mintmark 0.2.0__tar.gz → 0.3.0__tar.gz

{mintmark-0.2.0 → mintmark-0.3.0}/CHANGELOG.md

@@ -11,6 +11,46 @@ output for a fixed set of inputs. A change that alters emitted bytes for a fixed
 seed is a major version event even when no signature changed, because it breaks
 the reproducibility of every published manifest.
 
+## 0.3.0 - 2026-08-23
+
+### Security
+
+- HEALTH descriptors are condition classes only; treatment, surgery, therapy,
+  appointment, and examination phrases are no longer emitted under that label.
+- ETHNICITY no longer labels language/accommodation requests, and SEXUAL_LIFE no
+  longer labels marital or family-status phrases. Independent tests define the
+  forbidden semantic boundaries rather than trusting the descriptor file as its
+  own oracle.
+- CSV emission rejects string cells whose first meaningful character is a
+  spreadsheet formula trigger (`=`, `+`, `-`, or `@`), including whitespace- and
+  control-prefixed forms. It fails closed instead of changing the source value or
+  emitting active spreadsheet content.
+
+The descriptor corrections move document bytes for fixed seeds and therefore
+constitute a version event under this project's reproducibility contract.
+
+## 0.2.1 - 2026-08-23
+
+### Security
+
+- Template, enum, and reference-distribution weights are scaled once per
+  declaration rather than reparsed in record-count-multiplied hot loops. Pack
+  entity lexicon references are unique and bounded, and their surfaces are
+  composed once per mint.
+- Mint generation streams records and label sidecars into the private staging
+  directory. It retains only reference identifiers, enforces per-file and
+  aggregate output-byte budgets, and still commits atomically after the safe
+  identifier invariant succeeds.
+- Dataset verification streams JSONL, CSV, and label sidecars under line, file,
+  aggregate-byte, record, and diagnostic-count budgets. Semantic claims are
+  accumulated without retaining every parsed record.
+- Validator identifier mode now requires an affirmative caller choice even when
+  a recipe requires it. An omitted CLI or API policy is always `safe`; recipe
+  binding and exact-policy reproduction remain enforced.
+
+No generated record or sidecar bytes move for the same pack, recipe, seed,
+format, and explicit policy. Manifests record the patch engine version.
+
 ## 0.2.0 - 2026-08-23
 
 A security and conformance review of the three sector packs found one defect
@@ -319,3 +359,11 @@ and the name was still provisional. Both changed later the same day, in 0.1.1 an
 in the trademark clearance recorded in TRADEMARKS.md. This note is left as it was
 written rather than edited, because a changelog that quietly updates itself stops
 being a record.
+- Denylist comparisons now use compatibility normalization and remove combining
+  and default-ignorable characters, closing NFD, full-width, and zero-width
+  obfuscation bypasses without changing emitted span offsets.
+- Relationship allocation now enforces declared per-parent bounds and rejects
+  infeasible recipe totals before generation.
+- `packcheck` now exhaustively scans lexicons, field literals and parameters,
+  and every bounded template branch for denylisted entities and checksum-valid
+  identifier literals; its mini-mints preserve relationship constraints.

{mintmark-0.2.0 → mintmark-0.3.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: mintmark
-Version: 0.2.0
+Version: 0.3.0
 Summary: Deterministic, fully synthetic, Turkish-first labeled datasets with a provenance manifest.
 Author: Mintmark contributors
 License-Expression: Apache-2.0
@@ -33,9 +33,9 @@ Description-Content-Type: text/markdown
 
 <p align="center">
   <a href="https://github.com/lokomotifai/mintmark/actions/workflows/ci.yml"><img alt="CI" src="https://img.shields.io/github/actions/workflow/status/lokomotifai/mintmark/ci.yml?branch=main&amp;style=flat-square&amp;label=CI"></a>
-  <img alt="905 tests" src="https://img.shields.io/badge/tests-905-3C873A?style=flat-square">
+  <img alt="925 tests" src="https://img.shields.io/badge/tests-925-3C873A?style=flat-square">
   <img alt="18 invariants, each with a named test" src="https://img.shields.io/badge/invariants-18%20tested-3C873A?style=flat-square">
-  <a href="https://github.com/lokomotifai/mintmark/releases/tag/v0.2.0"><img alt="Release v0.2.0" src="https://img.shields.io/badge/release-v0.2.0-3C873A?style=flat-square"></a>
+  <a href="https://github.com/lokomotifai/mintmark/releases/tag/v0.3.0"><img alt="Release v0.3.0" src="https://img.shields.io/badge/release-v0.3.0-3C873A?style=flat-square"></a>
   <a href="https://pypi.org/project/mintmark/"><img alt="On PyPI" src="https://img.shields.io/pypi/v/mintmark?style=flat-square&amp;label=PyPI&amp;color=3C873A"></a>
   <a href="LICENSE"><img alt="Apache-2.0 license" src="https://img.shields.io/badge/license-Apache--2.0-3B3F46?style=flat-square"></a>
 </p>

{mintmark-0.2.0 → mintmark-0.3.0}/README.md

@@ -13,9 +13,9 @@
 
 <p align="center">
   <a href="https://github.com/lokomotifai/mintmark/actions/workflows/ci.yml"><img alt="CI" src="https://img.shields.io/github/actions/workflow/status/lokomotifai/mintmark/ci.yml?branch=main&amp;style=flat-square&amp;label=CI"></a>
-  <img alt="905 tests" src="https://img.shields.io/badge/tests-905-3C873A?style=flat-square">
+  <img alt="925 tests" src="https://img.shields.io/badge/tests-925-3C873A?style=flat-square">
   <img alt="18 invariants, each with a named test" src="https://img.shields.io/badge/invariants-18%20tested-3C873A?style=flat-square">
-  <a href="https://github.com/lokomotifai/mintmark/releases/tag/v0.2.0"><img alt="Release v0.2.0" src="https://img.shields.io/badge/release-v0.2.0-3C873A?style=flat-square"></a>
+  <a href="https://github.com/lokomotifai/mintmark/releases/tag/v0.3.0"><img alt="Release v0.3.0" src="https://img.shields.io/badge/release-v0.3.0-3C873A?style=flat-square"></a>
   <a href="https://pypi.org/project/mintmark/"><img alt="On PyPI" src="https://img.shields.io/pypi/v/mintmark?style=flat-square&amp;label=PyPI&amp;color=3C873A"></a>
   <a href="LICENSE"><img alt="Apache-2.0 license" src="https://img.shields.io/badge/license-Apache--2.0-3B3F46?style=flat-square"></a>
 </p>

{mintmark-0.2.0 → mintmark-0.3.0}/packs/example/pack.yaml

@@ -1,7 +1,7 @@
 schema_version: 1
 name: mintmark-example
 version: 0.1.0
-requires_core: ">=0.2,<0.3"
+requires_core: ">=0.3,<0.4"
 locale: tr-TR
 allowed_identifier_policies: [safe, validator]
 description: Fixture pack for the quickstart and the test suite. Not a sector pack.

{mintmark-0.2.0 → mintmark-0.3.0}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "hatchling.build"
 
 [project]
 name = "mintmark"
-version = "0.2.0"
+version = "0.3.0"
 description = "Deterministic, fully synthetic, Turkish-first labeled datasets with a provenance manifest."
 readme = "README.md"
 requires-python = ">=3.12,<3.13"

{mintmark-0.2.0 → mintmark-0.3.0}/schemas/manifest.schema.json

@@ -211,7 +211,7 @@
           "records": {
             "type": "integer",
             "minimum": 0,
-            "maximum": 100000
+            "maximum": 1000000
           }
         }
       }
@@ -292,12 +292,12 @@
               "target": {
                 "type": "integer",
                 "minimum": 0,
-                "maximum": 100000
+                "maximum": 1000000
               },
               "achieved": {
                 "type": "integer",
                 "minimum": 0,
-                "maximum": 100000
+                "maximum": 1000000
               },
               "met": {
                 "type": "boolean"

{mintmark-0.2.0 → mintmark-0.3.0}/schemas/pack.schema.json

@@ -76,6 +76,8 @@
       "additionalProperties": {
         "type": "array",
         "minItems": 1,
+        "maxItems": 128,
+        "uniqueItems": true,
         "items": {
           "type": "string",
           "pattern": "^[a-z][a-z0-9_]*$"
@@ -287,7 +289,7 @@
             "safe",
             "validator"
           ],
-          "description": "The policy this recipe is minted under. Binding: a mint that asks for a different one is refused. Absent means the caller chooses, defaulting to safe."
+          "description": "The policy this recipe is minted under. Binding: a mint that asks for a different one is refused. Validator mode still requires an explicit caller opt-in; an omitted caller choice is always safe."
         }
       }
     },

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/cli.py

@@ -31,7 +31,18 @@ from pathlib import Path
 from typing import Any
 
 from mintmark.annotate import Label, pin_digest
-from mintmark.identifiers import CHECKSUMMED
+from mintmark.engine.templates import (
+    Alternation,
+    EntitySlot,
+    FieldSlot,
+    IdentifierSlot,
+    LexiconSlot,
+    Literal,
+    Optional,
+    parse_template,
+)
+from mintmark.identifiers import ALL_ENGINES, CHECKSUMMED
+from mintmark.lexicons import load as load_denylist
 from mintmark.manifest import (
     MANIFEST_FILENAME,
     SUMS_FILENAME,
@@ -46,7 +57,8 @@ from mintmark.manifest.io import (
     DatasetIOError,
     DatasetReader,
 )
-from mintmark.mint import MintError, mint, packaged_pack_dir, resolve_pack, schema_dir
+from mintmark.manifest.safety import identifier_candidates, scalar_texts
+from mintmark.mint import MintError, asset_dir, mint, packaged_pack_dir, resolve_pack, schema_dir
 from mintmark.packs.loader import PackError
 from mintmark.packs.model import load_pack
 
@@ -98,8 +110,8 @@ def build_parser() -> argparse.ArgumentParser:
     mint_parser.add_argument(
         "--identifier-policy",
         choices=["safe", "validator"],
-        default=None,
-        help="only needed when the recipe does not name one; a recipe that does, decides",
+        default="safe",
+        help="safe by default; validator requires an explicit caller opt-in",
     )
     mint_parser.add_argument("--format", choices=["jsonl", "csv"], default="jsonl")
     mint_parser.add_argument("--json", action="store_true")
@@ -433,17 +445,116 @@ def _locate_pack(directory: Path, name: str) -> Path | None:
     return None
 
 
+MAX_TEMPLATE_SAFETY_EXPANSIONS = 10_000
+
+
+def _literal_expansions(nodes: tuple[Any, ...]) -> tuple[str, ...]:
+    """Enumerate every literal grammar branch under a strict cross-product cap."""
+    outputs = [""]
+    for node in nodes:
+        variants: tuple[str, ...]
+        match node:
+            case Literal(text=text):
+                variants = (text,)
+            case FieldSlot() | EntitySlot() | IdentifierSlot() | LexiconSlot():
+                variants = (" ",)
+            case Alternation(branches=branches):
+                variants = tuple(
+                    rendering for branch in branches for rendering in _literal_expansions(branch)
+                )
+            case Optional(body=body):
+                variants = ("", *_literal_expansions(body))
+            case _:
+                raise AssertionError(f"unknown template node {node!r}")
+        if len(outputs) * len(variants) > MAX_TEMPLATE_SAFETY_EXPANSIONS:
+            raise ValueError(
+                f"template expands beyond {MAX_TEMPLATE_SAFETY_EXPANSIONS} safety branches"
+            )
+        outputs = [prefix + suffix for prefix in outputs for suffix in variants]
+    return tuple(outputs)
+
+
+def _pack_safety(pack_root: Path, loaded: Any) -> tuple[Any, list[str]]:
+    """Exhaustively inspect declaration-controlled emitted strings."""
+    core = load_denylist(asset_dir("denylist") / "institutions-tr.txt")
+    extension_path = pack_root / "lexicons" / "denylist_extension.txt"
+    matcher = load_denylist(extension_path) if extension_path.exists() else core
+    problems: list[str] = []
+    if not matcher.covers(core):
+        missing = sorted(matcher.missing_from(core))[:5]
+        problems.append(f"denylist extension omits core entries: {missing}")
+
+    surfaces: list[tuple[str, str]] = []
+    for name, values in loaded.lexicons.items():
+        surfaces.extend((f"lexicon {name}", value) for value in values)
+    for record_type in loaded.record_types:
+        for field in record_type.fields:
+            source = f"field {record_type.type_name}.{field.name}"
+            if field.generator_kind == "literal":
+                surfaces.append((source, field.generator_argument))
+            surfaces.extend((source, text) for text in scalar_texts(field.params))
+    for set_name, entries in loaded.template_sets.items():
+        for entry in entries:
+            nodes = parse_template(
+                entry.text,
+                template_id=entry.id,
+                known_labels=frozenset(label.value for label in Label),
+                known_identifiers=frozenset(ALL_ENGINES),
+                known_lexicons=frozenset(loaded.lexicons),
+            )
+            try:
+                renderings = _literal_expansions(nodes)
+            except ValueError as exc:
+                problems.append(f"template {set_name}/{entry.id}: {exc}")
+                continue
+            surfaces.extend((f"template {set_name}/{entry.id}", text) for text in renderings)
+
+    for source, surface in surfaces:
+        for hit in matcher.scan(surface):
+            problems.append(f"{source}: {hit.render()}")
+        for candidate in identifier_candidates(surface):
+            valid_as = [
+                name for name, engine in CHECKSUMMED.items() if engine.is_checksum_valid(candidate)
+            ]
+            if valid_as:
+                problems.append(
+                    f"{source}: checksum-valid {'/'.join(valid_as)} literal {candidate!r} "
+                    "is forbidden by safe policy"
+                )
+    return matcher, problems
+
+
+def _mini_record_counts(loaded: Any, recipe: Any) -> dict[str, int]:
+    """Shrink a valid recipe while retaining every relationship bound."""
+    selected = {name: min(25, count) for name, count in recipe.records.items()}
+    for record_type in loaded.record_types:
+        for field in record_type.fields:
+            if field.ref is None:
+                continue
+            if selected.get(record_type.type_name, 0) == 0:
+                continue
+            parents = selected.get(field.ref.parent, 0)
+            low = parents * min(field.ref.counts)
+            high = parents * max(field.ref.counts)
+            selected[record_type.type_name] = min(
+                max(selected.get(record_type.type_name, 0), low), high
+            )
+    return selected
+
+
 def _cmd_packcheck(args: argparse.Namespace) -> int:
     pack_root = resolve_pack(args.pack)
     loaded = load_pack(pack_root)
 
     problems: list[str] = []
     warnings: list[str] = []
+    matcher, safety_problems = _pack_safety(pack_root, loaded)
+    problems.extend(safety_problems)
 
     with tempfile.TemporaryDirectory() as temporary:
         for recipe_name, recipe in sorted(loaded.recipes.items()):
             out = Path(temporary) / f"mini-{recipe_name}"
-            overrides = {name: min(25, count) for name, count in recipe.records.items()}
+            overrides = _mini_record_counts(loaded, recipe)
             try:
                 mint(
                     pack=pack_root,
@@ -467,12 +578,16 @@ def _cmd_packcheck(args: argparse.Namespace) -> int:
             problems.extend(f"recipe {recipe_name}: {p}" for p in report.problems)
 
             document = read_manifest(out)
+            for output in document["outputs"]:
+                output_path = out / output["path"]
+                for hit in matcher.scan(output_path.read_text(encoding="utf-8")):
+                    problems.append(f"recipe {recipe_name}: {output['path']}: {hit.render()}")
             for stat in document["stats"]["coverage_targets"]:
-                scaled = stat["target"] * 25
-                if stat["achieved"] * max(1, sum(recipe.records.values())) < scaled:
-                    warnings.append(
+                scaled_target = stat["target"] * max(1, sum(overrides.values()))
+                if stat["achieved"] * max(1, sum(recipe.records.values())) < scaled_target:
+                    problems.append(
                         f"recipe {recipe_name}: coverage target {stat['label']}="
-                        f"{stat['target']} may be infeasible at the declared counts"
+                        f"{stat['target']} is infeasible at the declared counts"
                     )
             shutil.rmtree(out, ignore_errors=True)
 

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/emit/__init__.py

@@ -10,12 +10,13 @@ from mintmark.emit.canonical import (
     FloatInOutputError,
     render_record,
 )
-from mintmark.emit.csv_writer import csv_header, render_csv_row
+from mintmark.emit.csv_writer import SpreadsheetFormulaError, csv_header, render_csv_row
 from mintmark.emit.writer import StagedOutput, staged_output
 
 __all__ = [
     "CanonicalEncoder",
     "FloatInOutputError",
+    "SpreadsheetFormulaError",
     "StagedOutput",
     "csv_header",
     "render_csv_row",

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/emit/csv_writer.py

@@ -13,9 +13,14 @@ from __future__ import annotations
 
 import csv
 import io
+import re
 from typing import Any
 
 
+class SpreadsheetFormulaError(ValueError):
+    """A string cell would be interpreted as active content by spreadsheets."""
+
+
 def csv_header(field_order: tuple[str, ...]) -> str:
     """The header row, in declaration order, RFC 4180 quoted, LF terminated."""
     return _render_row(list(field_order))
@@ -40,10 +45,29 @@ def render_csv_row(record: dict[str, Any], field_order: tuple[str, ...]) -> str:
                 f"field {name!r} carries a {type(value).__name__}, which has no CSV "
                 "representation. Emit this record type as JSONL, or flatten the field."
             )
+        if isinstance(value, str) and _formula_prefix(value) is not None:
+            raise SpreadsheetFormulaError(
+                f"field {name!r} begins with spreadsheet formula trigger "
+                f"{_formula_prefix(value)!r}; CSV emission refuses active cells"
+            )
         cells.append("" if value is None else str(value))
     return _render_row(cells)
 
 
+def _formula_prefix(value: str) -> str | None:
+    position = 0
+    while position < len(value) and (value[position].isspace() or ord(value[position]) < 0x20):
+        position += 1
+    candidate = value[position:]
+    if re.fullmatch(r"[+-]\d+(?:\.\d+)?", candidate) or re.fullmatch(
+        r"\+90 5\d{2} \d{3} \d{2} \d{2}", candidate
+    ):
+        return None
+    if candidate.startswith(("=", "+", "-", "@")):
+        return candidate[0]
+    return None
+
+
 def _render_row(cells: list[str]) -> str:
     buffer = io.StringIO()
     writer = csv.writer(buffer, lineterminator="\n", quoting=csv.QUOTE_MINIMAL)

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/engine/draws.py

@@ -16,6 +16,7 @@ tolerate.
 
 from __future__ import annotations
 
+from collections.abc import Sequence
 from dataclasses import dataclass
 from decimal import Decimal, InvalidOperation
 
@@ -126,8 +127,26 @@ def weighted_index(stream: SplitMix64, weights: list[str]) -> int:
     weights first would change which index a given draw selects and therefore
     change emitted bytes for a fixed seed.
     """
-    scaled = scale_weights(weights)
+    return weighted_index_scaled(stream, scale_weights(weights))
+
+
+def weighted_index_scaled(stream: SplitMix64, scaled: Sequence[int]) -> int:
+    """Draw from weights that were validated and scaled at declaration load time.
+
+    Generation hot paths call this form so a pack with thousands of choices
+    cannot multiply decimal parsing by its record count. The caller must retain
+    declaration order; doing so preserves the exact draw made by
+    :func:`weighted_index`.
+    """
+    if not scaled:
+        raise ValueError("scaled weights must not be empty")
+    if any(
+        not isinstance(weight, int) or isinstance(weight, bool) or weight < 0 for weight in scaled
+    ):
+        raise ValueError("scaled weights must be non-negative integers")
     total = sum(scaled)
+    if total <= 0:
+        raise ValueError("scaled weights sum to zero")
     target = bounded(stream, total)
     cumulative = 0
     for index, weight in enumerate(scaled):

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/engine/records.py

@@ -25,7 +25,7 @@ from dataclasses import dataclass
 from dataclasses import field as dataclass_field
 from typing import Any
 
-from mintmark.engine.draws import bounded, weighted_index
+from mintmark.engine.draws import bounded, scale_weights, weighted_index, weighted_index_scaled
 from mintmark.engine.streams import StreamFactory
 
 # One record is an ordered mapping of field name to emitted value. Order is the
@@ -118,9 +118,9 @@ def assign_children(
     """Return one parent index per child, honoring the count distribution.
 
     The distribution says how many children each parent should have. The recipe
-    says how many children exist. When they disagree, the distribution shapes the
-    allocation and the recipe's count wins on total, with the remainder walked
-    deterministically from parent zero upward.
+    says how many children exist. The requested total must fit the declared
+    per-parent minimum and maximum; deterministic adjustment never crosses those
+    bounds.
     """
     if parent_count <= 0:
         if child_count > 0:
@@ -129,11 +129,27 @@ def assign_children(
                 "a reference cannot dangle"
             )
         return []
+    if child_count == 0:
+        # A recipe may explicitly disable a record type. Cardinality applies
+        # when that relationship is present, not to a type the recipe omits.
+        return []
+
+    minimum = min(counts)
+    maximum = max(counts)
+    minimum_total = parent_count * minimum
+    maximum_total = parent_count * maximum
+    if not minimum_total <= child_count <= maximum_total:
+        raise ValueError(
+            f"{site}: {child_count} children cannot satisfy {minimum}..{maximum} "
+            f"children for each of {parent_count} parents; feasible total is "
+            f"{minimum_total}..{maximum_total}"
+        )
 
     per_parent: list[int] = []
+    scaled_weights = scale_weights(list(weights))
     for parent_index in range(parent_count):
         stream = factory.stream(f"{site}/{parent_index}/count")
-        per_parent.append(counts[weighted_index(stream, list(weights))])
+        per_parent.append(counts[weighted_index_scaled(stream, scaled_weights)])
 
     total = sum(per_parent)
     if total > child_count:
@@ -143,21 +159,26 @@ def assign_children(
         for parent_index in range(parent_count - 1, -1, -1):
             if surplus == 0:
                 break
-            take = min(per_parent[parent_index], surplus)
+            take = min(per_parent[parent_index] - minimum, surplus)
             per_parent[parent_index] -= take
             surplus -= take
     elif total < child_count:
         deficit = child_count - total
         parent_index = 0
         while deficit > 0:
-            per_parent[parent_index % parent_count] += 1
+            selected = parent_index % parent_count
             parent_index += 1
-            deficit -= 1
+            if per_parent[selected] < maximum:
+                per_parent[selected] += 1
+                deficit -= 1
+
+    if sum(per_parent) != child_count:
+        raise AssertionError(f"{site}: bounded relationship allocation did not converge")
 
     assignment: list[int] = []
     for parent_index, count in enumerate(per_parent):
         assignment.extend([parent_index] * count)
-    return assignment[:child_count]
+    return assignment
 
 
 @dataclass(slots=True)

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/engine/version.py

@@ -2,4 +2,4 @@
 
 from __future__ import annotations
 
-VERSION = "0.2.0"
+VERSION = "0.3.0"

{mintmark-0.2.0 → mintmark-0.3.0}/src/mintmark/lexicons/data/special_descriptors.yaml

@@ -11,24 +11,18 @@ labels:
     values:
     - kronik rahatsızlık
     - geçici iş göremezlik
-    - rutin kontrol
-    - istirahat raporu
-    - kontrol randevusu
-    - ameliyat sonrası izin
-    - sağlık raporu
-    - tedavi süreci
-    - fizik tedavi programı
-    - doğum öncesi izin
-    - iş kazası sonrası rapor
-    - sevk belgesi
-    - kontrol izni
-    - rapor uzatma talebi
-    - iş göremezlik bildirimi
     - engel durumu beyanı
-    - refakat izni
-    - periyodik muayene kaydı
-    - sağlık kurulu belgesi
-    - gebelik izni bildirimi
+    - sağlık durumu beyanı
+    - mesleki sağlık kısıtı
+    - hareket kısıtlılığı beyanı
+    - görme desteği ihtiyacı
+    - işitme desteği ihtiyacı
+    - alerji durumu beyanı
+    - solunum hassasiyeti beyanı
+    - kas iskelet kısıtı beyanı
+    - süreli sağlık kısıtı
+    - kalıcı sağlık kısıtı
+    - bulaşıcı olmayan rahatsızlık sınıfı
   RELIGION:
     note: Observance and accommodation, not belief content.
     values:
@@ -45,20 +39,16 @@ labels:
     - inanç temelli mazeret bildirimi
     - ibadet alanı talebi
   ETHNICITY:
-    note: Declared origin and language preference, as an HR or claims form would record them.
+    note: Declared ethnic or cultural origin only, as a form would record it.
     values:
     - etnik köken beyanı
-    - anadil tercihi
-    - iki dilli iletişim talebi
-    - tercüman talebi
-    - yerel dil desteği
-    - köken beyanı formu
-    - azınlık dili desteği talebi
-    - yazışma dili tercihi
-    - kültürel uyum notu
-    - isim yazımı düzeltme talebi
-    - çeviri desteği kaydı
-    - ana dilde bilgilendirme talebi
+    - soy kökeni beyanı
+    - kültürel köken beyanı
+    - etnik kimlik beyanı
+    - azınlık grubu aidiyeti beyanı
+    - etnik topluluk aidiyeti
+    - aile kökeni bölgesi beyanı
+    - köken bilgisi formu
   POLITICAL:
     note: Fictional parties and civic bodies. No real party is ever named.
     values:
@@ -77,18 +67,14 @@ labels:
   SEXUAL_LIFE:
     note: Form-level references only. Never narrative, never detail.
     values:
-    - özel yaşam beyanı
-    - medeni hâl güncellemesi
-    - aile durumu bildirimi
-    - eş bilgisi güncellemesi
-    - birlikte yaşanan kişi beyanı
-    - hayat arkadaşı bildirimi
-    - nikah durumu kaydı
-    - aile birleşimi talebi
-    - eş sağlık kapsamı talebi
-    - boşanma durumu güncellemesi
-    - birliktelik beyanı formu
-    - aile içi durum bildirimi
+    - cinsel yaşam verisi beyanı
+    - cinsel yönelim beyanı
+    - mahrem ilişki bilgisi formu
+    - cinsel yaşam mahremiyet talebi
+    - cinsel yaşama ilişkin veri düzeltme talebi
+    - mahrem partner bilgisi beyanı
+    - cinsel yaşama ilişkin erişim kısıtı
+    - cinsel yaşam verisi silme talebi
   CRIMINAL:
     note: Check categories, never findings. 'A record was requested' is a process step; 'a record
       exists' would be an accusation about a person, and this project does not make those even