data-designer-engine 0.7.0__tar.gz → 0.8.0__tar.gz

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: data-designer-engine
-Version: 0.7.0
+Version: 0.8.0
 Summary: Generation engine for DataDesigner synthetic data generation
 License-Expression: Apache-2.0
 Classifier: Development Status :: 4 - Beta
@@ -15,15 +15,17 @@ Classifier: Programming Language :: Python :: 3.14
 Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
 Requires-Python: >=3.10
 Requires-Dist: anyascii<1,>=0.3.3
+Requires-Dist: asciichartpy<2,>=1.5.25
 Requires-Dist: chardet<6,>=3.0.2
-Requires-Dist: cryptography<47,>=46.0.7
-Requires-Dist: data-designer-config==0.7.0
+Requires-Dist: cryptography<=49,>=48.0.1
+Requires-Dist: data-designer-config==0.8.0
 Requires-Dist: duckdb<2,>=1.5.0
 Requires-Dist: faker<21,>=20.1.0
 Requires-Dist: fsspec<2026,>=2025.3.0
 Requires-Dist: httpx-retries<1,>=0.4.2
 Requires-Dist: httpx<1,>=0.27.2
 Requires-Dist: huggingface-hub<2,>=1.0.1
+Requires-Dist: jinja2<4,>=3.1.6
 Requires-Dist: json-repair<1,>=0.48.0
 Requires-Dist: jsonpath-rust-bindings<2,>=1.0
 Requires-Dist: jsonschema<5,>=4.0.0
@@ -32,13 +34,17 @@ Requires-Dist: marko<3,>=2.1.2
 Requires-Dist: mcp<2,>=1.26.0
 Requires-Dist: networkx<4,>=3.0
 Requires-Dist: numpy<3,>=1.23.5
+Requires-Dist: pandas<3,>=2.3.3
+Requires-Dist: pyarrow<25,>=24
+Requires-Dist: pydantic<3,>=2.9.2
 Requires-Dist: pyjwt[crypto]<3,>=2.13.0
-Requires-Dist: python-multipart<1,>=0.0.29
+Requires-Dist: python-multipart<1,>=0.0.32
 Requires-Dist: ruff<1,>=0.14.10
 Requires-Dist: scipy<2,>=1.11.0
 Requires-Dist: sqlfluff<5,>=4.1.0
 Requires-Dist: starlette<2,>=1.2.0
 Requires-Dist: tiktoken<1,>=0.8.0
+Requires-Dist: wcwidth<1,>=0.2.13
 Description-Content-Type: text/markdown
 
 # data-designer-engine

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/pyproject.toml

@@ -34,8 +34,9 @@ bump = true
 [tool.hatch.metadata.hooks.uv-dynamic-versioning]
 dependencies = [
     "anyascii>=0.3.3,<1",
+    "asciichartpy>=1.5.25,<2",
     "chardet>=3.0.2,<6",  # Pulled in by sqlfluff
-    "cryptography>=46.0.7,<47", # 46.0.7 fixes CVE-2026-39892 pulled in by mcp
+    "cryptography>=48.0.1,<=49",
     "data-designer-config=={{ version }}",
     "duckdb>=1.5.0,<2",
     "faker>=20.1.0,<21",
@@ -43,6 +44,7 @@ dependencies = [
     "httpx>=0.27.2,<1",
     "httpx-retries>=0.4.2,<1",
     "huggingface-hub>=1.0.1,<2",
+    "jinja2>=3.1.6,<4",
     "json-repair>=0.48.0,<1",
     "jsonpath-rust-bindings>=1.0,<2",
     "jsonschema>=4.0.0,<5",
@@ -51,13 +53,17 @@ dependencies = [
     "mcp>=1.26.0,<2",
     "networkx>=3.0,<4",
     "numpy>=1.23.5,<3",
+    "pandas>=2.3.3,<3",
+    "pyarrow>=24,<25", # 24.0.0 fixes PyArrow security advisories
+    "pydantic>=2.9.2,<3",
     "pyjwt[crypto]>=2.13.0,<3", # 2.13.0 fixes PyJWT security advisories pulled in by mcp
-    "python-multipart>=0.0.29,<1", # 0.0.29 addresses multipart security advisories pulled in by mcp
+    "python-multipart>=0.0.32,<1", # 0.0.32 addresses multipart security advisories pulled in by mcp
     "ruff>=0.14.10,<1",
     "scipy>=1.11.0,<2",
     "sqlfluff>=4.1.0,<5",
     "starlette>=1.2.0,<2", # 1.2.0 fixes security advisory pulled in by mcp
     "tiktoken>=0.8.0,<1",
+    "wcwidth>=0.2.13,<1",
 ]
 
 [tool.hatch.build.targets.wheel]

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/src/data_designer/engine/analysis/utils/column_statistics_calculations.py

@@ -32,8 +32,6 @@ if TYPE_CHECKING:
 RANDOM_SEED = 42
 MAX_PROMPT_SAMPLE_SIZE = 1000
 WARNING_PREFIX = "⚠️ Error during column profile calculation: "
-TEXT_FIELD_AVG_SPACE_COUNT_THRESHOLD = 0.1
-
 logger = logging.getLogger(__name__)
 
 

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/src/data_designer/engine/column_generators/generators/custom.py

@@ -9,6 +9,7 @@ import asyncio
 import concurrent.futures
 import inspect
 import logging
+import time
 from typing import TYPE_CHECKING, Any
 
 import data_designer.lazy_heavy_imports as lazy
@@ -29,6 +30,7 @@ logger = logging.getLogger(__name__)
 # derivation in ``_compute_bridge_timeout`` exceeds this floor.
 # Module-level so tests can patch it for fast feedback.
 _BRIDGE_TIMEOUT_FLOOR_S: float = 60.0
+_BRIDGE_CANCEL_POLL_S: float = 0.1
 
 
 def _compute_bridge_timeout(
@@ -86,6 +88,7 @@ class _AsyncBridgedModelFacade:
                 "Use 'await model.agenerate()' in async custom columns."
             )
 
+        from data_designer.engine.context import is_run_cancellation_requested
         from data_designer.engine.dataset_builders.utils.async_concurrency import ensure_async_engine_loop
 
         # Honor a per-call ``timeout=`` override (passed straight through to the
@@ -99,19 +102,46 @@ class _AsyncBridgedModelFacade:
         conversation_restarts = int(kwargs.get("max_conversation_restarts", 0) or 0)
         bridge_timeout = _compute_bridge_timeout(per_request_timeout, correction_steps, conversation_restarts)
 
+        if is_run_cancellation_requested():
+            raise asyncio.CancelledError
+
+        async def agenerate_with_cancellation_check() -> tuple[Any, list]:
+            if is_run_cancellation_requested():
+                raise asyncio.CancelledError
+            return await facade.agenerate(*args, **kwargs)
+
         loop = ensure_async_engine_loop()
-        future = asyncio.run_coroutine_threadsafe(facade.agenerate(*args, **kwargs), loop)
+        coro = agenerate_with_cancellation_check()
         try:
-            return future.result(timeout=bridge_timeout)
-        except concurrent.futures.TimeoutError as exc:
-            future.cancel()
-            # Demoted to debug: the raised ModelTimeoutError already surfaces
-            # the timeout at the scheduler with full context, and the request-admission
-            # degraded-provider WARN is the user-facing signal under sustained
-            # bridge timeouts. Per-event WARN was noise on top of those.
-            logger.debug("Async model bridge timed out after %.0fs; coroutine cancelled", bridge_timeout)
-            # Raise as ModelTimeoutError so the scheduler classifies it retryable.
-            raise ModelTimeoutError(f"model.generate() bridge timed out after {bridge_timeout:.0f}s") from exc
+            future = asyncio.run_coroutine_threadsafe(coro, loop)
+        except RuntimeError as exc:
+            coro.close()
+            if is_run_cancellation_requested() or "interpreter shutdown" in str(exc):
+                raise asyncio.CancelledError from exc
+            raise
+        deadline = time.monotonic() + bridge_timeout
+        while True:
+            if is_run_cancellation_requested():
+                future.cancel()
+                raise asyncio.CancelledError
+            try:
+                return future.result(timeout=max(0.0, min(_BRIDGE_CANCEL_POLL_S, deadline - time.monotonic())))
+            except concurrent.futures.CancelledError as exc:
+                raise asyncio.CancelledError from exc
+            except concurrent.futures.TimeoutError as exc:
+                if is_run_cancellation_requested():
+                    future.cancel()
+                    raise asyncio.CancelledError from exc
+                if time.monotonic() < deadline:
+                    continue
+                future.cancel()
+                # Demoted to debug: the raised ModelTimeoutError already surfaces
+                # the timeout at the scheduler with full context, and the request-admission
+                # degraded-provider WARN is the user-facing signal under sustained
+                # bridge timeouts. Per-event WARN was noise on top of those.
+                logger.debug("Async model bridge timed out after %.0fs; coroutine cancelled", bridge_timeout)
+                # Raise as ModelTimeoutError so the scheduler classifies it retryable.
+                raise ModelTimeoutError(f"model.generate() bridge timed out after {bridge_timeout:.0f}s") from exc
 
     def __getattr__(self, name: str) -> Any:
         return getattr(object.__getattribute__(self, "_facade"), name)
@@ -277,38 +307,6 @@ class CustomColumnGenerator(ColumnGenerator[CustomColumnConfig]):
 
         return self._validate_output(result, keys_before, is_dataframe)
 
-    def _validate_cell_output(self, row: dict, keys_before: set[str]) -> dict:
-        """Validate a single row output (dict) for cell_by_cell; strip undeclared columns."""
-        expected_new = {self.config.name} | set(self.config.side_effect_columns)
-        result_keys = set(row.keys())
-
-        if self.config.name not in result_keys:
-            raise CustomColumnGenerationError(
-                f"Custom generator for column '{self.config.name}' did not create the expected column. "
-                f"The generator_function must add a column named '{self.config.name}' to the row."
-            )
-        missing = set(self.config.side_effect_columns) - result_keys
-        if missing:
-            raise CustomColumnGenerationError(
-                f"Custom generator for column '{self.config.name}' did not create declared side_effect_columns: "
-                f"{sorted(missing)}. Declared side_effect_columns must be added to the row."
-            )
-        removed = keys_before - result_keys
-        if removed:
-            raise CustomColumnGenerationError(
-                f"Custom generator for column '{self.config.name}' removed pre-existing columns: "
-                f"{sorted(removed)}. The generator_function must not remove any existing columns."
-            )
-        undeclared = (result_keys - keys_before) - expected_new
-        if undeclared:
-            logger.warning(
-                f"⚠️ Custom generator for column '{self.config.name}' created undeclared columns: "
-                f"{sorted(undeclared)}. These columns will be removed. "
-                f"To keep additional columns, declare them in @custom_column_generator(side_effect_columns=[...])."
-            )
-            row = {k: v for k, v in row.items() if k not in undeclared}
-        return row
-
     def _validate_output(
         self, result: dict | pd.DataFrame, keys_before: set[str], is_dataframe: bool
     ) -> dict | pd.DataFrame:

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/src/data_designer/engine/column_generators/utils/judge_score_factory.py

@@ -4,8 +4,9 @@
 from __future__ import annotations
 
 from enum import Enum
+from typing import Any
 
-from pydantic import BaseModel, ConfigDict, Field, create_model
+from pydantic import BaseModel, ConfigDict, Field, create_model, model_validator
 
 from data_designer.config.column_configs import Score
 
@@ -13,12 +14,54 @@ SCORING_FORMAT = "* {score}: {description}"
 SCORE_FIELD_DESCRIPTION_FORMAT = "Score Descriptions for {enum_name}:\n{scoring}"
 
 
+def _normalize_score_value(value: Any) -> str:
+    if isinstance(value, str):
+        return value.strip().casefold()
+    if isinstance(value, float) and value.is_integer():
+        return str(int(value))
+    return str(value).strip().casefold()
+
+
+def _coerce_score_value(value: Any, enum_type: type[Enum]) -> Any:
+    for member in enum_type:
+        # bool is a subclass of int in Python (True == 1, False == 0), so require
+        # matching bool-ness to keep True/False from silently matching 1/0 options.
+        if isinstance(value, bool) != isinstance(member.value, bool):
+            continue
+        if value == member.value:
+            return value
+
+    normalized_value = _normalize_score_value(value)
+    matches = [member.value for member in enum_type if _normalize_score_value(member.value) == normalized_value]
+    if len(matches) == 1:
+        return matches[0]
+    return value
+
+
 class BaseJudgeResponse(BaseModel):
     """Base model for all rubrics."""
 
     model_config = ConfigDict(use_enum_values=True)
     reasoning: str = Field(..., description="Reasoning for the assigned score.")
 
+    @model_validator(mode="before")
+    @classmethod
+    def coerce_score(cls, data: Any) -> Any:
+        if not isinstance(data, dict) or "score" not in data:
+            return data
+
+        score_field = cls.model_fields.get("score")
+        if score_field is None:
+            return data
+
+        score_type = score_field.annotation
+        if not isinstance(score_type, type) or not issubclass(score_type, Enum):
+            return data
+
+        coerced_data = data.copy()
+        coerced_data["score"] = _coerce_score_value(data["score"], score_type)
+        return coerced_data
+
 
 def _stringify_scoring(options: dict, enum_type: type[Enum]) -> str:
     """Convert score descriptions into a single text block."""

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/src/data_designer/engine/compiler.py

@@ -10,7 +10,7 @@ from data_designer.config.data_designer_config import DataDesignerConfig
 from data_designer.config.errors import InvalidConfigError
 from data_designer.config.sampler_params import UUIDSamplerParams
 from data_designer.engine.resources.resource_provider import ResourceProvider
-from data_designer.engine.resources.seed_reader import SeedReader
+from data_designer.engine.resources.seed_reader import SeedReader, SeedReaderConfigError
 from data_designer.engine.validation import ViolationLevel, rich_print_violations, validate_data_designer_config
 
 logger = logging.getLogger(__name__)
@@ -31,7 +31,10 @@ def _resolve_and_add_seed_columns(config: DataDesignerConfig, seed_reader: SeedR
     if not seed_reader:
         return
 
-    seed_col_names = seed_reader.get_column_names()
+    try:
+        seed_col_names = seed_reader.get_column_names()
+    except SeedReaderConfigError as error:
+        raise InvalidConfigError(str(error)) from error
     existing_columns = {column.name for column in config.columns}
     colliding_columns = {name for name in seed_col_names if name in existing_columns}
     if colliding_columns:

{data_designer_engine-0.7.0 → data_designer_engine-0.8.0}/src/data_designer/engine/context.py

@@ -4,11 +4,23 @@
 from __future__ import annotations
 
 from contextvars import ContextVar
+from threading import Event
 
 # Set per row group by the async scheduler before each task executes.
 # Value: (current_rg_index, total_rg_count) or None.
 current_row_group: ContextVar[tuple[int, int] | None] = ContextVar("current_row_group", default=None)
 
+# Shared cancellation signal for sync generator work running in thread-pool
+# workers. Context variables copy the Event object into worker threads, and the
+# scheduler flips the Event on cancellation.
+current_run_cancel_event: ContextVar[Event | None] = ContextVar("current_run_cancel_event", default=None)
+
+
+def is_run_cancellation_requested() -> bool:
+    cancel_event = current_run_cancel_event.get()
+    return cancel_event.is_set() if cancel_event is not None else False
+
+
 # Set while generating a row group. The value is the row group's planned start
 # offset in the full dataset, including row groups skipped during resume.
 current_row_group_start_offset: ContextVar[int | None] = ContextVar("current_row_group_start_offset", default=None)