlaps column-add)",
+        );
+        assert!(
+            added.cells[b].changed,
+            "the present b-cell in the Added row is a change",
+        );
     }
 
     // ----- Cross-source equivalence (the headline kill) -----
@@ -1146,8 +1245,8 @@ mod tests {
     #[test]
     fn cross_source_empty_csv_cell_converges_no_change() {
         // cute-dbt#127 (closes #124): the core csv-as-array-of-string-dicts
-        // NEW path now routes string cells through `type_csv_value` →
-        // `type_csv_token`, so an empty core cell maps to `Null` — the SAME as
+        // NEW path now routes string cells through `cell_from_csv_value` →
+        // `cell_from_csv_token`, so an empty core cell keys to `Null` — the SAME as
         // the OLD-YAML csv path's empty → `Null`. Both sides converge: `id`
         // "1" → Number on both, `note` "" → Null on both. The previously-
         // documented Str("") vs Null divergence is GONE; this now asserts a
@@ -1180,8 +1279,8 @@ mod tests {
         let new =
             table_from_manifest_rows(&serde_json::json!([{"id": "1.5"}]), Some("csv")).unwrap();
         // Both inferred to Number (the inference fired on each side)…
-        assert_eq!(old.rows[0].cells[0].value, CellValue::Number("1".into()));
-        assert_eq!(new.rows[0].cells[0].value, CellValue::Number("1.5".into()));
+        assert_eq!(old.rows[0].cells[0].key, CellValue::Number("1".into()));
+        assert_eq!(new.rows[0].cells[0].key, CellValue::Number("1.5".into()));
         // …and the genuine value change survives to the diff verdict.
         let diff = diff_fixture_tables(&old, &new);
         assert!(
@@ -1200,11 +1299,11 @@ mod tests {
         assert_eq!(fmt_a.as_deref(), Some("dict"));
         let tbl_a = table_from_yaml_fragment(&region_a, fmt_a.as_deref()).unwrap();
         assert_eq!(tbl_a.rows.len(), 1);
-        assert_eq!(tbl_a.rows[0].cells[0].value, n("1"));
+        assert_eq!(tbl_a.rows[0].cells[0].key, n("1"));
 
         let (region_b, _) = slice_named_region(old_text, "input", Some("ref('b')")).unwrap();
         let tbl_b = table_from_yaml_fragment(&region_b, Some("dict")).unwrap();
-        assert_eq!(tbl_b.rows[0].cells[0].value, n("2"));
+        assert_eq!(tbl_b.rows[0].cells[0].key, n("2"));
     }
 
     #[test]
@@ -1213,7 +1312,7 @@ mod tests {
         let (region, _) = slice_named_region(old_text, "expect", None).unwrap();
         let tbl = table_from_yaml_fragment(&region, Some("dict")).unwrap();
         assert_eq!(tbl.rows.len(), 1);
-        assert_eq!(tbl.rows[0].cells[0].value, n("99"));
+        assert_eq!(tbl.rows[0].cells[0].key, n("99"));
     }
 
     #[test]
@@ -1280,7 +1379,7 @@ mod tests {
         let (region, _) = slice_named_region(old_text, "input", Some("ref('q')")).unwrap();
         let tbl = table_from_yaml_fragment(&region, Some("dict")).unwrap();
         assert_eq!(tbl.rows.len(), 1);
-        assert_eq!(tbl.rows[0].cells[0].value, n("7"));
+        assert_eq!(tbl.rows[0].cells[0].key, n("7"));
     }
 
     // ----- table_diff_if_changed gating -----
@@ -1373,8 +1472,8 @@ mod tests {
                     rows: vec![RowChange {
                         kind: RowChangeKind::Modified,
                         cells: vec![CellChange {
-                            old: n("1"),
-                            new: n("2"),
+                            old: Cell::new(n("1")),
+                            new: Cell::new(n("2")),
                             changed: true,
                         }],
                     }],
@@ -1388,8 +1487,8 @@ mod tests {
                 rows: vec![RowChange {
                     kind: RowChangeKind::Added,
                     cells: vec![CellChange {
-                        old: CellValue::Absent,
-                        new: n("9"),
+                        old: Cell::new(CellValue::Absent),
+                        new: Cell::new(n("9")),
                         changed: true,
                     }],
                 }],
@@ -1449,8 +1548,8 @@ mod tests {
                                 rows: vec![RowChange {
                                     kind,
                                     cells: vec![CellChange {
-                                        old: old.clone(),
-                                        new: new.clone(),
+                                        old: Cell::new(old.clone()),
+                                        new: Cell::new(new.clone()),
                                         changed: old != new,
                                     }],
                                 }],
diff --git a/src/domain/unit_test_table.rs b/src/domain/unit_test_table.rs
index c01fde3a..72cb83ae 100644
--- a/src/domain/unit_test_table.rs
+++ b/src/domain/unit_test_table.rs
@@ -121,21 +121,66 @@ impl TableRow {
     }
 }
 
-/// One cell — a thin newtype over its [typed value](CellValue).
+/// One cell — split into its two axes (cute-dbt#138):
 ///
-/// Kept a struct (not a bare `CellValue`) so the File-2 diff can later hang
-/// per-cell render hints off it without a wire-shape break.
+/// - [`display`](Self::display) — the **authored token** (truth), rendered in
+///   BOTH the Current and Diff views. A csv `1.00` shows `1.00`, not the
+///   normalized `1`.
+/// - [`key`](Self::key) — the canonical [`CellValue`], the **equality** axis,
+///   used ONLY for the diff's change decision and row alignment. `1`, `1.0`,
+///   `1.00` all key to `Number("1")`, so a format-only reformat is not a diff.
+///
+/// Shipping both axes to JS is the foundation the settings normalize-toggle
+/// (cute-dbt#139) builds on — it re-flags client-side between `key` (normalized)
+/// and `display` (strict) without a Rust round-trip.
+///
+/// Kept a struct (not a bare `CellValue`) so the diff can hang per-cell render
+/// hints off it without a wire-shape break.
 #[derive(Debug, Clone, PartialEq, Eq, Serialize, Deserialize)]
 pub struct Cell {
-    /// The cell's semantically-typed value.
-    pub value: CellValue,
+    /// The authored token, rendered verbatim in both views. For an
+    /// [`Absent`](CellValue::Absent) or [`Null`](CellValue::Null) key the
+    /// display is the empty string (the renderer derives the NULL/absent
+    /// affordance from `key.t`, never from `display`).
+    pub display: String,
+    /// The cell's semantically-typed equality key.
+    pub key: CellValue,
 }
 
 impl Cell {
-    /// Construct from a typed value.
+    /// Construct from a typed value, deriving the display from the key (the
+    /// canonical form). Use this when there is no distinct authored token —
+    /// e.g. test literals, projected diff cells, and `Absent` placeholders.
     #[must_use]
-    pub fn new(value: CellValue) -> Self {
-        Self { value }
+    pub fn new(key: CellValue) -> Self {
+        Self {
+            display: display_from_key(&key),
+            key,
+        }
+    }
+
+    /// Construct from a distinct authored `display` token plus its canonical
+    /// `key`. This is the **fidelity** path: the normalizers capture the raw
+    /// token here so the Current/Diff views render what the author wrote, not
+    /// the canonicalized form.
+    #[must_use]
+    pub fn with_display(display: String, key: CellValue) -> Self {
+        Self { display, key }
+    }
+}
+
+/// The display string derived from a canonical [`CellValue`] when there is no
+/// distinct authored token: the value rendered as the author would have seen
+/// it. [`Null`](CellValue::Null) and [`Absent`](CellValue::Absent) yield the
+/// empty string — the renderer supplies their NULL/blank affordance from
+/// `key.t`, never from the display text.
+#[must_use]
+pub fn display_from_key(key: &CellValue) -> String {
+    match key {
+        CellValue::Null | CellValue::Absent => String::new(),
+        CellValue::Bool(b) => b.to_string(),
+        CellValue::Number(n) => n.clone(),
+        CellValue::Str(s) => s.clone(),
     }
 }
 
@@ -322,19 +367,81 @@ pub fn type_csv_token(token: &str) -> CellValue {
     CellValue::Str(token.to_owned())
 }
 
-/// Type a NEW-side cell from a csv-format manifest `Value` (dbt-core's
-/// csv-as-array-of-string-dicts encoding). A [`Value::String`] cell — what
-/// core ships for every csv field — routes through [`type_csv_token`] so its
-/// value is inferred (the csv-Array format discriminator, cute-dbt#127 DELTA
-/// 2); any other `Value` shape (a defensive non-string cell) falls back to
-/// [`type_cell_value`]. This is the `type_fn` [`table_from_value_objects`]
-/// threads when `format: csv`, mirroring how `format: dict` threads
-/// [`type_cell_value`] verbatim — `format` is the only discriminator, so a
-/// deliberately-quoted dict `'1'` stays `Str` while a csv `1` infers `Number`.
-fn type_csv_value(v: &Value) -> CellValue {
+// ---------------------------------------------------------------------
+// Authored-token cell builders — display (truth) + key (equality)
+// ---------------------------------------------------------------------
+
+/// Build a NEW-side **dict** [`Cell`] from an already-typed JSON `Value`,
+/// capturing the authored token as the display (cute-dbt#138).
+///
+/// `key` is [`type_cell_value`]'s canonical value; `display` is the value as
+/// authored: a string verbatim, a bool's `true`/`false`, a number's authored
+/// digits (`Value::Number`'s own `to_string`, which — absent `serde_json`'s
+/// `arbitrary_precision` — is the f64 round-trip, so a dict-numeric `1.00`
+/// already arrived as `1.0` upstream of this layer), or compact JSON for a
+/// nested value. A `Null` displays empty (the renderer styles it from `key.t`).
+#[must_use]
+pub fn cell_from_value(v: &Value) -> Cell {
+    let key = type_cell_value(v);
+    let display = match v {
+        Value::Null => String::new(),
+        Value::String(s) => s.clone(),
+        Value::Bool(b) => b.to_string(),
+        Value::Number(n) => n.to_string(),
+        Value::Array(_) | Value::Object(_) => serde_json::to_string(v).unwrap_or_default(),
+    };
+    Cell::with_display(display, key)
+}
+
+/// Build an OLD-side **dict** [`Cell`] from a YAML scalar token, capturing the
+/// authored token as the display (cute-dbt#138).
+///
+/// `key` is [`type_cell_scalar`]'s canonical value. `display` is the authored
+/// token as a reader sees it: a quoted scalar's inner text (quote-stripped via
+/// the same `parse_yaml_scalar` path that `type_cell_scalar` uses), else the
+/// trimmed token. A token whose key is [`Null`](CellValue::Null) (`""`,
+/// `null`, `~`) displays empty so the renderer styles it as NULL from `key.t`.
+#[must_use]
+pub fn cell_from_scalar(token: &str) -> Cell {
+    let key = type_cell_scalar(token);
+    let trimmed = token.trim();
+    let display = if is_quoted(trimmed) {
+        parse_yaml_scalar(trimmed)
+    } else if matches!(key, CellValue::Null) {
+        String::new()
+    } else {
+        trimmed.to_owned()
+    };
+    Cell::with_display(display, key)
+}
+
+/// Build a csv [`Cell`] from a field token, capturing the raw token as the
+/// display (cute-dbt#138).
+///
+/// `key` is [`type_csv_token`]'s value-inferred canonical value; `display` is
+/// the raw csv token **verbatim** (so a csv `1.00` renders as `1.00` even
+/// though its key is `Number("1")` — the headline fidelity fix). An empty
+/// field (`key == Null`) displays empty.
+#[must_use]
+pub fn cell_from_csv_token(token: &str) -> Cell {
+    let key = type_csv_token(token);
+    let display = if token.is_empty() {
+        String::new()
+    } else {
+        token.to_owned()
+    };
+    Cell::with_display(display, key)
+}
+
+/// The csv-format NEW-side `Value` cell builder (cute-dbt#138): a
+/// [`Value::String`] routes through [`cell_from_csv_token`] (raw token kept as
+/// display, value-inferred key); any other shape falls back to
+/// [`cell_from_value`]. The `cell_fn` analogue of the dict-path
+/// [`cell_from_value`] — `format` is the only discriminator.
+fn cell_from_csv_value(v: &Value) -> Cell {
     match v {
-        Value::String(s) => type_csv_token(s),
-        other => type_cell_value(other),
+        Value::String(s) => cell_from_csv_token(s),
+        other => cell_from_value(other),
     }
 }
 
@@ -424,13 +531,15 @@ fn canonicalize_float(f: f64) -> String {
 /// ([`table_from_manifest_rows`] on a `Value::String`) and the csv OLD side
 /// ([`table_from_yaml_fragment`] on a dedented `rows: |` body).
 ///
-/// Behavior, mirroring the JS twin's `tests/headless_csv_parser.rs` cases:
-/// strip exactly one trailing `\n` (and a preceding `\r`); quoted fields;
-/// `""` → a literal `"`; CRLF as one terminator; commas/newlines inside
-/// quotes verbatim; the first row is the header; fewer than two rows
-/// (empty or header-only) → `[]`; an unterminated final row is accepted;
-/// a row shorter than the header fills the missing trailing columns with
-/// `""` (which [`type_csv_token`] then maps to `Null`).
+/// Behavior (since cute-dbt#138 this is the SOLE RFC 4180 implementation — the
+/// JS `parseCsvRows` twin was retired when the Current view started rendering
+/// the Rust-computed [`FixtureTable`] POD; the `g22`–`g26` unit tests own its
+/// correctness): strip exactly one trailing `\n` (and a preceding `\r`);
+/// quoted fields; `""` → a literal `"`; CRLF as one terminator;
+/// commas/newlines inside quotes verbatim; the first row is the header; fewer
+/// than two rows (empty or header-only) → `[]`; an unterminated final row is
+/// accepted; a row shorter than the header fills the missing trailing columns
+/// with `""` (which [`type_csv_token`] then maps to `Null`).
 #[must_use]
 pub fn parse_csv_rows(text: &str) -> Vec<Vec<(String, String)>> {
     if text.is_empty() {
@@ -787,17 +896,18 @@ pub fn table_from_manifest_rows(rows: &Value, format: Option<&str>) -> Option<Fi
     let fmt = FixtureFormat::from_opt(format);
     match rows {
         Value::Null => Some(FixtureTable::default()),
-        // The Array arm is FORMAT-AWARE (cute-dbt#127 DELTA 2): dbt-core
-        // encodes csv as a `Value::Array` of all-string dicts, so a csv-format
-        // array threads the value-inferring `type_csv_value`; a dict-format
-        // array keeps `type_cell_value` verbatim (a deliberately-quoted dict
-        // `'1'` stays `Str`). `format` is the only discriminator.
+        // The Array arm is FORMAT-AWARE (cute-dbt#127 DELTA 2, cute-dbt#138):
+        // dbt-core encodes csv as a `Value::Array` of all-string dicts, so a
+        // csv-format array threads the value-inferring `cell_from_csv_value`;
+        // a dict-format array keeps `cell_from_value` verbatim (a
+        // deliberately-quoted dict `'1'` stays `Str`). Each cell carries its
+        // authored display + canonical key. `format` is the only discriminator.
         Value::Array(elems) => {
-            let type_fn: fn(&Value) -> CellValue = match fmt {
-                FixtureFormat::Csv => type_csv_value,
-                FixtureFormat::Dict | FixtureFormat::Sql => type_cell_value,
+            let cell_fn: fn(&Value) -> Cell = match fmt {
+                FixtureFormat::Csv => cell_from_csv_value,
+                FixtureFormat::Dict | FixtureFormat::Sql => cell_from_value,
             };
-            Some(table_from_value_objects(elems, type_fn))
+            Some(table_from_value_objects(elems, cell_fn))
         }
         Value::String(s) => match fmt {
             FixtureFormat::Csv => Some(table_from_csv_text(s)),
@@ -831,7 +941,7 @@ pub fn table_from_yaml_fragment(rows_region: &str, format: Option<&str>) -> Opti
         }
         FixtureFormat::Dict => {
             let keyed = parse_block_dict_rows(rows_region);
-            Some(table_from_keyed_rows(&keyed, type_cell_scalar))
+            Some(table_from_keyed_rows(&keyed, cell_from_scalar))
         }
     }
 }
@@ -841,11 +951,13 @@ pub fn table_from_yaml_fragment(rows_region: &str, format: Option<&str>) -> Opti
 /// through `type_fn`. Columns are the first-seen union of object keys; absent
 /// keys → [`CellValue::Absent`].
 ///
-/// `type_fn` is the **format discriminator** (cute-dbt#127 DELTA 2): the
-/// caller passes [`type_cell_value`] for `format: dict` (a quoted `'1'` stays
-/// `Str`) and [`type_csv_value`] for `format: csv` (a `"1"` string cell
-/// infers `Number`). Mirrors the `type_fn` thread in [`table_from_keyed_rows`].
-fn table_from_value_objects(elems: &[Value], type_fn: fn(&Value) -> CellValue) -> FixtureTable {
+/// `cell_fn` is the **format discriminator** (cute-dbt#127 DELTA 2,
+/// cute-dbt#138): the caller passes [`cell_from_value`] for `format: dict`
+/// (a quoted `'1'` stays `Str`) and [`cell_from_csv_value`] for `format: csv`
+/// (a `"1"` string cell infers `Number`). Each cell carries its authored
+/// display plus the canonical key. Mirrors the `cell_fn` thread in
+/// [`table_from_keyed_rows`].
+fn table_from_value_objects(elems: &[Value], cell_fn: fn(&Value) -> Cell) -> FixtureTable {
     // First-seen union of keys across all object rows.
     let mut columns: Vec<String> = Vec::new();
     let mut seen: std::collections::HashSet<String> = std::collections::HashSet::new();
@@ -863,12 +975,11 @@ fn table_from_value_objects(elems: &[Value], type_fn: fn(&Value) -> CellValue) -
         .map(|elem| {
             let cells = columns
                 .iter()
-                .map(|col| {
-                    let value = match elem {
-                        Value::Object(map) => map.get(col).map_or(CellValue::Absent, type_fn),
-                        _ => CellValue::Absent,
-                    };
-                    Cell::new(value)
+                .map(|col| match elem {
+                    Value::Object(map) => map
+                        .get(col)
+                        .map_or_else(|| Cell::new(CellValue::Absent), cell_fn),
+                    _ => Cell::new(CellValue::Absent),
                 })
                 .collect();
             TableRow::new(cells)
@@ -878,18 +989,20 @@ fn table_from_value_objects(elems: &[Value], type_fn: fn(&Value) -> CellValue) -
 }
 
 /// Normalize a raw csv body into a [`FixtureTable`] (cells value-inferred via
-/// [`type_csv_token`]). Shared by the fusion NEW side and the csv OLD side.
+/// [`cell_from_csv_token`], raw token kept as display). Shared by the fusion
+/// NEW side and the csv OLD side.
 fn table_from_csv_text(text: &str) -> FixtureTable {
     let keyed = parse_csv_rows(text);
-    table_from_keyed_rows(&keyed, type_csv_token)
+    table_from_keyed_rows(&keyed, cell_from_csv_token)
 }
 
-/// Turn header-keyed string rows into a [`FixtureTable`], typing each token
-/// through `type_fn`. Columns are the first-seen union of keys; a row that
-/// lacks a column → [`CellValue::Absent`].
+/// Turn header-keyed string rows into a [`FixtureTable`], building each cell
+/// through `cell_fn` (authored display + canonical key). Columns are the
+/// first-seen union of keys; a row that lacks a column → a `Cell` with an
+/// [`Absent`](CellValue::Absent) key.
 fn table_from_keyed_rows(
     keyed: &[Vec<(String, String)>],
-    type_fn: fn(&str) -> CellValue,
+    cell_fn: fn(&str) -> Cell,
 ) -> FixtureTable {
     let mut columns: Vec<String> = Vec::new();
     let mut seen: std::collections::HashSet<String> = std::collections::HashSet::new();
@@ -906,11 +1019,9 @@ fn table_from_keyed_rows(
             let cells = columns
                 .iter()
                 .map(|col| {
-                    let value = row
-                        .iter()
+                    row.iter()
                         .find(|(k, _)| k == col)
-                        .map_or(CellValue::Absent, |(_, v)| type_fn(v));
-                    Cell::new(value)
+                        .map_or_else(|| Cell::new(CellValue::Absent), |(_, v)| cell_fn(v))
                 })
                 .collect();
             TableRow::new(cells)
@@ -1105,9 +1216,9 @@ mod tests {
         // A sparse dict: row 1 has {a,b}, row 2 has only {a} → row 2's b is
         // Absent, NOT Null. And Absent != Null as CellValues.
         let elems = vec![json!({"a": 1, "b": 2}), json!({"a": 3})];
-        let table = table_from_value_objects(&elems, type_cell_value);
+        let table = table_from_value_objects(&elems, cell_from_value);
         assert_eq!(table.columns, vec!["a".to_string(), "b".to_string()]);
-        assert_eq!(table.rows[1].cells[1].value, CellValue::Absent);
+        assert_eq!(table.rows[1].cells[1].key, CellValue::Absent);
         assert_ne!(CellValue::Absent, CellValue::Null);
     }
 
@@ -1168,7 +1279,7 @@ mod tests {
         // through type_csv_token). The convergence is stronger than the old
         // both-Str: a format-only reformat is a true no-op.
         assert_eq!(
-            fusion_tbl.rows[0].cells[0].value,
+            fusion_tbl.rows[0].cells[0].key,
             CellValue::Number("1".into())
         );
     }
@@ -1188,7 +1299,7 @@ mod tests {
         // Build a one-row, one-column table for `id` and pull out cell[0][0].
         fn cell(rows: &Value, fmt: &str) -> CellValue {
             let t = table_from_manifest_rows(rows, Some(fmt)).unwrap();
-            t.rows[0].cells[0].value.clone()
+            t.rows[0].cells[0].key.clone()
         }
         // dict side: a typed JSON scalar value for `id`.
         fn dict(v: Value) -> CellValue {
@@ -1265,12 +1376,12 @@ mod tests {
         let dict_tbl = table_from_manifest_rows(&rows, Some("dict")).unwrap();
         let csv_tbl = table_from_manifest_rows(&rows, Some("csv")).unwrap();
         assert_eq!(
-            dict_tbl.rows[0].cells[0].value,
+            dict_tbl.rows[0].cells[0].key,
             CellValue::Str("1".into()),
             "a deliberately-quoted dict '1' stays Str"
         );
         assert_eq!(
-            csv_tbl.rows[0].cells[0].value,
+            csv_tbl.rows[0].cells[0].key,
             CellValue::Number("1".into()),
             "the same array under csv format infers Number"
         );
@@ -1324,7 +1435,7 @@ mod tests {
     /// `BTreeMap`, so the column order is alphabetical, not insertion order).
     fn cell_by_col(t: &FixtureTable, row: usize, col: &str) -> CellValue {
         let idx = t.columns.iter().position(|c| c == col).expect("column");
-        t.rows[row].cells[idx].value.clone()
+        t.rows[row].cells[idx].key.clone()
     }
 
     /// Kill the csv-Array routing in `table_from_value_objects`: the threaded
@@ -1356,7 +1467,8 @@ mod tests {
 
     // ----- G. Parsers -----
 
-    /// Mirror tests/headless_csv_parser.rs's exact RFC 4180 cases.
+    /// The canonical RFC 4180 cases (since cute-dbt#138 the Rust parser is the
+    /// sole implementation — the retired JS twin's cases live here now).
     #[test]
     fn g22_parse_csv_mirrors_headless_edge_cases() {
         // Helper: parse and project to a simple Vec<Vec<(k,v)>> string form.
@@ -1524,11 +1636,8 @@ mod tests {
             table.columns,
             vec!["payer_key".to_string(), "payer_id".to_string()]
         );
-        assert_eq!(table.rows[0].cells[0].value, CellValue::Number("-1".into()));
-        assert_eq!(
-            table.rows[0].cells[1].value,
-            CellValue::Str("UNKNOWN".into())
-        );
+        assert_eq!(table.rows[0].cells[0].key, CellValue::Number("-1".into()));
+        assert_eq!(table.rows[0].cells[1].key, CellValue::Str("UNKNOWN".into()));
     }
 
     #[test]
@@ -1544,7 +1653,7 @@ mod tests {
         assert_eq!(row, vec![("note".into(), "'a, b'".into())]);
         // Through the normalizer the quoted value strips to Str("a, b").
         let table = table_from_yaml_fragment("      - {note: 'a, b'}", Some("dict")).unwrap();
-        assert_eq!(table.rows[0].cells[0].value, CellValue::Str("a, b".into()));
+        assert_eq!(table.rows[0].cells[0].key, CellValue::Str("a, b".into()));
     }
 
     #[test]
@@ -1630,7 +1739,7 @@ mod tests {
         let table = table_from_yaml_fragment(region, Some("csv")).unwrap();
         assert_eq!(table.columns, vec!["id".to_string(), "name".to_string()]);
         assert_eq!(table.rows.len(), 2);
-        assert_eq!(table.rows[0].cells[1].value, CellValue::Str("alice".into()));
+        assert_eq!(table.rows[0].cells[1].key, CellValue::Str("alice".into()));
     }
 
     #[test]
diff --git a/templates/report.html b/templates/report.html
index 53074821..d2ff2aec 100644
--- a/templates/report.html
+++ b/templates/report.html
@@ -1213,16 +1213,17 @@ <h2>Expected</h2>
       $sec.append(buildExternalFixtureAffordance(given.fixture));
       return $sec;
     }
-    var rows = extractRowsFromFixture(given);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = given.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
     if (given.format && given.format !== "dict") {
       $hdr.append($("<span>").addClass("format-badge").text("format: " + given.format));
     }
     $hdr.append($("<span>").addClass("row-count-badge")
-      .text(rows.length + " row" + (rows.length === 1 ? "" : "s")));
+      .text(rowCount + " row" + (rowCount === 1 ? "" : "s")));
     $sec.append($hdr);
-    var cols = collectColumns(rows);
     var diff = givenDataDiff(dataDiff, given.input);
-    $sec.append(buildFixtureView(cols, rows, "given-table", diff));
+    $sec.append(buildFixtureView(table, "given-table", diff));
     return $sec;
   }
 
@@ -1250,14 +1251,15 @@ <h2>Expected</h2>
       $body.append(buildExternalFixtureAffordance(t.expected.fixture));
       return;
     }
-    var rows = extractRowsFromFixture(t.expected);
-    $panel.find(".expected-rowcount").text(rows.length + " row" + (rows.length === 1 ? "" : "s"));
-    var cols = collectColumns(rows);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = t.expected.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
+    $panel.find(".expected-rowcount").text(rowCount + " row" + (rowCount === 1 ? "" : "s"));
     // The expect side's cell diff, when the test's expect fixture carried a
     // real change (cute-dbt#98). `data_diff.expect` already passed
     // `has_real_change()`, so its presence === "default this table to Diff".
     var diff = (t.data_diff && t.data_diff.expect) || undefined;
-    $body.append(buildFixtureView(cols, rows, "expected-table", diff));
+    $body.append(buildFixtureView(table, "expected-table", diff));
     initDataTablesIn($body);
   }
 
@@ -1281,26 +1283,13 @@ <h2>Expected</h2>
     return Boolean(obj && obj.format === "sql" && typeof obj.rows === "string");
   }
 
-  // Return a tabulable row array from a fixture, handling both
-  // array shapes (dict/csv-from-core) and raw-CSV-string shapes
-  // (csv-from-fusion). Returns `[]` for sql fixtures and other
-  // unexpected shapes — the sql path is dispatched separately by the
-  // caller via `isSqlStringFixture`.
-  function extractRowsFromFixture(fixture) {
-    if (!fixture) return [];
-    var rows = fixture.rows;
-    if (rows == null) return [];
-    if (Array.isArray(rows)) return rows;
-    if (typeof rows === "string" && fixture.format === "csv") {
-      return parseCsvRows(rows);
-    }
-    return [];
-  }
-
   // Render a sql-format fixture as a syntax-highlighted code block.
   // Reuses the compiled-SQL drawer's `.sql-block-wrap` shell so visual
   // style stays consistent. No copy button: the source of truth is the
-  // fixture in the manifest, not the rendered DOM.
+  // fixture in the manifest, not the rendered DOM. This is the ONE
+  // remaining non-tabulatable fallback path: a `format: sql` fixture has no
+  // cells, so the domain returns no `table` POD (cute-dbt#138) and the JS
+  // renders the raw SELECT here.
   function buildSqlCodeBlock(sql) {
     var $wrap = $("<div>").addClass("sql-block-wrap fixture-sql-block-wrap");
     var $pre = $("<pre>").addClass("sql-block").append(
@@ -1310,129 +1299,14 @@ <h2>Expected</h2>
     return $wrap;
   }
 
-  // RFC 4180 CSV parser — state machine over the input string.
-  //
-  // dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as
-  // a raw CSV body in the manifest. dbt-core 1.11+ pre-parses csv to
-  // an array of dicts. cute-dbt does the parse at render time so both
-  // engines render identically (cute-dbt#66).
-  //
-  // Behavior summary:
-  //   * First row is headers; subsequent rows are body values.
-  //   * Quoted fields strip the surrounding double-quotes.
-  //   * Double-quote inside a quoted field is the RFC 4180 escape:
-  //     `""` decodes to a literal `"`.
-  //   * Quoted fields may contain commas and newlines verbatim.
-  //   * A single trailing newline (LF or CRLF) is consumed — dbt YAML
-  //     `rows: |` block scalars always end in `\n`; preserving it
-  //     would create a spurious final empty row.
-  //   * An unterminated final row is accepted (graceful — the
-  //     renderer is not a validator).
-  //   * CRLF (`\r\n`) terminates a row as a single unit.
-  //   * Empty input or header-only input returns `[]`.
-  //
-  // Correctness is unit-tested via `tests/headless_csv_parser.rs` —
-  // the parser is exposed on `window.__cuteParseCsv` so that test can
-  // invoke it through the headless Chrome runtime.
-  function parseCsvRows(text) {
-    if (text == null) return [];
-    var t = String(text);
-    if (t.length === 0) return [];
-    // Strip exactly one trailing newline (LF or CRLF).
-    if (t.charAt(t.length - 1) === "\n") {
-      t = t.slice(0, -1);
-      if (t.length > 0 && t.charAt(t.length - 1) === "\r") {
-        t = t.slice(0, -1);
-      }
-    }
-    if (t.length === 0) return [];
-
-    var matrix = [];
-    var fields = [];
-    var field = "";
-    var inQuotes = false;
-    for (var i = 0; i < t.length; i++) {
-      var c = t.charAt(i);
-      if (inQuotes) {
-        if (c === '"') {
-          // RFC 4180 escape: doubled quote inside a quoted field
-          // collapses to a single literal quote.
-          if (i + 1 < t.length && t.charAt(i + 1) === '"') {
-            field += '"';
-            i++;
-            continue;
-          }
-          inQuotes = false;
-          continue;
-        }
-        field += c;
-        continue;
-      }
-      // Outside quotes.
-      if (c === '"' && field === "") {
-        inQuotes = true;
-        continue;
-      }
-      if (c === ',') {
-        fields.push(field);
-        field = "";
-        continue;
-      }
-      if (c === '\n' || c === '\r') {
-        fields.push(field);
-        matrix.push(fields);
-        fields = [];
-        field = "";
-        if (c === '\r' && i + 1 < t.length && t.charAt(i + 1) === '\n') {
-          i++;
-        }
-        continue;
-      }
-      field += c;
-    }
-    // Flush the final field/row. We always run this — the trailing-
-    // newline strip above guarantees at least one character of
-    // content survives to reach this branch in non-empty input.
-    fields.push(field);
-    matrix.push(fields);
-
-    if (matrix.length < 2) return [];
-    var headers = matrix[0];
-    var out = [];
-    for (var ri = 1; ri < matrix.length; ri++) {
-      var row = matrix[ri];
-      var obj = {};
-      for (var hi = 0; hi < headers.length; hi++) {
-        obj[headers[hi]] = hi < row.length ? row[hi] : "";
-      }
-      out.push(obj);
-    }
-    return out;
-  }
-
-  // Expose the parser on `window.__cuteParseCsv` so the headless
-  // integration test at `tests/headless_csv_parser.rs` can invoke it
-  // via `Runtime.Evaluate` and assert RFC 4180 edge cases (quoted
-  // fields, embedded commas, embedded newlines, escaped quotes,
-  // CRLF, empty input, header-only input). The `__` prefix marks it
-  // as a test-only seam; production code reaches it via the closure
-  // binding above.
-  if (typeof window !== "undefined") {
-    window.__cuteParseCsv = parseCsvRows;
-  }
-
-  function collectColumns(rows) {
-    var seen = {};
-    var order = [];
-    rows.forEach(function (row) {
-      if (row && typeof row === "object" && !Array.isArray(row)) {
-        Object.keys(row).forEach(function (k) {
-          if (!seen[k]) { seen[k] = true; order.push(k); }
-        });
-      }
-    });
-    return order;
-  }
+  // cute-dbt#138 — the JS csv/dict table parsers (`parseCsvRows`,
+  // `extractRowsFromFixture`, `collectColumns`) are RETIRED. The Current view
+  // now renders the Rust-computed `FixtureTable` POD that the domain ships on
+  // each given/expect payload (`table`), so csv/dict tabulation — and the
+  // engine-divergence handling above — happens once, in Rust
+  // (`table_from_manifest_rows`), and the template is a pure renderer. CSV
+  // RFC 4180 correctness is now owned by the Rust `parse_csv_rows` unit tests
+  // (the JS `__cuteParseCsv` headless seam retired with the function).
 
   function escapeHtml(s) {
     return String(s)
@@ -1789,7 +1663,7 @@ <h2>Expected</h2>
   function highlightYaml(yaml) { return emitTokens(tokenizeYaml(yaml)); }
 
   // Expose for headless JS unit tests (cute-dbt#69 / cute-dbt#132 — parallels
-  // the `__cuteParseCsv` seam introduced in cute-dbt#66).
+  // the `__cuteRenderYamlDiff` headless seam).
   window.__cuteHighlightYaml = highlightYaml;
   window.__cuteTokenizeSql = tokenizeSql;
   window.__cuteTokenizeYaml = tokenizeYaml;
@@ -2060,7 +1934,16 @@ <h2>Expected</h2>
     done(ok);
   }
 
-  function buildTable(columns, rows, cls) {
+  // The Current-view grid, rendered PURELY from the Rust-computed
+  // `FixtureTable` POD (cute-dbt#138). Each cell is a `{display, key}` Cell:
+  // the text is the authored `display` (so a csv `1.00` shows `1.00`, not the
+  // canonicalized `1`), while the numeric / NULL styling + sort order come
+  // from the canonical `key` (a `Number` key sorts numerically even when its
+  // display is `1.00`). The JS no longer parses csv/dict — the domain's
+  // `table_from_manifest_rows` owns that, and ships this POD on the wire.
+  function buildTable(table, cls) {
+    var columns = (table && table.columns) || [];
+    var rows = (table && table.rows) || [];
     var $wrap = $("<div>").addClass("table-fit");
     var $tbl = $("<table>").addClass(cls);
     var $thead = $("<thead>").appendTo($tbl);
@@ -2072,23 +1955,34 @@ <h2>Expected</h2>
     var $tbody = $("<tbody>").appendTo($tbl);
     rows.forEach(function (r) {
       var $tr = $("<tr>").appendTo($tbody);
-      columns.forEach(function (col) {
-        var cell = (r && typeof r === "object" && !Array.isArray(r)) ? r[col] : undefined;
-        var $td = $("<td>");
-        if (cell === null || cell === undefined) {
-          $td.addClass("cell-null").text("NULL");
-        } else if (typeof cell === "number") {
-          $td.addClass("cell-num").text(cell);
-        } else {
-          $td.text(String(cell));
-        }
-        $tr.append($td);
+      (r.cells || []).forEach(function (cell) {
+        $tr.append(decorateCurrentCell($("<td>"), cell));
       });
     });
     $wrap.append($tbl);
     return $wrap;
   }
 
+  // Render one Current-view Cell (`{display, key}`) into a <td>: the text is
+  // the authored `display`; the styling/sort affordance is read from `key.t`
+  // (the canonical type), never from the display token (cute-dbt#138). Shares
+  // the null/number styling with `decorateValueCell` so Current and Diff read
+  // identically.
+  function decorateCurrentCell($td, cell) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
+      return $td.addClass("cell-null").text("NULL");
+    }
+    if (key.t === "absent") {
+      return $td.addClass("cell-absent").text("");
+    }
+    if (key.t === "number") {
+      // Sort numerically on the canonical value; show the authored display.
+      return $td.addClass("cell-num").attr("data-order", key.v).text(cell.display);
+    }
+    return $td.text(cell.display);
+  }
+
   // cute-dbt#98 ----------------------------------------------------------
   //
   // The structured cell-diff renderer + the per-table Current↔Diff toggle.
@@ -2120,14 +2014,14 @@ <h2>Expected</h2>
   // is present — also build the Diff grid and a Current↔Diff toggle that flips
   // exactly one view visible at a time (the inactive view carries `hidden`).
   // Default-Diff when a diff exists, else Current with no toggle.
-  function buildFixtureView(columns, rows, cls, diff) {
+  function buildFixtureView(table, cls, diff) {
     if (!diff) {
-      // No cell diff for this table → today's behavior, byte-for-byte.
-      return buildTable(columns, rows, cls);
+      // No cell diff for this table → the plain Current grid from the POD.
+      return buildTable(table, cls);
     }
     var $wrap = $("<div>").addClass("fixture-view");
     var $diffView = buildDiffTable(diff, cls);
-    var $currentView = buildTable(columns, rows, cls).prop("hidden", true);
+    var $currentView = buildTable(table, cls).prop("hidden", true);
 
     var $toggle = $("<div>").addClass("cell-diff-toggle");
     var $diffBtn = $("<button>").attr("type", "button")
@@ -2200,67 +2094,56 @@ <h2>Expected</h2>
       return decorateValueCell($td, cell.old);
     }
     if (rowKind === "added" || rowKind === "unchanged" || !cell.changed) {
+      // Not flagged (added/unchanged, or a format-only change where the keys
+      // are equal): show the NEW authored value, no red/green (cute-dbt#138).
       return decorateValueCell($td, cell.new);
     }
-    // Modified + changed → inline old → new. Each side is type-aware via
-    // diffToken: a real null/absent keeps its own neutral styling and never
-    // borrows the value-change red/green (cute-dbt#132).
-    $td.addClass("cell-changed").attr("data-order", cellValueText(cell.new));
+    // Modified + changed → inline old → new, each side showing its authored
+    // display. Each side is type-aware via diffToken: a real null/absent keeps
+    // its own neutral styling and never borrows the value-change red/green
+    // (cute-dbt#132). The sort key is the NEW cell's authored display.
+    $td.addClass("cell-changed").attr("data-order", cellText(cell.new));
     $td.append(diffToken(cell.old, "cell-old"));
     $td.append($("<span>").addClass("cell-arrow").html("&rarr;"));
     $td.append(diffToken(cell.new, "cell-new"));
     return $td;
   }
 
-  // One side of an inline `old → new` changed cell. Type-aware so a real null
-  // or absent cell reads in its own neutral styling rather than the brilliant
-  // red/green reserved for a real value change: null → italic muted-gray
-  // "NULL"; absent → blank; any concrete value (including the string 'null')
-  // → the side's color via `sideClass` (.cell-old = red, .cell-new = green).
-  // This is what keeps a real NULL visually distinct from a string 'null'
-  // even mid-diff (cute-dbt#132).
-  function diffToken(cv, sideClass) {
-    if (!cv || cv.t === "null") {
+  // One side of an inline `old → new` changed cell. `cell` is a `{display,
+  // key}` Cell (cute-dbt#138): the text is the authored `display`; the
+  // neutral-vs-value styling is read from `key.t` so a real null or absent
+  // cell reads in its own neutral styling rather than the brilliant red/green
+  // reserved for a real value change: null → italic muted-gray "NULL"; absent
+  // → blank; any concrete value (including the string 'null') → the side's
+  // color via `sideClass` (.cell-old = red, .cell-new = green). This keeps a
+  // real NULL visually distinct from a string 'null' even mid-diff (#132).
+  function diffToken(cell, sideClass) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
       return $("<span>").addClass("cell-null").text("NULL");
     }
-    if (cv.t === "absent") {
+    if (key.t === "absent") {
       return $("<span>").addClass("cell-absent").text("");
     }
-    return $("<span>").addClass(sideClass).text(cellValueText(cv));
+    return $("<span>").addClass(sideClass).text(cell.display);
   }
 
-  // Render a single CellValue into a plain value cell, matching `buildTable`'s
-  // null/number styling so the Current and Diff views read identically.
-  function decorateValueCell($td, cv) {
-    if (!cv || cv.t === "null") {
-      // A literal null reads as NULL, matching the Current grid's
-      // `cell === null` styling.
-      return $td.addClass("cell-null").text("NULL");
-    }
-    if (cv.t === "absent") {
-      // An absent cell (a unified-diff column the row's source table lacks)
-      // is NOT a null — render it blank so an Absent->Null change stays
-      // visible and added/removed-column cells don't read as literal nulls.
-      return $td.addClass("cell-absent").text("");
-    }
-    if (cv.t === "number") {
-      return $td.addClass("cell-num").attr("data-order", cv.v).text(cv.v);
-    }
-    return $td.text(cellValueText(cv));
+  // Render a single `{display, key}` Cell into a plain value cell, matching
+  // `buildTable`'s null/number styling so the Current and Diff views read
+  // identically. Text is the authored `display`; styling/sort from `key`.
+  function decorateValueCell($td, cell) {
+    return decorateCurrentCell($td, cell);
   }
 
-  // The display text of an adjacently-tagged CellValue ({t, v}). `absent`
-  // renders empty; `null` renders the word NULL; bool/number/str verbatim.
-  function cellValueText(cv) {
-    if (!cv) return "";
-    switch (cv.t) {
-      case "null":   return "NULL";
-      case "absent": return "";
-      case "bool":   return cv.v ? "true" : "false";
-      case "number": return String(cv.v);
-      case "str":    return String(cv.v);
-      default:       return cv.v == null ? "" : String(cv.v);
-    }
+  // The display text of a `{display, key}` Cell (cute-dbt#138): a `null` key
+  // renders the word NULL, an `absent` key renders empty, otherwise the
+  // authored `display` token verbatim. Used for the changed-cell sort key.
+  function cellText(cell) {
+    var key = cell && cell.key;
+    if (!key) return "";
+    if (key.t === "null") return "NULL";
+    if (key.t === "absent") return "";
+    return cell.display == null ? "" : String(cell.display);
   }
 
   // Re-measure DataTables column widths for any (now-visible) table under
@@ -2275,13 +2158,15 @@ <h2>Expected</h2>
     });
   }
   // Exposed for headless JS unit tests (cute-dbt#98 — parallels the
-  // `__cuteRenderYamlDiff` / `__cuteParseCsv` seams). Returns the outer
-  // wrapper element so a test can assert the toggle + tinted markup.
+  // `__cuteRenderYamlDiff` seam). Takes the Rust-computed `FixtureTable` POD
+  // (cute-dbt#138) + an optional cell diff and returns the outer wrapper
+  // element so a test can assert the authored-display rendering, the toggle,
+  // and the tinted markup.
   if (typeof window !== "undefined") {
-    window.__cuteBuildFixtureView = function (columns, rows, cls, diff) {
-      return buildFixtureView(columns, rows, cls, diff)[0];
+    window.__cuteBuildFixtureView = function (table, cls, diff) {
+      return buildFixtureView(table, cls, diff)[0];
     };
-    window.__cuteCellValueText = cellValueText;
+    window.__cuteCellText = cellText;
   }
 
   function initDataTablesIn($scope) {
diff --git a/tests/cell_table_diff.rs b/tests/cell_table_diff.rs
index 2ff7ba5e..1367596f 100644
--- a/tests/cell_table_diff.rs
+++ b/tests/cell_table_diff.rs
@@ -178,11 +178,11 @@ fn h31_block_style_dict_single_cell_edit_is_one_modified_cell() {
     // Old payer_key was 1 (from the Removed line), new is 2 (manifest).
     use cute_dbt::domain::unit_test_table::CellValue;
     assert_eq!(
-        td.rows[0].cells[pk].old,
+        td.rows[0].cells[pk].old.key,
         CellValue::Number("1".into()),
         "old payer_key reconstructed from the Removed line"
     );
-    assert_eq!(td.rows[0].cells[pk].new, CellValue::Number("2".into()));
+    assert_eq!(td.rows[0].cells[pk].new.key, CellValue::Number("2".into()));
 }
 
 /// Inline-flow dict, single-cell edit. Each list item is a whole row, so
@@ -261,8 +261,14 @@ fn expect_side_single_cell_edit_populates_data_expect() {
         !td.rows[0].cells[label].changed,
         "label unchanged (Context)"
     );
-    assert_eq!(td.rows[0].cells[count].old, CellValue::Number("5".into()));
-    assert_eq!(td.rows[0].cells[count].new, CellValue::Number("7".into()));
+    assert_eq!(
+        td.rows[0].cells[count].old.key,
+        CellValue::Number("5".into())
+    );
+    assert_eq!(
+        td.rows[0].cells[count].new.key,
+        CellValue::Number("7".into())
+    );
 }
 
 /// csv block-scalar edit: the header line survives as Context (only the data
@@ -545,25 +551,33 @@ fn fusion_csv_raw_string_new_path_value_infers_through_the_adapter() {
     // Value inference (cute-dbt#127) fired on the raw-string fields:
     // `encounters` → Str, `1` → Number, `false` → Bool.
     assert_eq!(
-        given_tbl.rows[0].cells[0].value,
+        given_tbl.rows[0].cells[0].key,
         CellValue::Str("encounters".into())
     );
     assert_eq!(
-        given_tbl.rows[0].cells[1].value,
+        given_tbl.rows[0].cells[1].key,
         CellValue::Number("1".into()),
         "a fusion csv numeric field infers Number, not Str"
     );
     assert_eq!(
-        given_tbl.rows[0].cells[2].value,
+        given_tbl.rows[0].cells[2].key,
         CellValue::Bool(false),
         "a fusion csv `false` field infers Bool"
     );
-    assert_eq!(given_tbl.rows[1].cells[2].value, CellValue::Bool(true));
+    assert_eq!(given_tbl.rows[1].cells[2].key, CellValue::Bool(true));
+
+    // The given side authored `1` (row 0) / `true` (row 1) — and the authored
+    // DISPLAY token survives the IR (cute-dbt#138). `1` displays `1`, lowercase
+    // `true` displays `true`.
+    assert_eq!(given_tbl.rows[0].cells[1].display, "1");
+    assert_eq!(given_tbl.rows[1].cells[2].display, "true");
 
     // --- EXPECT side: the SAME logical data, reformatted (1 → 1.00, true →
-    // TRUE). Because csv is value-inferred, the expect table is EQUAL to the
-    // given table — a reformat-only change is a zero data diff (the #127
-    // headline guarantee), proven end-to-end on a committed fixture.
+    // TRUE). Because csv is value-inferred, the expect table's KEYS are EQUAL
+    // to the given table's keys — a reformat-only change is a zero data diff
+    // (the #127 equality guarantee). But the authored DISPLAY tokens DIFFER
+    // (`1` vs `1.00`), and cute-dbt#138 preserves them: the diff shows the
+    // authored value yet is NOT flagged.
     let expect = ut.expect();
     assert_eq!(expect.format(), Some("csv"));
     assert!(
@@ -572,9 +586,38 @@ fn fusion_csv_raw_string_new_path_value_infers_through_the_adapter() {
     );
     let expect_tbl = table_from_manifest_rows(expect.rows(), expect.format())
         .expect("the raw-string csv expect body parses");
+
+    // cute-dbt#138: the two tables are NO LONGER byte-equal — the `display`
+    // axis preserves the authored reformat (`1` vs `1.00`). The expect side
+    // authored `1.00`; that token survives.
     assert_eq!(
+        expect_tbl.rows[0].cells[1].display, "1.00",
+        "the authored `1.00` token is preserved (the #138 fidelity fix)"
+    );
+    assert_eq!(
+        expect_tbl.rows[1].cells[2].display, "TRUE",
+        "the authored uppercase `TRUE` token is preserved too (#138)"
+    );
+    assert_ne!(
         given_tbl, expect_tbl,
-        "1 vs 1.00 and false vs TRUE are value-equal: the reformatted expect \
-         table converges with the given table (cute-dbt#127)"
+        "the tables now differ on the authored DISPLAY axis (1 vs 1.00) — #138"
+    );
+
+    // But on the canonical KEY (equality) axis they converge cell-for-cell:
+    // a reformat-only change is still a zero data diff.
+    for (g_row, e_row) in given_tbl.rows.iter().zip(expect_tbl.rows.iter()) {
+        for (g_cell, e_cell) in g_row.cells.iter().zip(e_row.cells.iter()) {
+            assert_eq!(
+                g_cell.key, e_cell.key,
+                "keys converge: 1≡1.00, TRUE≡true (the #127 equality guarantee)"
+            );
+        }
+    }
+    // End-to-end: the diff verdict is zero change (keys equal) even though the
+    // displays differ — the headline #138 behavior on a committed fixture.
+    let diff = cute_dbt::domain::cell_diff::diff_fixture_tables(&given_tbl, &expect_tbl);
+    assert!(
+        !diff.has_real_change(),
+        "format-only reformat (1 → 1.00) is NOT flagged: keys equal despite display change"
     );
 }
diff --git a/tests/headless_csv_parser.rs b/tests/headless_csv_parser.rs
deleted file mode 100644
index bf13eb5c..00000000
--- a/tests/headless_csv_parser.rs
+++ /dev/null
@@ -1,273 +0,0 @@
-//! RFC 4180 CSV parser correctness — exercised through the rendered
-//! report via a real headless Chromium runtime.
-//!
-//! Why this test exists:
-//!
-//! dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as a
-//! raw CSV body in the manifest (verified empirically against a
-//! minimal probe project, 2026-05-26). dbt-core 1.11+ pre-parses csv
-//! to an array of dicts. cute-dbt's JS renderer (`templates/report.html`)
-//! contains a hand-rolled RFC 4180 parser so reports look identical
-//! regardless of which engine compiled the manifest. The parser is
-//! exposed as `window.__cuteParseCsv` for this test.
-//!
-//! The locked design (cute-dbt#66) requires comprehensive coverage
-//! for: quoted fields, embedded commas, escaped double-quotes,
-//! embedded newlines, trailing newline, empty input, header-only
-//! input, CRLF terminator. Each case is a separate sub-assertion so a
-//! parser regression points at the exact failing case.
-//!
-//! ## Runtime cost
-//!
-//! Shares the same CI job as `headless_zero_egress` and is `#[ignore]`
-//! by default. One Chrome cold-start covers both tests. Locally:
-//!
-//! ```bash
-//! cargo test --test headless_csv_parser -- --ignored
-//! ```
-//!
-//! Tracked: breezy-bays-labs/cute-dbt#66.
-
-#[path = "common/mod.rs"]
-mod common;
-
-use std::ffi::OsStr;
-use std::path::PathBuf;
-
-use headless_chrome::Browser;
-use headless_chrome::LaunchOptionsBuilder;
-use headless_chrome::protocol::cdp::Runtime;
-
-fn report_file_url(filename: &str) -> String {
-    let path = common::example_path(filename);
-    let p = path.to_str().expect("report path must be valid UTF-8");
-    format!("file://{p}")
-}
-
-/// One CSV parser case — the input fed to `window.__cuteParseCsv` and
-/// the expected JSON output as a Rust string literal. We compare the
-/// `JSON.stringify`'d parser output to `expected_json` for an exact
-/// structural match (key order in the row dicts is determined by the
-/// header row, so the comparison is well-defined).
-struct ParserCase {
-    name: &'static str,
-    input: &'static str,
-    /// Exact `JSON.stringify` form the parser must produce. Use
-    /// canonical no-whitespace JSON to match what the JS engine emits.
-    expected_json: &'static str,
-}
-
-const CASES: &[ParserCase] = &[
-    ParserCase {
-        name: "empty input",
-        input: "",
-        expected_json: "[]",
-    },
-    ParserCase {
-        name: "header-only input",
-        input: "id,name",
-        expected_json: "[]",
-    },
-    ParserCase {
-        name: "header-only with trailing newline",
-        input: "id,name\n",
-        expected_json: "[]",
-    },
-    ParserCase {
-        name: "single row, no trailing newline",
-        input: "id,name\n1,alice",
-        expected_json: r#"[{"id":"1","name":"alice"}]"#,
-    },
-    ParserCase {
-        name: "single row, trailing LF",
-        input: "id,name\n1,alice\n",
-        expected_json: r#"[{"id":"1","name":"alice"}]"#,
-    },
-    ParserCase {
-        name: "two rows, trailing LF",
-        input: "id,name\n1,alice\n2,bob\n",
-        expected_json: r#"[{"id":"1","name":"alice"},{"id":"2","name":"bob"}]"#,
-    },
-    ParserCase {
-        name: "CRLF terminator",
-        input: "id,name\r\n1,alice\r\n2,bob\r\n",
-        expected_json: r#"[{"id":"1","name":"alice"},{"id":"2","name":"bob"}]"#,
-    },
-    ParserCase {
-        name: "quoted field stripped",
-        input: "id,name\n1,\"alice\"\n",
-        expected_json: r#"[{"id":"1","name":"alice"}]"#,
-    },
-    ParserCase {
-        name: "quoted field with embedded comma",
-        input: "id,name\n1,\"alice, the brave\"\n",
-        expected_json: r#"[{"id":"1","name":"alice, the brave"}]"#,
-    },
-    ParserCase {
-        name: "RFC 4180 double-quote escape",
-        input: "id,note\n1,\"she said \"\"hello\"\"\"\n",
-        expected_json: r#"[{"id":"1","note":"she said \"hello\""}]"#,
-    },
-    ParserCase {
-        name: "quoted field with embedded LF",
-        input: "id,memo\n1,\"line one\nline two\"\n",
-        expected_json: r#"[{"id":"1","memo":"line one\nline two"}]"#,
-    },
-    ParserCase {
-        name: "quoted field with embedded CRLF",
-        input: "id,memo\n1,\"line one\r\nline two\"\n",
-        expected_json: r#"[{"id":"1","memo":"line one\r\nline two"}]"#,
-    },
-    ParserCase {
-        name: "empty fields via consecutive commas",
-        input: "a,b,c\n1,,3\n",
-        expected_json: r#"[{"a":"1","b":"","c":"3"}]"#,
-    },
-    ParserCase {
-        name: "trailing empty field",
-        input: "a,b,c\n1,2,\n",
-        expected_json: r#"[{"a":"1","b":"2","c":""}]"#,
-    },
-    ParserCase {
-        name: "mixed quoted and unquoted",
-        input: "a,b,c\n\"x\",y,\"z\"\n",
-        expected_json: r#"[{"a":"x","b":"y","c":"z"}]"#,
-    },
-    ParserCase {
-        name: "row shorter than header — missing trailing fields filled with empty string",
-        input: "a,b,c\n1,2\n",
-        expected_json: r#"[{"a":"1","b":"2","c":""}]"#,
-    },
-];
-
-#[test]
-#[ignore = "requires Chrome; runs explicitly in the headless-zero-egress CI job via `-- --ignored`"]
-fn cute_parse_csv_handles_rfc4180_edge_cases() {
-    // Reuse one of the committed example reports — any of them works
-    // since the parser is exposed on `window.__cuteParseCsv` regardless
-    // of which fixture seeded the report.
-    let filename = "playground-report.html";
-    let path = common::example_path(filename);
-    assert!(
-        path.exists(),
-        "examples/{filename} missing — regenerate via the example-report-up-to-date \
-         CI step or run: cargo run --bin cute-dbt -- \
-         --manifest tests/fixtures/playground-current.json \
-         --baseline-manifest tests/fixtures/playground-baseline.json \
-         --out examples/{filename}",
-    );
-
-    let chrome_path = std::env::var_os("CHROME").map(PathBuf::from);
-    let host_resolver = OsStr::new("--host-resolver-rules=MAP * ~NOTFOUND");
-    let no_first_run = OsStr::new("--no-first-run");
-    let no_default_check = OsStr::new("--no-default-browser-check");
-    let disable_breakpad = OsStr::new("--disable-breakpad");
-
-    let mut builder = LaunchOptionsBuilder::default();
-    builder.headless(true).sandbox(false).args(vec![
-        host_resolver,
-        no_first_run,
-        no_default_check,
-        disable_breakpad,
-    ]);
-    if let Some(p) = chrome_path.as_ref() {
-        builder.path(Some(p.clone()));
-    }
-    let opts = builder.build().expect("LaunchOptions must build");
-    let browser = Browser::new(opts).expect("Chromium must launch");
-
-    let tab = browser.new_tab().expect("new tab");
-    let url = report_file_url(filename);
-    assert!(url.starts_with("file://"));
-    tab.navigate_to(&url).expect("navigate to file:// URL");
-    tab.wait_until_navigated().expect("await navigation");
-
-    // Confirm the parser is exposed before driving cases — a missing
-    // seam is a different failure mode from a parser bug and the error
-    // message should reflect that.
-    let probe = tab
-        .call_method(Runtime::Evaluate {
-            expression: "typeof window.__cuteParseCsv === 'function'".to_string(),
-            object_group: None,
-            include_command_line_api: None,
-            silent: Some(true),
-            context_id: None,
-            return_by_value: Some(true),
-            generate_preview: None,
-            user_gesture: None,
-            await_promise: Some(false),
-            throw_on_side_effect: None,
-            timeout: None,
-            disable_breaks: None,
-            repl_mode: None,
-            allow_unsafe_eval_blocked_by_csp: None,
-            unique_context_id: None,
-            serialization_options: None,
-        })
-        .expect("evaluate parser-exposure probe");
-    let exposed = probe
-        .result
-        .value
-        .and_then(|v| v.as_bool())
-        .unwrap_or(false);
-    assert!(
-        exposed,
-        "window.__cuteParseCsv is not exposed — the parser cannot be \
-         driven through this seam. Re-check the test-seam guard at the \
-         bottom of the IIFE in templates/report.html.",
-    );
-
-    let mut failures: Vec<String> = Vec::new();
-    for case in CASES {
-        // JSON.stringify the input to safely embed it in the eval
-        // expression — covers embedded quotes, newlines, etc.
-        let input_json = serde_json::to_string(case.input).expect("input encodes as JSON");
-        let expression = format!(
-            "JSON.stringify(window.__cuteParseCsv({input_json}))",
-            input_json = input_json,
-        );
-        let eval = tab
-            .call_method(Runtime::Evaluate {
-                expression,
-                object_group: None,
-                include_command_line_api: None,
-                silent: Some(true),
-                context_id: None,
-                return_by_value: Some(true),
-                generate_preview: None,
-                user_gesture: None,
-                await_promise: Some(false),
-                throw_on_side_effect: None,
-                timeout: None,
-                disable_breaks: None,
-                repl_mode: None,
-                allow_unsafe_eval_blocked_by_csp: None,
-                unique_context_id: None,
-                serialization_options: None,
-            })
-            .expect("evaluate parser case");
-        let actual_json = eval
-            .result
-            .value
-            .and_then(|v| v.as_str().map(str::to_owned))
-            .unwrap_or_else(|| "<no value>".to_string());
-        if actual_json != case.expected_json {
-            failures.push(format!(
-                "case `{name}`:\n  input    = {input:?}\n  expected = {expected}\n  actual   = {actual}",
-                name = case.name,
-                input = case.input,
-                expected = case.expected_json,
-                actual = actual_json,
-            ));
-        }
-    }
-
-    let _ = tab.close(true);
-
-    assert!(
-        failures.is_empty(),
-        "RFC 4180 parser failed {n} case(s):\n{listing}",
-        n = failures.len(),
-        listing = failures.join("\n\n"),
-    );
-}
diff --git a/tests/headless_toggle.rs b/tests/headless_toggle.rs
index c135574f..f20d05bd 100644
--- a/tests/headless_toggle.rs
+++ b/tests/headless_toggle.rs
@@ -25,9 +25,8 @@
 //!
 //! ## Runtime cost
 //!
-//! Shares the same CI job as `headless_zero_egress` / `headless_csv_parser`
-//! and is `#[ignore]` by default. One Chrome cold-start covers all the
-//! headless tests. Locally:
+//! Shares the same CI job as `headless_zero_egress` and is `#[ignore]` by
+//! default. One Chrome cold-start covers all the headless tests. Locally:
 //!
 //! ```bash
 //! cargo test --test headless_toggle -- --ignored
@@ -1475,7 +1474,7 @@ fn show_all_inputs(tab: &Tab) {
 /// `id` value goes 1 → 2 (the synthetic toggle fixture).
 fn one_cell_modified_data_diff(input: &str) -> UnitTestDataDiff {
     use cute_dbt::domain::{
-        CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
+        Cell, CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
         RowChange, RowChangeKind,
     };
     UnitTestDataDiff {
@@ -1489,8 +1488,8 @@ fn one_cell_modified_data_diff(input: &str) -> UnitTestDataDiff {
                 rows: vec![RowChange {
                     kind: RowChangeKind::Modified,
                     cells: vec![CellChange {
-                        old: CellValue::Number("1".into()),
-                        new: CellValue::Number("2".into()),
+                        old: Cell::new(CellValue::Number("1".into())),
+                        new: Cell::new(CellValue::Number("2".into())),
                         changed: true,
                     }],
                 }],
@@ -1505,7 +1504,7 @@ fn one_cell_modified_data_diff(input: &str) -> UnitTestDataDiff {
 /// the string literal `"null"` — the cute-dbt#132 null-vs-string distinction.
 fn null_vs_string_null_data_diff(input: &str) -> UnitTestDataDiff {
     use cute_dbt::domain::{
-        CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
+        Cell, CellChange, CellValue, ColumnStatus, DiffColumn, FixtureTableDiff, NamedTableDiff,
         RowChange, RowChangeKind,
     };
     UnitTestDataDiff {
@@ -1527,15 +1526,15 @@ fn null_vs_string_null_data_diff(input: &str) -> UnitTestDataDiff {
                     cells: vec![
                         // new side is a REAL null -> italic muted-gray .cell-null.
                         CellChange {
-                            old: CellValue::Number("1".into()),
-                            new: CellValue::Null,
+                            old: Cell::new(CellValue::Number("1".into())),
+                            new: Cell::new(CellValue::Null),
                             changed: true,
                         },
                         // new side is the STRING "null" -> a normal value
                         // (.cell-new), NOT .cell-null.
                         CellChange {
-                            old: CellValue::Str("completed".into()),
-                            new: CellValue::Str("null".into()),
+                            old: Cell::new(CellValue::Str("completed".into())),
+                            new: Cell::new(CellValue::Str("null".into())),
                             changed: true,
                         },
                     ],
@@ -1885,6 +1884,42 @@ fn fusion_csv_format_only_shows_no_diff_cell_but_value_change_shows_old_to_new()
         eval_bool(&tab, "document.querySelector('.cell-old') === null"),
         "format-only reformat shows NO inline old → new diff cell",
     );
+    // cute-dbt#138 — the FIDELITY fix, asserted on the EXPECT grid. The
+    // fixture's NEW `expect` csv authors `1.00` for `quarantined_count` and
+    // `TRUE` for `is_dq_valid`; those authored tokens must survive to the cell
+    // text even though their canonical keys are Number("1") / Bool(true).
+    // Pre-#138 the Current grid normalized them to `1` / `true`.
+    assert!(
+        eval_bool(
+            &tab,
+            "Array.prototype.some.call(\
+               document.querySelectorAll('.expected-table td.cell-num'),\
+               function (td) { return td.textContent.trim() === '1.00'; })"
+        ),
+        "the EXPECT Current grid shows the AUTHORED `1.00`, not the normalized `1` (cute-dbt#138)",
+    );
+    assert!(
+        eval_bool(
+            &tab,
+            "Array.prototype.some.call(\
+               document.querySelectorAll('.expected-table td'),\
+               function (td) { return td.textContent.trim() === 'TRUE'; })"
+        ),
+        "the EXPECT Current grid shows the AUTHORED `TRUE`, not the normalized `true` (cute-dbt#138)",
+    );
+    // The `1.00` cell is still styled numeric (the class comes from the
+    // canonical key, not the display token), so the column sorts numerically.
+    assert!(
+        eval_bool(
+            &tab,
+            "Array.prototype.every.call(\
+               document.querySelectorAll('.expected-table td'),\
+               function (td) {\
+                 return td.textContent.trim() !== '1.00' || td.classList.contains('cell-num');\
+               })"
+        ),
+        "the authored `1.00` cell keeps numeric styling from its key (cute-dbt#138)",
+    );
 
     // ===== Value-change report: Diff view tints the changed cell 9 → 1 =====
     tab.navigate_to(&url_value_change)
diff --git a/tests/snapshots/render_integration__rendered_chrome_jaffle_shop.snap b/tests/snapshots/render_integration__rendered_chrome_jaffle_shop.snap
index f19cd0d1..1dc24418 100644
--- a/tests/snapshots/render_integration__rendered_chrome_jaffle_shop.snap
+++ b/tests/snapshots/render_integration__rendered_chrome_jaffle_shop.snap
@@ -435,7 +435,7 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
 
 </div>
 
-<script type="application/json" id="cute-dbt-data">{"baseline":"jaffle-shop-baseline.json","models":[{"name":"stg_customers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"renamed","role":"transform"},{"id":"stg_customers","role":"final"}],"edges":[{"from":"source","to":"renamed","edge_type":"from"},{"from":"renamed","to":"stg_customers","edge_type":"from"}]},"compiled_sql":{"renamed":"renamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)","source":"source as (\n    select * from \"jaffle_shop\".\"main\".\"raw_customers\"\n\n)","stg_customers":"select * from renamed"},"raw_sql":"with source as (\n\n    {#-\n    Normally we would select from the table here, but we are using seeds to load\n    our data in this project. Modified body — adds a trim() pass on names so\n    PR 5's StateComparator sees the body checksum change.\n    #}\n    select * from {{ ref('raw_customers') }}\n\n),\n\nrenamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)\n\nselect * from renamed","tests":[{"id":"unit_test.jaffle_shop.stg_customers.test_stg_customers_renames_columns","name":"test_stg_customers_renames_columns","target_model":"stg_customers","changed":true,"description":"stg_customers renames raw_customers.id/first_name/last_name to\ncustomer_id/first_name/last_name and otherwise passes rows through.\nSynthetic fixture rows — no real customer data.\n","tags":["quality","smoke"],"meta":{"owner":"data-eng","priority":"high"},"defined_in":"models/staging/unit_tests.yml","given":[{"input":"ref('raw_customers')","bound_to_node":"source","rows":[{"first_name":"Alice","id":1,"last_name":"A."},{"first_name":"Bob","id":2,"last_name":"B."}],"format":"dict"}],"expected":{"rows":[{"customer_id":1,"first_name":"Alice","last_name":"A."},{"customer_id":2,"first_name":"Bob","last_name":"B."}],"format":"dict"}}],"is_recursive":false}]}</script>
+<script type="application/json" id="cute-dbt-data">{"baseline":"jaffle-shop-baseline.json","models":[{"name":"stg_customers","dag":{"nodes":[{"id":"source","role":"import"},{"id":"renamed","role":"transform"},{"id":"stg_customers","role":"final"}],"edges":[{"from":"source","to":"renamed","edge_type":"from"},{"from":"renamed","to":"stg_customers","edge_type":"from"}]},"compiled_sql":{"renamed":"renamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)","source":"source as (\n    select * from \"jaffle_shop\".\"main\".\"raw_customers\"\n\n)","stg_customers":"select * from renamed"},"raw_sql":"with source as (\n\n    {#-\n    Normally we would select from the table here, but we are using seeds to load\n    our data in this project. Modified body — adds a trim() pass on names so\n    PR 5's StateComparator sees the body checksum change.\n    #}\n    select * from {{ ref('raw_customers') }}\n\n),\n\nrenamed as (\n\n    select\n        id as customer_id,\n        trim(first_name) as first_name,\n        trim(last_name) as last_name\n\n    from source\n\n)\n\nselect * from renamed","tests":[{"id":"unit_test.jaffle_shop.stg_customers.test_stg_customers_renames_columns","name":"test_stg_customers_renames_columns","target_model":"stg_customers","changed":true,"description":"stg_customers renames raw_customers.id/first_name/last_name to\ncustomer_id/first_name/last_name and otherwise passes rows through.\nSynthetic fixture rows — no real customer data.\n","tags":["quality","smoke"],"meta":{"owner":"data-eng","priority":"high"},"defined_in":"models/staging/unit_tests.yml","given":[{"input":"ref('raw_customers')","bound_to_node":"source","table":{"columns":["first_name","id","last_name"],"rows":[{"cells":[{"display":"Alice","key":{"t":"str","v":"Alice"}},{"display":"1","key":{"t":"number","v":"1"}},{"display":"A.","key":{"t":"str","v":"A."}}]},{"cells":[{"display":"Bob","key":{"t":"str","v":"Bob"}},{"display":"2","key":{"t":"number","v":"2"}},{"display":"B.","key":{"t":"str","v":"B."}}]}]},"rows":[{"first_name":"Alice","id":1,"last_name":"A."},{"first_name":"Bob","id":2,"last_name":"B."}],"format":"dict"}],"expected":{"table":{"columns":["customer_id","first_name","last_name"],"rows":[{"cells":[{"display":"1","key":{"t":"number","v":"1"}},{"display":"Alice","key":{"t":"str","v":"Alice"}},{"display":"A.","key":{"t":"str","v":"A."}}]},{"cells":[{"display":"2","key":{"t":"number","v":"2"}},{"display":"Bob","key":{"t":"str","v":"Bob"}},{"display":"B.","key":{"t":"str","v":"B."}}]}]},"rows":[{"customer_id":1,"first_name":"Alice","last_name":"A."},{"customer_id":2,"first_name":"Bob","last_name":"B."}],"format":"dict"}}],"is_recursive":false}]}</script>
 
 <script>
 <<inlined-asset>>
@@ -1197,16 +1197,17 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
       $sec.append(buildExternalFixtureAffordance(given.fixture));
       return $sec;
     }
-    var rows = extractRowsFromFixture(given);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = given.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
     if (given.format && given.format !== "dict") {
       $hdr.append($("<span>").addClass("format-badge").text("format: " + given.format));
     }
     $hdr.append($("<span>").addClass("row-count-badge")
-      .text(rows.length + " row" + (rows.length === 1 ? "" : "s")));
+      .text(rowCount + " row" + (rowCount === 1 ? "" : "s")));
     $sec.append($hdr);
-    var cols = collectColumns(rows);
     var diff = givenDataDiff(dataDiff, given.input);
-    $sec.append(buildFixtureView(cols, rows, "given-table", diff));
+    $sec.append(buildFixtureView(table, "given-table", diff));
     return $sec;
   }
 
@@ -1234,14 +1235,15 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
       $body.append(buildExternalFixtureAffordance(t.expected.fixture));
       return;
     }
-    var rows = extractRowsFromFixture(t.expected);
-    $panel.find(".expected-rowcount").text(rows.length + " row" + (rows.length === 1 ? "" : "s"));
-    var cols = collectColumns(rows);
+    // cute-dbt#138 — the Current view renders the Rust-computed POD directly.
+    var table = t.expected.table || { columns: [], rows: [] };
+    var rowCount = table.rows.length;
+    $panel.find(".expected-rowcount").text(rowCount + " row" + (rowCount === 1 ? "" : "s"));
     // The expect side's cell diff, when the test's expect fixture carried a
     // real change (cute-dbt#98). `data_diff.expect` already passed
     // `has_real_change()`, so its presence === "default this table to Diff".
     var diff = (t.data_diff && t.data_diff.expect) || undefined;
-    $body.append(buildFixtureView(cols, rows, "expected-table", diff));
+    $body.append(buildFixtureView(table, "expected-table", diff));
     initDataTablesIn($body);
   }
 
@@ -1265,26 +1267,13 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
     return Boolean(obj && obj.format === "sql" && typeof obj.rows === "string");
   }
 
-  // Return a tabulable row array from a fixture, handling both
-  // array shapes (dict/csv-from-core) and raw-CSV-string shapes
-  // (csv-from-fusion). Returns `[]` for sql fixtures and other
-  // unexpected shapes — the sql path is dispatched separately by the
-  // caller via `isSqlStringFixture`.
-  function extractRowsFromFixture(fixture) {
-    if (!fixture) return [];
-    var rows = fixture.rows;
-    if (rows == null) return [];
-    if (Array.isArray(rows)) return rows;
-    if (typeof rows === "string" && fixture.format === "csv") {
-      return parseCsvRows(rows);
-    }
-    return [];
-  }
-
   // Render a sql-format fixture as a syntax-highlighted code block.
   // Reuses the compiled-SQL drawer's `.sql-block-wrap` shell so visual
   // style stays consistent. No copy button: the source of truth is the
-  // fixture in the manifest, not the rendered DOM.
+  // fixture in the manifest, not the rendered DOM. This is the ONE
+  // remaining non-tabulatable fallback path: a `format: sql` fixture has no
+  // cells, so the domain returns no `table` POD (cute-dbt#138) and the JS
+  // renders the raw SELECT here.
   function buildSqlCodeBlock(sql) {
     var $wrap = $("<div>").addClass("sql-block-wrap fixture-sql-block-wrap");
     var $pre = $("<pre>").addClass("sql-block").append(
@@ -1294,129 +1283,14 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
     return $wrap;
   }
 
-  // RFC 4180 CSV parser — state machine over the input string.
-  //
-  // dbt-fusion 2.0-preview emits `format: csv` unit-test fixtures as
-  // a raw CSV body in the manifest. dbt-core 1.11+ pre-parses csv to
-  // an array of dicts. cute-dbt does the parse at render time so both
-  // engines render identically (cute-dbt#66).
-  //
-  // Behavior summary:
-  //   * First row is headers; subsequent rows are body values.
-  //   * Quoted fields strip the surrounding double-quotes.
-  //   * Double-quote inside a quoted field is the RFC 4180 escape:
-  //     `""` decodes to a literal `"`.
-  //   * Quoted fields may contain commas and newlines verbatim.
-  //   * A single trailing newline (LF or CRLF) is consumed — dbt YAML
-  //     `rows: |` block scalars always end in `\n`; preserving it
-  //     would create a spurious final empty row.
-  //   * An unterminated final row is accepted (graceful — the
-  //     renderer is not a validator).
-  //   * CRLF (`\r\n`) terminates a row as a single unit.
-  //   * Empty input or header-only input returns `[]`.
-  //
-  // Correctness is unit-tested via `tests/headless_csv_parser.rs` —
-  // the parser is exposed on `window.__cuteParseCsv` so that test can
-  // invoke it through the headless Chrome runtime.
-  function parseCsvRows(text) {
-    if (text == null) return [];
-    var t = String(text);
-    if (t.length === 0) return [];
-    // Strip exactly one trailing newline (LF or CRLF).
-    if (t.charAt(t.length - 1) === "\n") {
-      t = t.slice(0, -1);
-      if (t.length > 0 && t.charAt(t.length - 1) === "\r") {
-        t = t.slice(0, -1);
-      }
-    }
-    if (t.length === 0) return [];
-
-    var matrix = [];
-    var fields = [];
-    var field = "";
-    var inQuotes = false;
-    for (var i = 0; i < t.length; i++) {
-      var c = t.charAt(i);
-      if (inQuotes) {
-        if (c === '"') {
-          // RFC 4180 escape: doubled quote inside a quoted field
-          // collapses to a single literal quote.
-          if (i + 1 < t.length && t.charAt(i + 1) === '"') {
-            field += '"';
-            i++;
-            continue;
-          }
-          inQuotes = false;
-          continue;
-        }
-        field += c;
-        continue;
-      }
-      // Outside quotes.
-      if (c === '"' && field === "") {
-        inQuotes = true;
-        continue;
-      }
-      if (c === ',') {
-        fields.push(field);
-        field = "";
-        continue;
-      }
-      if (c === '\n' || c === '\r') {
-        fields.push(field);
-        matrix.push(fields);
-        fields = [];
-        field = "";
-        if (c === '\r' && i + 1 < t.length && t.charAt(i + 1) === '\n') {
-          i++;
-        }
-        continue;
-      }
-      field += c;
-    }
-    // Flush the final field/row. We always run this — the trailing-
-    // newline strip above guarantees at least one character of
-    // content survives to reach this branch in non-empty input.
-    fields.push(field);
-    matrix.push(fields);
-
-    if (matrix.length < 2) return [];
-    var headers = matrix[0];
-    var out = [];
-    for (var ri = 1; ri < matrix.length; ri++) {
-      var row = matrix[ri];
-      var obj = {};
-      for (var hi = 0; hi < headers.length; hi++) {
-        obj[headers[hi]] = hi < row.length ? row[hi] : "";
-      }
-      out.push(obj);
-    }
-    return out;
-  }
-
-  // Expose the parser on `window.__cuteParseCsv` so the headless
-  // integration test at `tests/headless_csv_parser.rs` can invoke it
-  // via `Runtime.Evaluate` and assert RFC 4180 edge cases (quoted
-  // fields, embedded commas, embedded newlines, escaped quotes,
-  // CRLF, empty input, header-only input). The `__` prefix marks it
-  // as a test-only seam; production code reaches it via the closure
-  // binding above.
-  if (typeof window !== "undefined") {
-    window.__cuteParseCsv = parseCsvRows;
-  }
-
-  function collectColumns(rows) {
-    var seen = {};
-    var order = [];
-    rows.forEach(function (row) {
-      if (row && typeof row === "object" && !Array.isArray(row)) {
-        Object.keys(row).forEach(function (k) {
-          if (!seen[k]) { seen[k] = true; order.push(k); }
-        });
-      }
-    });
-    return order;
-  }
+  // cute-dbt#138 — the JS csv/dict table parsers (`parseCsvRows`,
+  // `extractRowsFromFixture`, `collectColumns`) are RETIRED. The Current view
+  // now renders the Rust-computed `FixtureTable` POD that the domain ships on
+  // each given/expect payload (`table`), so csv/dict tabulation — and the
+  // engine-divergence handling above — happens once, in Rust
+  // (`table_from_manifest_rows`), and the template is a pure renderer. CSV
+  // RFC 4180 correctness is now owned by the Rust `parse_csv_rows` unit tests
+  // (the JS `__cuteParseCsv` headless seam retired with the function).
 
   function escapeHtml(s) {
     return String(s)
@@ -1773,7 +1647,7 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
   function highlightYaml(yaml) { return emitTokens(tokenizeYaml(yaml)); }
 
   // Expose for headless JS unit tests (cute-dbt#69 / cute-dbt#132 — parallels
-  // the `__cuteParseCsv` seam introduced in cute-dbt#66).
+  // the `__cuteRenderYamlDiff` headless seam).
   window.__cuteHighlightYaml = highlightYaml;
   window.__cuteTokenizeSql = tokenizeSql;
   window.__cuteTokenizeYaml = tokenizeYaml;
@@ -2044,7 +1918,16 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
     done(ok);
   }
 
-  function buildTable(columns, rows, cls) {
+  // The Current-view grid, rendered PURELY from the Rust-computed
+  // `FixtureTable` POD (cute-dbt#138). Each cell is a `{display, key}` Cell:
+  // the text is the authored `display` (so a csv `1.00` shows `1.00`, not the
+  // canonicalized `1`), while the numeric / NULL styling + sort order come
+  // from the canonical `key` (a `Number` key sorts numerically even when its
+  // display is `1.00`). The JS no longer parses csv/dict — the domain's
+  // `table_from_manifest_rows` owns that, and ships this POD on the wire.
+  function buildTable(table, cls) {
+    var columns = (table && table.columns) || [];
+    var rows = (table && table.rows) || [];
     var $wrap = $("<div>").addClass("table-fit");
     var $tbl = $("<table>").addClass(cls);
     var $thead = $("<thead>").appendTo($tbl);
@@ -2056,23 +1939,34 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
     var $tbody = $("<tbody>").appendTo($tbl);
     rows.forEach(function (r) {
       var $tr = $("<tr>").appendTo($tbody);
-      columns.forEach(function (col) {
-        var cell = (r && typeof r === "object" && !Array.isArray(r)) ? r[col] : undefined;
-        var $td = $("<td>");
-        if (cell === null || cell === undefined) {
-          $td.addClass("cell-null").text("NULL");
-        } else if (typeof cell === "number") {
-          $td.addClass("cell-num").text(cell);
-        } else {
-          $td.text(String(cell));
-        }
-        $tr.append($td);
+      (r.cells || []).forEach(function (cell) {
+        $tr.append(decorateCurrentCell($("<td>"), cell));
       });
     });
     $wrap.append($tbl);
     return $wrap;
   }
 
+  // Render one Current-view Cell (`{display, key}`) into a <td>: the text is
+  // the authored `display`; the styling/sort affordance is read from `key.t`
+  // (the canonical type), never from the display token (cute-dbt#138). Shares
+  // the null/number styling with `decorateValueCell` so Current and Diff read
+  // identically.
+  function decorateCurrentCell($td, cell) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
+      return $td.addClass("cell-null").text("NULL");
+    }
+    if (key.t === "absent") {
+      return $td.addClass("cell-absent").text("");
+    }
+    if (key.t === "number") {
+      // Sort numerically on the canonical value; show the authored display.
+      return $td.addClass("cell-num").attr("data-order", key.v).text(cell.display);
+    }
+    return $td.text(cell.display);
+  }
+
   // cute-dbt#98 ----------------------------------------------------------
   //
   // The structured cell-diff renderer + the per-table Current↔Diff toggle.
@@ -2104,14 +1998,14 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
   // is present — also build the Diff grid and a Current↔Diff toggle that flips
   // exactly one view visible at a time (the inactive view carries `hidden`).
   // Default-Diff when a diff exists, else Current with no toggle.
-  function buildFixtureView(columns, rows, cls, diff) {
+  function buildFixtureView(table, cls, diff) {
     if (!diff) {
-      // No cell diff for this table → today's behavior, byte-for-byte.
-      return buildTable(columns, rows, cls);
+      // No cell diff for this table → the plain Current grid from the POD.
+      return buildTable(table, cls);
     }
     var $wrap = $("<div>").addClass("fixture-view");
     var $diffView = buildDiffTable(diff, cls);
-    var $currentView = buildTable(columns, rows, cls).prop("hidden", true);
+    var $currentView = buildTable(table, cls).prop("hidden", true);
 
     var $toggle = $("<div>").addClass("cell-diff-toggle");
     var $diffBtn = $("<button>").attr("type", "button")
@@ -2184,67 +2078,56 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
       return decorateValueCell($td, cell.old);
     }
     if (rowKind === "added" || rowKind === "unchanged" || !cell.changed) {
+      // Not flagged (added/unchanged, or a format-only change where the keys
+      // are equal): show the NEW authored value, no red/green (cute-dbt#138).
       return decorateValueCell($td, cell.new);
     }
-    // Modified + changed → inline old → new. Each side is type-aware via
-    // diffToken: a real null/absent keeps its own neutral styling and never
-    // borrows the value-change red/green (cute-dbt#132).
-    $td.addClass("cell-changed").attr("data-order", cellValueText(cell.new));
+    // Modified + changed → inline old → new, each side showing its authored
+    // display. Each side is type-aware via diffToken: a real null/absent keeps
+    // its own neutral styling and never borrows the value-change red/green
+    // (cute-dbt#132). The sort key is the NEW cell's authored display.
+    $td.addClass("cell-changed").attr("data-order", cellText(cell.new));
     $td.append(diffToken(cell.old, "cell-old"));
     $td.append($("<span>").addClass("cell-arrow").html("&rarr;"));
     $td.append(diffToken(cell.new, "cell-new"));
     return $td;
   }
 
-  // One side of an inline `old → new` changed cell. Type-aware so a real null
-  // or absent cell reads in its own neutral styling rather than the brilliant
-  // red/green reserved for a real value change: null → italic muted-gray
-  // "NULL"; absent → blank; any concrete value (including the string 'null')
-  // → the side's color via `sideClass` (.cell-old = red, .cell-new = green).
-  // This is what keeps a real NULL visually distinct from a string 'null'
-  // even mid-diff (cute-dbt#132).
-  function diffToken(cv, sideClass) {
-    if (!cv || cv.t === "null") {
+  // One side of an inline `old → new` changed cell. `cell` is a `{display,
+  // key}` Cell (cute-dbt#138): the text is the authored `display`; the
+  // neutral-vs-value styling is read from `key.t` so a real null or absent
+  // cell reads in its own neutral styling rather than the brilliant red/green
+  // reserved for a real value change: null → italic muted-gray "NULL"; absent
+  // → blank; any concrete value (including the string 'null') → the side's
+  // color via `sideClass` (.cell-old = red, .cell-new = green). This keeps a
+  // real NULL visually distinct from a string 'null' even mid-diff (#132).
+  function diffToken(cell, sideClass) {
+    var key = cell && cell.key;
+    if (!key || key.t === "null") {
       return $("<span>").addClass("cell-null").text("NULL");
     }
-    if (cv.t === "absent") {
+    if (key.t === "absent") {
       return $("<span>").addClass("cell-absent").text("");
     }
-    return $("<span>").addClass(sideClass).text(cellValueText(cv));
+    return $("<span>").addClass(sideClass).text(cell.display);
   }
 
-  // Render a single CellValue into a plain value cell, matching `buildTable`'s
-  // null/number styling so the Current and Diff views read identically.
-  function decorateValueCell($td, cv) {
-    if (!cv || cv.t === "null") {
-      // A literal null reads as NULL, matching the Current grid's
-      // `cell === null` styling.
-      return $td.addClass("cell-null").text("NULL");
-    }
-    if (cv.t === "absent") {
-      // An absent cell (a unified-diff column the row's source table lacks)
-      // is NOT a null — render it blank so an Absent->Null change stays
-      // visible and added/removed-column cells don't read as literal nulls.
-      return $td.addClass("cell-absent").text("");
-    }
-    if (cv.t === "number") {
-      return $td.addClass("cell-num").attr("data-order", cv.v).text(cv.v);
-    }
-    return $td.text(cellValueText(cv));
+  // Render a single `{display, key}` Cell into a plain value cell, matching
+  // `buildTable`'s null/number styling so the Current and Diff views read
+  // identically. Text is the authored `display`; styling/sort from `key`.
+  function decorateValueCell($td, cell) {
+    return decorateCurrentCell($td, cell);
   }
 
-  // The display text of an adjacently-tagged CellValue ({t, v}). `absent`
-  // renders empty; `null` renders the word NULL; bool/number/str verbatim.
-  function cellValueText(cv) {
-    if (!cv) return "";
-    switch (cv.t) {
-      case "null":   return "NULL";
-      case "absent": return "";
-      case "bool":   return cv.v ? "true" : "false";
-      case "number": return String(cv.v);
-      case "str":    return String(cv.v);
-      default:       return cv.v == null ? "" : String(cv.v);
-    }
+  // The display text of a `{display, key}` Cell (cute-dbt#138): a `null` key
+  // renders the word NULL, an `absent` key renders empty, otherwise the
+  // authored `display` token verbatim. Used for the changed-cell sort key.
+  function cellText(cell) {
+    var key = cell && cell.key;
+    if (!key) return "";
+    if (key.t === "null") return "NULL";
+    if (key.t === "absent") return "";
+    return cell.display == null ? "" : String(cell.display);
   }
 
   // Re-measure DataTables column widths for any (now-visible) table under
@@ -2259,13 +2142,15 @@ th.col-removed { text-decoration: line-through; text-decoration-color: rgba(190,
     });
   }
   // Exposed for headless JS unit tests (cute-dbt#98 — parallels the
-  // `__cuteRenderYamlDiff` / `__cuteParseCsv` seams). Returns the outer
-  // wrapper element so a test can assert the toggle + tinted markup.
+  // `__cuteRenderYamlDiff` seam). Takes the Rust-computed `FixtureTable` POD
+  // (cute-dbt#138) + an optional cell diff and returns the outer wrapper
+  // element so a test can assert the authored-display rendering, the toggle,
+  // and the tinted markup.
   if (typeof window !== "undefined") {
-    window.__cuteBuildFixtureView = function (columns, rows, cls, diff) {
-      return buildFixtureView(columns, rows, cls, diff)[0];
+    window.__cuteBuildFixtureView = function (table, cls, diff) {
+      return buildFixtureView(table, cls, diff)[0];
     };
-    window.__cuteCellValueText = cellValueText;
+    window.__cuteCellText = cellText;
   }
 
   function initDataTablesIn($scope) {
diff --git a/tests/steps/cell_table_diff.rs b/tests/steps/cell_table_diff.rs
index c7046bb9..a0e0c70e 100644
--- a/tests/steps/cell_table_diff.rs
+++ b/tests/steps/cell_table_diff.rs
@@ -471,8 +471,25 @@ fn data_diff_one_changed_cell(world: &mut World, name: String, old: String, new:
         .filter(|c| c["changed"] == Value::Bool(true))
         .collect();
     assert_eq!(changed.len(), 1, "exactly one changed cell; got {cells:?}");
-    assert_eq!(changed[0]["old"]["v"], old, "old value is {old:?}");
-    assert_eq!(changed[0]["new"]["v"], new, "new value is {new:?}");
+    // cute-dbt#138 — each cell side is a `{display, key}` Cell; the value
+    // (equality) axis lives under `key`. The authored `display` echoes it for
+    // these string values, so assert both for completeness.
+    assert_eq!(
+        changed[0]["old"]["key"]["v"], old,
+        "old key value is {old:?}"
+    );
+    assert_eq!(
+        changed[0]["new"]["key"]["v"], new,
+        "new key value is {new:?}"
+    );
+    assert_eq!(
+        changed[0]["old"]["display"], old,
+        "old authored display is {old:?}"
+    );
+    assert_eq!(
+        changed[0]["new"]["display"], new,
+        "new authored display is {new:?}"
+    );
 }
 
 #[then(regex = r#"^the test "([^"]+)" carries no data diff$"#)]