Merge origin/main into gatekeeper/fix-2057-1

python/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "lancedb-python"
-version = "0.37.1-beta.0"
+version = "0.38.0-beta.2"
 publish = false
 edition.workspace = true
 description = "Python bindings for LanceDB"
@@ -15,10 +15,10 @@ name = "_lancedb"
 crate-type = ["cdylib"]
 
 [dependencies]
-arrow = { version = "58.0.0", features = ["pyarrow"] }
-async-trait = "0.1"
-bytes = "1"
-lancedb = { path = "../rust/lancedb", default-features = false }
+arrow = { workspace = true, features = ["pyarrow"] }
+async-trait.workspace = true
+bytes.workspace = true
+lancedb.workspace = true
 datafusion-common.workspace = true
 lance-core.workspace = true
 lance-namespace.workspace = true
@@ -26,24 +26,24 @@ lance-namespace-impls.workspace = true
 lance-io.workspace = true
 env_logger.workspace = true
 log.workspace = true
-pyo3 = { version = "0.28", features = ["extension-module", "abi3-py39", "chrono"] }
-chrono = { version = "0.4", default-features = false, features = ["clock"] }
+pyo3 = { version = "0.28", features = ["extension-module", "abi3-py310", "chrono"] }
+chrono.workspace = true
 pyo3-async-runtimes = { version = "0.28", features = [
     "attributes",
     "tokio-runtime",
 ] }
-pin-project = "1.1.5"
+pin-project.workspace = true
 futures.workspace = true
-serde = "1"
-serde_json = "1"
+serde.workspace = true
+serde_json.workspace = true
 snafu.workspace = true
-tokio = { version = "1.40", features = ["sync", "rt-multi-thread"] }
+tokio.workspace = true
 libc = "0.2"
 
 [build-dependencies]
 pyo3-build-config = { version = "0.28", features = [
     "extension-module",
-    "abi3-py39",
+    "abi3-py310",
 ] }
 
 [features]

python/pyproject.toml

@@ -8,7 +8,7 @@ dependencies = [
     "overrides>=0.7; python_version<'3.12'",
     "packaging>=23.0",
     "pyarrow>=16",
-    "pydantic>=1.10",
+    "pydantic>=2.7.4,<3",
     "tqdm>=4.27.0",
     "lance-namespace>=0.3.2"
 ]
@@ -60,7 +60,7 @@ tests = [
     "pytest-asyncio>=0.21",
     "duckdb>=0.9.0",
     "pytz>=2023.3",
-    "polars>=0.19, <=1.3.0",
+    "polars>=0.19, <=1.32.3",
     "pyarrow<25",
     "pyarrow-stubs>=16.0",
     "pylance==9.0.0rc1",
@@ -140,6 +140,7 @@ include = [
     "python/lancedb/remote/errors.py",
     "python/lancedb/embeddings/__init__.py",
     "python/lancedb/_lancedb.pyi",
+    "python/type_tests/connect.py",
 ]
 exclude = ["python/tests/"]
 pythonVersion = "3.13"

python/python/lancedb/__init__.py

@@ -12,6 +12,7 @@ __version__ = importlib.metadata.version("lancedb")
 
 from ._lancedb import connect as lancedb_connect
 from ._lancedb import FtsToken
+from ._lancedb import LsmWriteSpec
 from ._lancedb import tokenize as _tokenize
 from .common import URI, sanitize_uri
 from urllib.parse import urlparse
@@ -21,6 +22,12 @@ from .remote.db import RemoteDBConnection
 from .expr import Expr, col, lit, func
 from .schema import blob, vector, BlobType
 from .job import AsyncJob, Job
+from .functions import (
+    FunctionApplication as FunctionApplication,
+    FunctionBinding as FunctionBinding,
+    FunctionVersion as FunctionVersion,
+    PythonRuntimeSpec as PythonRuntimeSpec,
+)
 from .table import AsyncTable, Table
 from .types import BaseTokenizerType
 from ._lancedb import Session
@@ -518,6 +525,7 @@ __all__ = [
     "Job",
     "LanceDBConnection",
     "LanceNamespaceDBConnection",
+    "LsmWriteSpec",
     "RemoteDBConnection",
     "Session",
     "Table",

python/python/lancedb/_lancedb.pyi

@@ -198,6 +198,9 @@ class Connection(object):
     async def drop_table(
         self, name: str, namespace_path: Optional[List[str]] = None
     ) -> None: ...
+    async def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job: ...
     async def drop_all_tables(
         self, namespace_path: Optional[List[str]] = None
     ) -> None: ...
@@ -335,6 +338,14 @@ class Table:
     ) -> list[FtsToken]: ...
     async def delete(self, filter: Union[str, PyExpr]) -> DeleteResult: ...
     async def add_columns(self, columns: list[tuple[str, str]]) -> AddColumnsResult: ...
+    async def add_computed_columns(
+        self, columns: list[tuple[str, str]]
+    ) -> AddColumnsResult: ...
+    async def add_function_columns(
+        self, application_json: str, output_name: Optional[str]
+    ) -> AddColumnsResult: ...
+    async def refresh_column(self, column: str) -> RefreshColumnResult: ...
+    async def refresh_column_async(self, column: str) -> Job: ...
     async def add_columns_with_schema(self, schema: pa.Schema) -> AddColumnsResult: ...
     async def alter_columns(
         self, columns: list[dict[str, Any]]
@@ -355,6 +366,10 @@ class Table:
     async def set_lsm_write_spec(self, spec: LsmWriteSpec) -> None: ...
     async def unset_lsm_write_spec(self) -> None: ...
     async def get_lsm_write_spec(self) -> Optional[LsmWriteSpec]: ...
+    async def checkpoint_lsm(self) -> None: ...
+    async def flush_lsm(self) -> None: ...
+    async def compact_lsm(self) -> None: ...
+    async def get_lsm_stats(self, include_generation_rows: bool) -> Optional[dict]: ...
     async def close_lsm_writers(self) -> None: ...
     @property
     def tags(self) -> Tags: ...
@@ -649,9 +664,10 @@ class LsmWriteSpec:
     def identity(column: str) -> "LsmWriteSpec": ...
     @staticmethod
     def unsharded() -> "LsmWriteSpec": ...
-    def with_maintained_indexes(self, indexes: List[str]) -> "LsmWriteSpec":
-        """Return a copy of this spec asking the MemWAL to keep the named
-        indexes up to date as rows are appended."""
+    def with_maintained_indexes(self, indexes: Optional[List[str]]) -> "LsmWriteSpec":
+        """Set which indexes the MemWAL keeps up to date. None resolves every
+        index on the table at install, failing if one cannot be maintained;
+        a list is verbatim, empty means none."""
         ...
     def with_writer_config_defaults(self, defaults: Dict[str, str]) -> "LsmWriteSpec":
         """Return a copy of this spec recording the given default
@@ -666,13 +682,19 @@ class LsmWriteSpec:
     @property
     def num_buckets(self) -> Optional[int]: ...
     @property
-    def maintained_indexes(self) -> List[str]: ...
+    def maintained_indexes(self) -> Optional[List[str]]:
+        """Indexes the MemWAL keeps up to date, or None for every supported one."""
+        ...
     @property
     def writer_config_defaults(self) -> Dict[str, str]: ...
 
 class AddColumnsResult:
     version: int
 
+class RefreshColumnResult:
+    rows_filled: int
+    version: int
+
 class AlterColumnsResult:
     version: int
 

python/python/lancedb/db.py

@@ -524,6 +524,12 @@ class DBConnection(EnforceOverrides):
             namespace_path = []
         raise NotImplementedError
 
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        raise NotImplementedError
+
     def rename_table(
         self,
         cur_name: str,
@@ -1186,6 +1192,20 @@ class LanceDBConnection(DBConnection):
             )
         )
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job.
+
+        The table may become unavailable before its data files are removed.
+        Call :meth:`Job.wait` to wait for cleanup to finish.
+        """
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(self._conn.drop_table_async(name, namespace_path=namespace_path))
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def drop_all_tables(self, namespace_path: Optional[List[str]] = None):
         if namespace_path is None:
@@ -1963,6 +1983,23 @@ class AsyncConnection(object):
             if f"Table '{name}' was not found" not in str(e):
                 raise e
 
+    async def drop_table_async(
+        self,
+        name: str,
+        *,
+        namespace_path: Optional[List[str]] = None,
+    ) -> AsyncJob:
+        """Start dropping a table and return its cleanup job.
+
+        The table may become unavailable before its data files are removed.
+        Await :meth:`AsyncJob.wait` to wait for cleanup to finish.
+        """
+        if namespace_path is None:
+            namespace_path = []
+        return AsyncJob(
+            await self._inner.drop_table_async(name, namespace_path=namespace_path)
+        )
+
     async def drop_all_tables(self, namespace_path: Optional[List[str]] = None):
         """Drop all tables from the database.
 

python/python/lancedb/embeddings/base.py

@@ -26,7 +26,6 @@ class EmbeddingFunction(BaseModel, ABC):
     3. ndims() which returns the number of dimensions of the vector column
     """
 
-    __slots__ = ("__weakref__",)  # pydantic 1.x compatibility
     max_retries: int = (
         7  # Setting 0 disables retires. Maybe this should not be enabled by default,
     )

python/python/lancedb/embeddings/bedrock.py

@@ -7,8 +7,7 @@ from functools import cached_property
 from typing import List, Union
 
 import numpy as np
-
-from lancedb.pydantic import PYDANTIC_VERSION
+from pydantic import ConfigDict
 
 from ..util import attempt_import_or_raise
 from .base import TextEmbeddingFunction
@@ -67,13 +66,7 @@ class BedRockText(TextEmbeddingFunction):
     source_input_type: str = "search_document"
     query_input_type: str = "search_query"
 
-    if PYDANTIC_VERSION.major < 2:  # Pydantic 1.x compat
-
-        class Config:
-            keep_untouched = (cached_property,)
-    else:
-        model_config = dict()
-        model_config["ignored_types"] = (cached_property,)
+    model_config = ConfigDict(ignored_types=(cached_property,))
 
     def ndims(self):
         # return len(self._generate_embedding("test"))

python/python/lancedb/embeddings/gemini_text.py

@@ -7,8 +7,7 @@ from functools import cached_property
 from typing import List, Optional, Union
 
 import numpy as np
-
-from lancedb.pydantic import PYDANTIC_VERSION
+from pydantic import ConfigDict
 
 from ..util import attempt_import_or_raise
 from .base import TextEmbeddingFunction
@@ -87,13 +86,7 @@ class GeminiText(TextEmbeddingFunction):
     query_task_type: str = "retrieval_query"
     source_task_type: str = "retrieval_document"
 
-    if PYDANTIC_VERSION.major < 2:  # Pydantic 1.x compat
-
-        class Config:
-            keep_untouched = (cached_property,)
-    else:
-        model_config = dict()
-        model_config["ignored_types"] = (cached_property,)
+    model_config = ConfigDict(ignored_types=(cached_property,))
 
     def ndims(self):
         if self.dim:

python/python/lancedb/embeddings/imagebind.py

@@ -7,14 +7,13 @@ from typing import List, Union
 
 import numpy as np
 import pyarrow as pa
+from pydantic import ConfigDict
 
 from ..util import attempt_import_or_raise
 from .base import EmbeddingFunction
 from .registry import register
 from .utils import AUDIO, IMAGES, TEXT
 
-from lancedb.pydantic import PYDANTIC_VERSION
-
 
 @register("imagebind")
 class ImageBindEmbeddings(EmbeddingFunction):
@@ -31,13 +30,7 @@ class ImageBindEmbeddings(EmbeddingFunction):
     device: str = "cpu"
     normalize: bool = False
 
-    if PYDANTIC_VERSION.major < 2:  # Pydantic 1.x compat
-
-        class Config:
-            keep_untouched = (cached_property,)
-    else:
-        model_config = dict()
-        model_config["ignored_types"] = (cached_property,)
+    model_config = ConfigDict(ignored_types=(cached_property,))
 
     def __init__(self, *args, **kwargs):
         super().__init__(*args, **kwargs)

python/python/lancedb/embeddings/instructor.py

@@ -101,8 +101,7 @@ class InstructorEmbeddingFunction(TextEmbeddingFunction):
 
     @weak_lru(maxsize=1)
     def ndims(self):
-        model = self.get_model()
-        return model.encode("foo").shape[0]
+        return len(self.generate_embeddings([[self.source_instruction, "foo"]])[0])
 
     def compute_query_embeddings(self, query: str, *args, **kwargs) -> List[np.array]:
         return self.generate_embeddings([[self.query_instruction, query]])

python/python/lancedb/embeddings/jinaai.py

@@ -87,12 +87,13 @@ class JinaEmbeddings(EmbeddingFunction):
         if isinstance(image, bytes):
             image_dict = {"image": base64.b64encode(image).decode("utf-8")}
         elif isinstance(image, (str, Path)):
-            parsed = urlparse.urlparse(image)
-            # TODO handle drive letter on windows.
+            parsed = urlparse(str(image))
             PIL_Image = attempt_import_or_raise("PIL.Image", "pillow")
             if parsed.scheme == "file":
                 pil_image = PIL_Image.open(parsed.path)
-            elif parsed.scheme == "":
+            elif parsed.scheme == "" or (os.name == "nt" and len(parsed.scheme) == 1):
+                # A Windows drive letter parses as a one-character scheme
+                # ("C:\\img.png" -> scheme="c"), so treat it as a local path.
                 pil_image = PIL_Image.open(image if os.name == "nt" else parsed.path)
             elif parsed.scheme.startswith("http"):
                 pil_image = PIL_Image.open(io.BytesIO(url_retrieve(image)))

python/python/lancedb/embeddings/transformers.py

@@ -7,8 +7,7 @@ from typing import List, Any
 
 import numpy as np
 
-from pydantic import PrivateAttr
-from lancedb.pydantic import PYDANTIC_VERSION
+from pydantic import ConfigDict, PrivateAttr
 
 from ..util import attempt_import_or_raise
 from .base import EmbeddingFunction
@@ -59,13 +58,7 @@ class TransformersEmbeddingFunction(EmbeddingFunction):
         )
         self._model.to(self.device)
 
-    if PYDANTIC_VERSION.major < 2:  # Pydantic 1.x compat
-
-        class Config:
-            keep_untouched = (cached_property,)
-    else:
-        model_config = dict()
-        model_config["ignored_types"] = (cached_property,)
+    model_config = ConfigDict(ignored_types=(cached_property,))
 
     def ndims(self):
         self._ndims = self._model.config.hidden_size

python/python/lancedb/functions.py

@@ -0,0 +1,380 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+"""Canonical values exchanged with LanceDB Enterprise Function services.
+
+These immutable models contain client/wire state only. Catalog persistence,
+environment bake, secret resolution, and execution are owned by Sophon.
+"""
+
+from __future__ import annotations
+
+import json
+from collections.abc import Mapping
+from typing import Any, Optional
+
+from pydantic import (
+    BaseModel,
+    ConfigDict,
+    Field,
+    conint,
+    field_validator,
+    model_validator,
+)
+
+_Int32 = conint(strict=True, ge=-(2**31), le=2**31 - 1)
+_UInt32 = conint(strict=True, ge=0, le=2**32 - 1)
+_UInt64 = conint(strict=True, ge=0, le=2**64 - 1)
+
+
+class _FrozenDict(dict):
+    def _immutable(self, *args, **kwargs):
+        raise TypeError("remote canonical values are immutable")
+
+    __setitem__ = _immutable
+    __delitem__ = _immutable
+    clear = _immutable
+    pop = _immutable
+    popitem = _immutable
+    setdefault = _immutable
+    update = _immutable
+
+    def __ior__(self, other):
+        self._immutable()
+
+
+def _freeze_value(value):
+    if isinstance(value, Mapping):
+        return _FrozenDict({key: _freeze_value(child) for key, child in value.items()})
+    if isinstance(value, (list, tuple)):
+        return tuple(_freeze_value(child) for child in value)
+    return value
+
+
+def _validate_literal(value):
+    if isinstance(value, float):
+        raise ValueError(
+            "floating-point Function literals are not part of the Slice 1 "
+            "canonical wire contract"
+        )
+    if isinstance(value, int) and not isinstance(value, bool):
+        if not -(2**63) <= value <= 2**64 - 1:
+            raise ValueError(
+                "Function integer literal is outside the canonical JSON range"
+            )
+    elif isinstance(value, Mapping):
+        for child in value.values():
+            _validate_literal(child)
+    elif isinstance(value, (list, tuple)):
+        for child in value:
+            _validate_literal(child)
+    return value
+
+
+def _known_wire_value(value):
+    if isinstance(value, _RemoteValue):
+        return value._known_dict()
+    if isinstance(value, Mapping):
+        return {key: _known_wire_value(child) for key, child in value.items()}
+    if isinstance(value, (list, tuple)):
+        return [_known_wire_value(child) for child in value]
+    return value
+
+
+class _RemoteValue(BaseModel):
+    model_config = ConfigDict(extra="ignore", frozen=True)
+
+    @model_validator(mode="after")
+    def _freeze_mappings(self):
+        for name, value in self.__dict__.items():
+            object.__setattr__(self, name, _freeze_value(value))
+        return self
+
+    @classmethod
+    def from_json(cls, payload: str):
+        return cls.model_validate_json(payload)
+
+    def _known_dict(self) -> dict[str, Any]:
+        known = {}
+        for name, field in self.__class__.model_fields.items():
+            value = getattr(self, name)
+            if value is None:
+                continue
+            if not field.is_required():
+                default_factory = field.default_factory
+                if default_factory is not None and value == default_factory():
+                    continue
+                if default_factory is None and value == field.default:
+                    continue
+            known[name] = _known_wire_value(value)
+        return known
+
+    def _copy(self, *, update: Mapping[str, Any]):
+        update = {name: _freeze_value(value) for name, value in update.items()}
+        return self.model_copy(update=update)
+
+    def to_canonical_json(self) -> str:
+        return json.dumps(
+            self._known_dict(),
+            ensure_ascii=False,
+            allow_nan=False,
+            sort_keys=True,
+            separators=(",", ":"),
+        )
+
+
+class _OpenRemoteValue(_RemoteValue):
+    """Forward-readable value whose extras stay out of canonical encoding."""
+
+    if _PYDANTIC_V2:
+        model_config = {"extra": "allow", "frozen": True}
+    else:
+
+        class Config:
+            allow_mutation = False
+            extra = "allow"
+
+    def _unknown_field_names(self) -> set[str]:
+        if _PYDANTIC_V2:
+            return set((self.__pydantic_extra__ or {}).keys())
+        return set(self.__dict__) - set(self.__fields__)
+
+
+class FunctionArtifact(_RemoteValue):
+    """Content-addressed Python artifact identity."""
+
+    kind: str
+    digest: str
+    entrypoint: str
+
+
+class FunctionParameter(_RemoteValue):
+    name: str
+    arrow_type: str
+    nullable: bool
+
+
+class FunctionResultField(_OpenRemoteValue):
+    name: str
+    arrow_type: str
+    nullable: bool
+
+
+class FunctionOutput(_OpenRemoteValue):
+    """Scalar or ordered named-struct output; unknown kinds remain decodable."""
+
+    kind: str
+    arrow_type: Optional[str] = None
+    nullable: Optional[bool] = None
+    fields: tuple[FunctionResultField, ...] = ()
+
+
+class FunctionSignature(_RemoteValue):
+    inputs: tuple[FunctionParameter, ...]
+    output: FunctionOutput
+
+
+class PythonEnvironmentSpec(_RemoteValue):
+    """One Sophon-managed Python environment source."""
+
+    kind: str
+    packages: tuple[str, ...] = ()
+    path: Optional[str] = None
+    modules: tuple[str, ...] = ()
+    image: Optional[str] = None
+
+
+class PythonRuntimeSpec(_RemoteValue):
+    """Remote runtime definition with non-secret environment values.
+
+    V1 supports ``kind="python"``. Newer runtime kinds remain readable, while
+    their unknown payload fields are intentionally not retained by the client.
+    """
+
+    kind: str
+    python_version: Optional[str] = None
+    environment: Optional[PythonEnvironmentSpec] = None
+    env: Optional[Mapping[str, str]] = None
+
+    @model_validator(mode="after")
+    def _validate_runtime_kind(self):
+        if self.kind == "python":
+            if self.python_version is None:
+                raise ValueError("python runtime requires python_version")
+            if self.environment is None:
+                raise ValueError("python runtime requires environment")
+        else:
+            object.__setattr__(self, "python_version", None)
+            object.__setattr__(self, "environment", None)
+            object.__setattr__(self, "env", None)
+        return self
+
+
+class FunctionVersion(_RemoteValue):
+    """An exact immutable Function version returned by Enterprise.
+
+    Scheduling resources, priority, concurrency, and retry policy belong to
+    the submitting Job and are not part of this identity.
+    """
+
+    name: str
+    version: str
+    artifact: FunctionArtifact
+    signature: FunctionSignature
+    runtime: PythonRuntimeSpec
+    runtime_digest: str
+    environment_digest: str
+    required_secrets: tuple[str, ...] = ()
+    created_at: str
+
+
+class FunctionVersionRef(_OpenRemoteValue):
+    name: str
+    version: str
+
+
+class ApplicationInput(_OpenRemoteValue):
+    """One parameter value.
+
+    Slice 1 freezes integers, strings, booleans, nulls, arrays, and objects.
+    Floating-point literal encoding is deferred until Python authoring is
+    introduced with a language-neutral numeric representation.
+    """
+
+    parameter: str
+    kind: str
+    value: Any
+
+    @field_validator("value")
+    @classmethod
+    def _validate_value(cls, value):
+        return _validate_literal(value)
+
+
+class FunctionApplication(_OpenRemoteValue):
+    """Immutable pre-declaration application of an exact Function version."""
+
+    function: FunctionVersionRef
+    inputs: tuple[ApplicationInput, ...]
+    output: FunctionOutput
+    group_id: str
+    columns: Mapping[str, str] = Field(default_factory=dict)
+
+    def _known_dict(self) -> dict[str, Any]:
+        value = super()._known_dict()
+        for name in self._unknown_field_names():
+            value.pop(name, None)
+        return value
+
+    def _ensure_declarable(self) -> None:
+        unknown = {f"application.{name}" for name in self._unknown_field_names()}
+        unknown.update(
+            f"function.{name}" for name in self.function._unknown_field_names()
+        )
+        for index, input_value in enumerate(self.inputs):
+            unknown.update(
+                f"inputs[{index}].{name}" for name in input_value._unknown_field_names()
+            )
+        unknown.update(f"output.{name}" for name in self.output._unknown_field_names())
+        for index, field in enumerate(self.output.fields):
+            unknown.update(
+                f"output.fields[{index}].{name}"
+                for name in field._unknown_field_names()
+            )
+        if unknown:
+            raise ValueError(
+                "Function application contains fields from a newer contract: "
+                f"{sorted(unknown)!r}"
+            )
+
+    def rename(self, *, columns: Mapping[str, str]) -> FunctionApplication:
+        """Return a copy with result-field to table-column aliases."""
+        if self.output.kind != "named_struct":
+            raise ValueError("rename(columns=...) requires a named-struct application")
+        result_fields = {field.name for field in self.output.fields}
+        unknown = set(columns) - result_fields
+        if unknown:
+            raise ValueError(f"unknown Function result fields: {sorted(unknown)!r}")
+        merged = dict(self.columns)
+        merged.update(columns)
+        destinations = tuple(
+            merged.get(field.name, field.name) for field in self.output.fields
+        )
+        if len(set(destinations)) != len(destinations):
+            raise ValueError("FunctionApplication rename destinations must be unique")
+        return self._copy(update={"columns": merged})
+
+
+class InputBinding(_RemoteValue):
+    parameter: str
+    field_id: _Int32
+    field_path: str
+    arrow_type: str
+    nullable: bool
+
+
+class OutputMapping(_RemoteValue):
+    """One stable result-field mapping.
+
+    Assignment state is outside the Slice 1 client contract. During the NULL
+    transition Lance exposes no public cell-flag identifier to persist here.
+    """
+
+    result_field: str
+    output_name: str
+    output_field_id: _Int32
+    output_ordinal: _UInt32
+    arrow_type: str
+    nullable: bool
+
+
+class FunctionBinding(_RemoteValue):
+    """Immutable grouped binding persisted by the Enterprise table service."""
+
+    binding_id: str
+    revision: _UInt64
+    function: FunctionVersionRef
+    group_id: str
+    inputs: tuple[InputBinding, ...]
+    outputs: tuple[OutputMapping, ...]
+    input_schema: Optional[Mapping[str, Any]] = None
+    output_schema: Optional[Mapping[str, Any]] = None
+
+
+class RefreshColumnResult(_RemoteValue):
+    """Terminal result of a remote Function-column refresh Job."""
+
+    rows_assigned: _UInt64
+    rows_failed: _UInt64
+    rows_remaining: _UInt64
+    source_version: _UInt64
+    published_version: Optional[_UInt64] = None
+
+    @property
+    def rows_filled(self) -> int:
+        """Deprecated compatibility alias for :attr:`rows_assigned`."""
+        return self.rows_assigned
+
+    @property
+    def version(self) -> Optional[int]:
+        """Deprecated compatibility alias for :attr:`published_version`."""
+        return self.published_version
+
+
+__all__ = [
+    "ApplicationInput",
+    "FunctionApplication",
+    "FunctionArtifact",
+    "FunctionBinding",
+    "FunctionOutput",
+    "FunctionParameter",
+    "FunctionResultField",
+    "FunctionSignature",
+    "FunctionVersion",
+    "FunctionVersionRef",
+    "InputBinding",
+    "OutputMapping",
+    "PythonEnvironmentSpec",
+    "PythonRuntimeSpec",
+    "RefreshColumnResult",
+]

python/python/lancedb/namespace.py

@@ -49,6 +49,7 @@ from lancedb._lancedb import (
 )
 from lancedb.background_loop import LOOP
 from lancedb.db import AsyncConnection, DBConnection
+from lancedb.job import AsyncJob, Job
 from lance_namespace import (
     LanceNamespace,
     connect as namespace_connect,
@@ -624,6 +625,18 @@ class LanceNamespaceDBConnection(DBConnection):
             namespace_path = []
         LOOP.run(self._inner.drop_table(name, namespace_path=namespace_path))
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(
+            self._inner.drop_table_async(name, namespace_path=namespace_path)
+        )
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def rename_table(
         self,
@@ -1134,6 +1147,14 @@ class AsyncLanceNamespaceDBConnection:
             namespace_path = []
         await self._inner.drop_table(name, namespace_path=namespace_path)
 
+    async def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> AsyncJob:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        return await self._inner.drop_table_async(name, namespace_path=namespace_path)
+
     async def rename_table(
         self,
         cur_name: str,

python/python/lancedb/py.typed

@@ -0,0 +1 @@
+

python/python/lancedb/pydantic.py

@@ -1,7 +1,7 @@
 # SPDX-License-Identifier: Apache-2.0
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
-"""Pydantic (v1 / v2) adapter for LanceDB"""
+"""Pydantic adapter for LanceDB."""
 
 from __future__ import annotations
 
@@ -14,9 +14,6 @@ from enum import Enum
 from typing import (
     TYPE_CHECKING,
     Any,
-    Callable,
-    Dict,
-    Generator,
     List,
     Type,
     Union,
@@ -24,17 +21,9 @@ from typing import (
     GenericAlias,
 )
 
-import numpy as np
 import pyarrow as pa
 import pydantic
-from packaging.version import Version
-
-PYDANTIC_VERSION = Version(pydantic.__version__)
-try:
-    from pydantic_core import CoreSchema, core_schema
-except ImportError:
-    if PYDANTIC_VERSION.major >= 2:
-        raise
+from pydantic_core import CoreSchema, core_schema
 
 if TYPE_CHECKING:
     from pydantic.fields import FieldInfo
@@ -131,28 +120,18 @@ def Vector(
                 ),
             )
 
-        @classmethod
-        def __get_validators__(cls) -> Generator[Callable, None, None]:
-            yield cls.validate
-
-        # For pydantic v1
-        @classmethod
-        def validate(cls, v):
-            if not isinstance(v, (list, range, np.ndarray)) or len(v) != dim:
-                raise TypeError("A list of numbers or numpy.ndarray is needed")
-            return cls(v)
-
-        if PYDANTIC_VERSION.major < 2:
-
-            @classmethod
-            def __modify_schema__(cls, field_schema: Dict[str, Any]):
-                field_schema["items"] = {"type": "number"}
-                field_schema["maxItems"] = dim
-                field_schema["minItems"] = dim
-
     return FixedSizeList
 
 
+def _raise_bare_vector_error(*_args):
+    raise TypeError("Vector must be parameterized with a dimension, e.g. Vector(128).")
+
+
+# Pydantic otherwise inspects the bare factory as a field type and produces
+# misleading errors about its internal annotations.
+setattr(Vector, "__get_pydantic_core_schema__", _raise_bare_vector_error)
+
+
 def MultiVector(
     dim: int, value_type: pa.DataType = pa.float32(), nullable: bool = True
 ) -> Type:
@@ -223,31 +202,6 @@ def MultiVector(
                 ),
             )
 
-        @classmethod
-        def __get_validators__(cls) -> Generator[Callable, None, None]:
-            yield cls.validate
-
-        # For pydantic v1
-        @classmethod
-        def validate(cls, v):
-            if not isinstance(v, (list, range)):
-                raise TypeError("A list of vectors is needed")
-            for vec in v:
-                if not isinstance(vec, (list, range, np.ndarray)) or len(vec) != dim:
-                    raise TypeError(f"Each vector must be a list of {dim} numbers")
-            return cls(v)
-
-        if PYDANTIC_VERSION.major < 2:
-
-            @classmethod
-            def __modify_schema__(cls, field_schema: Dict[str, Any]):
-                field_schema["items"] = {
-                    "type": "array",
-                    "items": {"type": "number"},
-                    "minItems": dim,
-                    "maxItems": dim,
-                }
-
     return MultiVectorList
 
 
@@ -293,20 +247,10 @@ def _py_type_to_arrow_type(py_type: Type[Any], field: FieldInfo) -> pa.DataType:
     )
 
 
-if PYDANTIC_VERSION.major < 2:
-
-    def _pydantic_model_to_fields(model: pydantic.BaseModel) -> List[pa.Field]:
-        return [
-            _pydantic_to_field(name, field) for name, field in model.__fields__.items()
-        ]
-
-else:
-
-    def _pydantic_model_to_fields(model: pydantic.BaseModel) -> List[pa.Field]:
-        return [
-            _pydantic_to_field(name, field)
-            for name, field in model.model_fields.items()
-        ]
+def _pydantic_model_to_fields(model: pydantic.BaseModel) -> List[pa.Field]:
+    return [
+        _pydantic_to_field(name, field) for name, field in model.model_fields.items()
+    ]
 
 
 def _pydantic_type_to_arrow_type(tp: Any, field: FieldInfo) -> pa.DataType:
@@ -499,8 +443,6 @@ class LanceModel(pydantic.BaseModel):
 
     @classmethod
     def safe_get_fields(cls):
-        if PYDANTIC_VERSION.major < 2:
-            return cls.__fields__
         return cls.model_fields
 
     @classmethod
@@ -538,23 +480,9 @@ def get_extras(field_info: FieldInfo, key: str) -> Any:
     """
     Get the extra metadata from a Pydantic FieldInfo.
     """
-    if PYDANTIC_VERSION.major >= 2:
-        return (field_info.json_schema_extra or {}).get(key)
-    return (field_info.field_info.extra or {}).get("json_schema_extra", {}).get(key)
+    return (field_info.json_schema_extra or {}).get(key)
 
 
-if PYDANTIC_VERSION.major < 2:
-
-    def model_to_dict(model: pydantic.BaseModel) -> Dict[str, Any]:
-        """
-        Convert a Pydantic model to a dictionary.
-        """
-        return model.dict()
-
-else:
-
-    def model_to_dict(model: pydantic.BaseModel) -> Dict[str, Any]:
-        """
-        Convert a Pydantic model to a dictionary.
-        """
-        return model.model_dump()
+def model_to_dict(model: pydantic.BaseModel) -> dict[str, Any]:
+    """Convert a Pydantic model to a dictionary."""
+    return model.model_dump()

python/python/lancedb/query.py

@@ -32,8 +32,6 @@ from typing_extensions import Annotated
 
 from lancedb._lancedb import fts_query_to_json
 from lancedb.background_loop import LOOP
-from lancedb.pydantic import PYDANTIC_VERSION
-
 from . import __version__
 from .arrow import AsyncRecordBatchReader
 from .dependencies import pandas as pd
@@ -827,12 +825,7 @@ class Query(pydantic.BaseModel):
 
     # This tells pydantic to allow custom types (needed for the `vector` query since
     # pa.Array wouln't be allowed otherwise)
-    if PYDANTIC_VERSION.major < 2:  # Pydantic 1.x compat
-
-        class Config:
-            arbitrary_types_allowed = True
-    else:
-        model_config = {"arbitrary_types_allowed": True}
+    model_config = pydantic.ConfigDict(arbitrary_types_allowed=True)
 
 
 class LanceQueryBuilder(ABC):
@@ -2235,6 +2228,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             reranker=self._reranker,
             limit=self._limit,
             with_row_ids=True,
+            offset=self._offset,
         )
         return self._finish_hybrid_results(results)
 
@@ -2256,6 +2250,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
         reranker,
         limit: int,
         with_row_ids: bool,
+        offset: Optional[int] = None,
     ) -> pa.Table:
         if norm == "rank":
             vector_results = LanceHybridQueryBuilder._rank(vector_results, "_distance")
@@ -2332,7 +2327,7 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
             score_i = results.column_names.index("_score")
             results = results.set_column(score_i, "_score", original_scores)
 
-        results = results.slice(length=limit)
+        results = results.slice(offset=offset or 0, length=limit)
 
         if not with_row_ids:
             results = results.drop(["_rowid"])
@@ -2679,8 +2674,12 @@ class LanceHybridQueryBuilder(LanceQueryBuilder):
 
         # Apply common configurations
         if self._limit:
-            self._vector_query.limit(self._limit)
-            self._fts_query.limit(self._limit)
+            # The final offset/limit window is sliced out of the combined,
+            # reranked results, so each sub-query must fetch enough rows to
+            # cover the skipped prefix as well as the window itself.
+            sub_query_limit = self._limit + (self._offset or 0)
+            self._vector_query.limit(sub_query_limit)
+            self._fts_query.limit(sub_query_limit)
         if self._columns:
             self._vector_query.select(self._columns)
             self._fts_query.select(self._columns)
@@ -3245,12 +3244,7 @@ class AsyncStandardQuery(AsyncQueryBase):
         if ordering is None:
             self._inner.order_by(None)
         else:
-            self._inner.order_by(
-                [
-                    o.model_dump() if hasattr(o, "model_dump") else o.dict()
-                    for o in ordering
-                ]
-            )
+            self._inner.order_by([o.model_dump() for o in ordering])
         return self
 
     def fast_search(self) -> Self:

python/python/lancedb/remote/db.py

@@ -23,7 +23,7 @@ import pyarrow as pa
 
 from ..common import DATA
 from ..db import DBConnection, LOOP
-from ..job import Job
+from ..job import AsyncJob, Job
 
 if TYPE_CHECKING:
     from .._lancedb import JobDescription, JobInfo
@@ -663,6 +663,16 @@ class RemoteDBConnection(DBConnection):
             namespace_path = []
         LOOP.run(self._conn.drop_table(name, namespace_path=namespace_path))
 
+    @override
+    def drop_table_async(
+        self, name: str, namespace_path: Optional[List[str]] = None
+    ) -> Job:
+        """Start dropping a table and return its cleanup job."""
+        if namespace_path is None:
+            namespace_path = []
+        job = LOOP.run(self._conn.drop_table_async(name, namespace_path=namespace_path))
+        return Job(job if isinstance(job, AsyncJob) else AsyncJob(job))
+
     @override
     def rename_table(
         self,

python/python/lancedb/remote/table.py

@@ -49,6 +49,7 @@ from lancedb.index import (
     LabelList,
 )
 from lancedb.job import Job
+from lancedb.functions import FunctionApplication
 from lancedb.remote.db import LOOP
 from lancedb.table import IndexConfigType, KNOWN_METRICS
 import pyarrow as pa
@@ -958,8 +959,21 @@ class RemoteTable(Table):
     def count_rows(self, filter: Optional[str] = None) -> int:
         return LOOP.run(self._table.count_rows(filter))
 
-    def add_columns(self, transforms: Dict[str, str]) -> AddColumnsResult:
-        return LOOP.run(self._table.add_columns(transforms))
+    def add_columns(
+        self,
+        transforms: Dict[str, str | FunctionApplication]
+        | FunctionApplication
+        | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
+    ) -> AddColumnsResult:
+        return LOOP.run(self._table.add_columns(transforms, computed=computed))
+
+    def refresh_column(self, column: str):
+        return LOOP.run(self._table.refresh_column(column))
+
+    def refresh_column_async(self, column: str) -> Job:
+        return Job(LOOP.run(self._table.refresh_column_async(column)))
 
     def alter_columns(
         self, *alterations: Iterable[Dict[str, str]]
@@ -979,17 +993,39 @@ class RemoteTable(Table):
         return LOOP.run(self._table.set_unenforced_primary_key(columns))
 
     def set_lsm_write_spec(self, spec: "LsmWriteSpec") -> None:
-        """Not supported on LanceDB Cloud."""
+        """Install an LsmWriteSpec."""
         return LOOP.run(self._table.set_lsm_write_spec(spec))
 
     def unset_lsm_write_spec(self) -> None:
-        """Not supported on LanceDB Cloud."""
+        """Remove the LsmWriteSpec."""
         return LOOP.run(self._table.unset_lsm_write_spec())
 
     def get_lsm_write_spec(self) -> Optional["LsmWriteSpec"]:
         """Read the installed LsmWriteSpec, or ``None``."""
         return LOOP.run(self._table.get_lsm_write_spec())
 
+    def checkpoint_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.checkpoint_lsm`][lancedb.AsyncTable.checkpoint_lsm]."""
+        return LOOP.run(self._table.checkpoint_lsm())
+
+    def flush_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.flush_lsm`][lancedb.AsyncTable.flush_lsm]."""
+        return LOOP.run(self._table.flush_lsm())
+
+    def compact_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.compact_lsm`][lancedb.AsyncTable.compact_lsm]."""
+        return LOOP.run(self._table.compact_lsm())
+
+    def get_lsm_stats(self, *, include_generation_rows: bool = False) -> Optional[dict]:
+        """Synchronous version of
+        [`AsyncTable.get_lsm_stats`][lancedb.AsyncTable.get_lsm_stats]."""
+        return LOOP.run(
+            self._table.get_lsm_stats(include_generation_rows=include_generation_rows)
+        )
+
     def close_lsm_writers(self) -> None:
         """No-op on LanceDB Cloud (no local shard writers)."""
         return LOOP.run(self._table.close_lsm_writers())

python/python/lancedb/streaming.py

@@ -11,6 +11,11 @@ Provides StreamingDataset, a PyTorch IterableDataset that guarantees:
 - **Resumability**: state_dict / load_state_dict capture per-split consumption
   counts so training can resume from an exact mid-epoch position even when the
   distributed topology changes between runs.
+
+Transform failures on bad rows (e.g. nulls or NaNs from incomplete data) can
+be tolerated with ``on_transform_error="skip"``; see the parameter
+documentation on StreamingDataset for how this interacts with the guarantees
+above.
 """
 
 import ctypes
@@ -22,7 +27,7 @@ import time
 from collections import deque
 from concurrent.futures import ThreadPoolExecutor
 from multiprocessing import RawArray
-from typing import Any, Callable, Iterator, Optional
+from typing import Any, Callable, Iterator, Optional, Union
 
 from torch.utils.data import IterableDataset, get_worker_info
 
@@ -127,6 +132,49 @@ class StreamingDataset(IterableDataset):
         Maximum number of transforms to run concurrently.  Must be greater
         than zero.  When ``None`` (the default), uses ``os.cpu_count()`` or 1
         when the CPU count is unavailable.
+    on_transform_error:
+        What to do when the transform raises an exception:
+
+        - ``"raise"`` (the default): the exception propagates and iteration
+          aborts.
+        - ``"skip"``: the failing rows are dropped and iteration continues.
+        - ``"warn"``: like ``"skip"``, but a warning is logged for each
+          failing batch.
+        - a callable ``handler(exc) -> bool``: called with the exception;
+          return ``True`` to skip the failing rows or ``False`` to re-raise.
+          Useful to skip only expected error types (compatible with
+          ``webdataset.handlers`` style handlers).
+
+        When a batch fails, the transform is re-invoked on each single-row
+        slice of the batch so that only the rows that actually fail are
+        dropped.  Transforms should therefore be deterministic and accept
+        batches of any size (including one row).  Skipped rows are counted in
+        ``rows_skipped``.
+
+        Skipping weakens the elastic-determinism guarantee at the end of the
+        epoch: splits that lose more rows than others run dry earlier, and
+        each rank's iterator ends at the last cycle where every split *it
+        owns* still has a row.  Because bad rows are not distributed evenly
+        across splits, this means one rank's iterator can yield noticeably
+        fewer or more steps than another rank's *in the same run* — there is
+        no cross-rank coordination that stops every rank at the same global
+        step.  This is generally safe for asynchronous or single-rank use,
+        but synchronous distributed training (e.g. ranks that call
+        ``all_reduce`` every step) can hang or deadlock if one rank's
+        iterator is exhausted while others are still stepping; callers doing
+        synchronous multi-rank training with ``on_transform_error != "raise"``
+        are responsible for their own cross-rank stopping mechanism (e.g.
+        broadcasting a stop signal on ``StopIteration``).  The final few
+        global steps can also differ across topologies (bounded by the skew
+        in bad-row counts across splits).  The sequence of samples yielded
+        from each split remains deterministic.  Mid-epoch
+        checkpoints remain exact provided the transform fails
+        deterministically; in multi-rank training each rank must save its
+        own ``state_dict`` and the states must be combined with
+        ``merge_state_dicts`` before resuming on a different topology.
+        Prefer the ``filter`` parameter when bad rows can be expressed as a
+        SQL predicate (e.g. ``"col IS NOT NULL"``) — filtering happens before
+        splits are built, so every guarantee is fully preserved.
     worker_info_override:
         If set, used in place of ``torch.utils.data.get_worker_info()`` to
         determine the DataLoader worker assignment.  Intended for unit tests
@@ -152,6 +200,7 @@ class StreamingDataset(IterableDataset):
         filter: Optional[str] = None,
         transform: Optional[Callable] = None,
         transform_parallelism: Optional[int] = None,
+        on_transform_error: Union[str, Callable[[Exception], bool]] = "raise",
         connection_factory: Optional[Callable[[str], Any]] = None,
         worker_info_override=None,
     ):
@@ -167,6 +216,13 @@ class StreamingDataset(IterableDataset):
             )
         if transform_parallelism is not None and transform_parallelism <= 0:
             raise ValueError("transform_parallelism must be greater than 0")
+        if on_transform_error not in ("raise", "skip", "warn") and not callable(
+            on_transform_error
+        ):
+            raise ValueError(
+                "on_transform_error must be 'raise', 'skip', 'warn', or a "
+                f"callable, got {on_transform_error!r}"
+            )
 
         self._table = table
         self._num_splits = num_splits
@@ -182,6 +238,7 @@ class StreamingDataset(IterableDataset):
         self._filter = filter
         self._transform = transform
         self._transform_parallelism = transform_parallelism
+        self._on_transform_error = on_transform_error
         self._connection_factory = connection_factory
         self._worker_info_override = worker_info_override
 
@@ -199,19 +256,28 @@ class StreamingDataset(IterableDataset):
         # in the main process.  RawArray is picklable via the forkserver
         # reduction protocol so it survives the dataset pickle round-trip.
         # Layout: [unscanned_rows, raw_rows, cooked_rows, consumed_rows,
-        #          bytes_loaded, fetch_time_us, transform_time_us]
-        self._worker_stats: RawArray = RawArray(ctypes.c_int64, 7)
+        #          bytes_loaded, fetch_time_us, transform_time_us,
+        #          rows_skipped]
+        self._worker_stats: RawArray = RawArray(ctypes.c_int64, 8)
 
         # Cumulative bytes of Arrow buffer data fetched across all iterations.
         self._bytes_loaded: int = 0
         # Cumulative seconds spent in LanceDB I/O and in transform functions.
         self._fetch_time: float = 0.0
         self._transform_time: float = 0.0
+        # Cumulative rows dropped by on_transform_error across all iterations.
+        self._rows_skipped: int = 0
 
         # Number of samples each split has already been consumed.  At global
         # step boundaries all splits have consumed this many samples, so a
         # single scalar captures the topology-independent checkpoint state.
         self._resume_offset: int = 0
+        # Permutation position each split has consumed through, keyed by
+        # global split index.  Equal to _resume_offset for every split unless
+        # on_transform_error skipped rows, in which case skipped positions
+        # push the watermark of the affected splits further ahead.  Splits
+        # this instance has never iterated have no entry.
+        self._resume_positions: dict[int, int] = {}
 
         # Build the permutation table once, deterministically.
         builder = permutation_builder(table)
@@ -275,6 +341,7 @@ class StreamingDataset(IterableDataset):
         # Set identity transform on each Permutation so __getitems__ returns
         # the raw RecordBatch.  Stage 2 applies the real transform.
         permutations: list[Permutation] = []
+        initial_positions: list[int] = []
         for split_idx in my_splits:
             perm = Permutation.from_tables(
                 self._table, self._perm_table, split=split_idx
@@ -282,14 +349,20 @@ class StreamingDataset(IterableDataset):
             if self._columns is not None:
                 perm = perm.select_columns(self._columns)
             perm = perm.with_transform(lambda batch: batch)
-            if self._resume_offset > 0:
-                perm = perm.with_skip(self._resume_offset)
+            start_pos = self._resume_positions.get(split_idx, self._resume_offset)
+            if start_pos > 0:
+                perm = perm.with_skip(start_pos)
+            initial_positions.append(start_pos)
             permutations.append(perm)
 
         n = len(permutations)
         split_sizes = [perm.num_rows for perm in permutations]
         initial_offset = self._resume_offset
         local_consumed = [0] * n
+        # Permutation position each split has consumed through (absolute,
+        # i.e. counted from the start of the unskipped split).  Runs ahead of
+        # initial + local_consumed when rows are skipped.
+        pos_consumed = list(initial_positions)
 
         batch_size = self._read_batch_size
         max_prefetch = self._prefetch_batches
@@ -302,12 +375,14 @@ class StreamingDataset(IterableDataset):
             self._transform if self._transform is not None else Transforms.arrow2python
         )
 
-        # Per-split pipeline state.
+        # Per-split pipeline state.  Batches are paired with the absolute
+        # permutation position of their first row so that skipped rows can be
+        # accounted for in pos_consumed.
         fetch_head = [0] * n
-        io_pending = [deque() for _ in range(n)]  # Future[RecordBatch]
-        raw_batches = [deque() for _ in range(n)]  # RecordBatch — fetched, awaiting tx
-        tx_pending = [deque() for _ in range(n)]  # Future[list[Any]]
-        cooked = [deque() for _ in range(n)]  # rows ready to yield
+        io_pending = [deque() for _ in range(n)]  # (abs_start, Future[RecordBatch])
+        raw_batches = [deque() for _ in range(n)]  # (abs_start, RecordBatch)
+        tx_pending = [deque() for _ in range(n)]  # Future[list[(abs_pos, row)]]
+        cooked = [deque() for _ in range(n)]  # (abs_pos, row) ready to yield
 
         # Limit simultaneous transforms to transform_workers across all splits.
         tx_semaphore = threading.Semaphore(transform_workers)
@@ -330,7 +405,8 @@ class StreamingDataset(IterableDataset):
             fetch_head[i] += fetch
             perm_i = permutations[i]
             indices = list(range(start, start + fetch))
-            io_pending[i].append(io_pool.submit(_io_call, perm_i, indices))
+            abs_start = initial_positions[i] + start
+            io_pending[i].append((abs_start, io_pool.submit(_io_call, perm_i, indices)))
 
         def _fill_io(i: int) -> None:
             while len(io_pending[i]) < max_prefetch and fetch_head[i] < split_sizes[i]:
@@ -338,15 +414,72 @@ class StreamingDataset(IterableDataset):
 
         def _drain_io(i: int) -> None:
             """Move completed I/O futures into raw_batches non-blockingly."""
-            while io_pending[i] and io_pending[i][0].done():
-                raw_batches[i].append(io_pending[i].popleft().result())
+            while io_pending[i] and io_pending[i][0][1].done():
+                abs_start, fut = io_pending[i].popleft()
+                raw_batches[i].append((abs_start, fut.result()))
 
         # ── Stage 2 helpers ───────────────────────────────────────────────────
 
-        def _tx_call_guarded(batch):
+        on_error = self._on_transform_error
+
+        def _should_skip(exc: Exception) -> bool:
+            if on_error == "raise":
+                return False
+            if callable(on_error):
+                return bool(on_error(exc))
+            return True  # "skip" or "warn"
+
+        def _check_row_count(rows: list, num_rows: int) -> None:
+            if len(rows) != num_rows:
+                raise ValueError(
+                    f"transform returned {len(rows)} rows for a batch of "
+                    f"{num_rows}; transforms must return exactly one output "
+                    "row per input row.  To drop bad rows, raise inside the "
+                    "transform and pass on_transform_error='skip'."
+                )
+
+        def _transform_isolated(abs_start, batch, batch_exc):
+            """Re-run the transform on single-row slices, dropping failures."""
+            out = []
+            skipped = 0
+            first_exc = None
+            for j in range(batch.num_rows):
+                try:
+                    rows = list(final_transform(batch.slice(j, 1)))
+                except Exception as exc:
+                    if not _should_skip(exc):
+                        raise
+                    skipped += 1
+                    if first_exc is None:
+                        first_exc = exc
+                    continue
+                _check_row_count(rows, 1)
+                out.append((abs_start + j, rows[0]))
+            self._rows_skipped += skipped
+            if skipped and on_error == "warn":
+                logger.warning(
+                    "Skipped %d of %d rows whose transform failed (first error: %r)",
+                    skipped,
+                    batch.num_rows,
+                    first_exc if first_exc is not None else batch_exc,
+                )
+            return out
+
+        def _transform_batch(abs_start, batch):
+            """Apply the transform, returning [(abs_pos, row), ...]."""
+            try:
+                rows = list(final_transform(batch))
+            except Exception as exc:
+                if not _should_skip(exc):
+                    raise
+                return _transform_isolated(abs_start, batch, exc)
+            _check_row_count(rows, batch.num_rows)
+            return [(abs_start + j, row) for j, row in enumerate(rows)]
+
+        def _tx_call_guarded(abs_start, batch):
             try:
                 t0 = time.perf_counter()
-                result = final_transform(batch)
+                result = _transform_batch(abs_start, batch)
                 self._transform_time += time.perf_counter() - t0
                 return result
             finally:
@@ -355,8 +488,8 @@ class StreamingDataset(IterableDataset):
         def _try_submit_tx(i: int) -> None:
             """Submit transforms for raw_batches[i] up to available capacity."""
             while raw_batches[i] and tx_semaphore.acquire(blocking=False):
-                batch = raw_batches[i].popleft()
-                tx_pending[i].append(tx_pool.submit(_tx_call_guarded, batch))
+                abs_start, batch = raw_batches[i].popleft()
+                tx_pending[i].append(tx_pool.submit(_tx_call_guarded, abs_start, batch))
 
         def _drain_tx(i: int) -> None:
             """Move completed transform futures into cooked non-blockingly."""
@@ -384,11 +517,14 @@ class StreamingDataset(IterableDataset):
                     # Acquire a transform slot (may block briefly if all
                     # transform_workers are busy with other splits).
                     tx_semaphore.acquire()
-                    batch = raw_batches[i].popleft()
-                    tx_pending[i].append(tx_pool.submit(_tx_call_guarded, batch))
+                    abs_start, batch = raw_batches[i].popleft()
+                    tx_pending[i].append(
+                        tx_pool.submit(_tx_call_guarded, abs_start, batch)
+                    )
                 elif io_pending[i]:
                     # Block on the oldest in-flight I/O fetch.
-                    raw_batches[i].append(io_pending[i].popleft().result())
+                    abs_start, fut = io_pending[i].popleft()
+                    raw_batches[i].append((abs_start, fut.result()))
                     _advance(i)
                 else:
                     break  # split exhausted
@@ -407,15 +543,28 @@ class StreamingDataset(IterableDataset):
                         _fill_io(i)
 
                     while True:
-                        # Stop when any split is exhausted (all exhaust
-                        # simultaneously: equal split sizes + round-robin).
-                        if any(local_consumed[i] >= split_sizes[i] for i in range(n)):
+                        # A cycle only runs if every split can still produce a
+                        # row.  Without skips all splits exhaust simultaneously
+                        # (equal split sizes + round-robin); when
+                        # on_transform_error drops rows a split can run dry
+                        # early, ending the epoch at the last complete cycle.
+                        # This check only sees splits owned by this rank/worker
+                        # (my_splits) — there is no cross-rank coordination, so
+                        # a different rank with fewer skipped rows keeps going;
+                        # see the on_transform_error docstring.
+                        exhausted = False
+                        for i in range(n):
+                            _ensure_cooked(i)
+                            if not cooked[i]:
+                                exhausted = True
+                                break
+                        if exhausted:
                             break
 
                         for i in range(n):
-                            _ensure_cooked(i)
-                            row = cooked[i].popleft()
+                            pos, row = cooked[i].popleft()
                             local_consumed[i] += 1
+                            pos_consumed[i] = pos + 1
                             _advance(i)
 
                             # After the last split in each cycle: update the
@@ -424,21 +573,39 @@ class StreamingDataset(IterableDataset):
                             # even when __iter__ runs in a worker process.
                             if i == n - 1:
                                 self._resume_offset = initial_offset + local_consumed[i]
+                                for j, split_idx in enumerate(my_splits):
+                                    self._resume_positions[split_idx] = pos_consumed[j]
                                 ws = self._worker_stats
                                 ws[0] = sum(
                                     split_sizes[j] - fetch_head[j] for j in range(n)
                                 )
                                 ws[1] = sum(
-                                    batch.num_rows for q in raw_batches for batch in q
+                                    batch.num_rows
+                                    for q in raw_batches
+                                    for _, batch in q
                                 )
                                 ws[2] = sum(len(q) for q in cooked)
                                 ws[3] = sum(local_consumed)
                                 ws[4] = self._bytes_loaded
                                 ws[5] = int(self._fetch_time * 1_000_000)
                                 ws[6] = int(self._transform_time * 1_000_000)
+                                ws[7] = self._rows_skipped
 
                             yield row
                 finally:
+                    # Final stats flush: the per-cycle write above never runs
+                    # when iteration ends mid-cycle (e.g. a split whose rows
+                    # were all skipped before completing a single cycle), so
+                    # counters like rows_skipped would otherwise be stale.
+                    ws = self._worker_stats
+                    ws[0] = sum(split_sizes[j] - fetch_head[j] for j in range(n))
+                    ws[1] = 0  # queue-depth properties document 0 when idle
+                    ws[2] = 0
+                    ws[3] = sum(local_consumed)
+                    ws[4] = self._bytes_loaded
+                    ws[5] = int(self._fetch_time * 1_000_000)
+                    ws[6] = int(self._transform_time * 1_000_000)
+                    ws[7] = self._rows_skipped
                     self._raw_batches_ref = None
                     self._cooked_ref = None
                     self._fetch_head_ref = None
@@ -492,7 +659,7 @@ class StreamingDataset(IterableDataset):
         batches.  Returns 0 when not iterating.
         """
         if self._raw_batches_ref is not None:
-            return sum(batch.num_rows for q in self._raw_batches_ref for batch in q)
+            return sum(batch.num_rows for q in self._raw_batches_ref for _, batch in q)
         return int(self._worker_stats[1])
 
     @property
@@ -522,6 +689,19 @@ class StreamingDataset(IterableDataset):
             )
         return int(self._worker_stats[0])
 
+    @property
+    def rows_skipped(self) -> int:
+        """Number of rows dropped because their transform raised an exception.
+
+        Only ever non-zero when ``on_transform_error`` is set to ``"skip"``,
+        ``"warn"``, or a callable that returned ``True``.  Accumulates across
+        multiple iterations of the same dataset instance and is never reset
+        automatically.
+        """
+        if self._raw_batches_ref is not None:
+            return self._rows_skipped
+        return int(self._worker_stats[7])
+
     @property
     def consumed_rows(self) -> int:
         """Number of rows already yielded to the caller across all splits.
@@ -587,12 +767,27 @@ class StreamingDataset(IterableDataset):
         every split has been consumed the same number of times (by the
         round-robin design), so the per-split count is a single uniform value
         that is identical across all ranks and DataLoader workers.
+
+        ``positions_consumed_per_split`` records how far into each split's
+        permutation iteration has advanced.  It only differs from
+        ``samples_consumed_per_split`` when ``on_transform_error`` skipped
+        rows, in which case entries are exact for the splits this instance
+        iterated and a lower bound (the sample count) for splits owned by
+        other ranks or workers.  Combine the state dicts from all ranks with
+        [merge_state_dicts][lancedb.streaming.StreamingDataset.merge_state_dicts]
+        to recover the exact value for every split before resuming on a
+        different topology.
         """
+        positions = [
+            self._resume_positions.get(split, self._resume_offset)
+            for split in range(self._num_splits)
+        ]
         return {
             "shuffle_seed": self._shuffle_seed,
             "num_splits": self._num_splits,
             "epoch": self._epoch,
             "samples_consumed_per_split": [self._resume_offset] * self._num_splits,
+            "positions_consumed_per_split": positions,
         }
 
     def load_state_dict(self, state: dict) -> None:
@@ -618,3 +813,96 @@ class StreamingDataset(IterableDataset):
             self._resume_offset = consumed[0] if consumed else 0
         else:
             self._resume_offset = int(consumed)
+        # Older checkpoints predate positions_consumed_per_split; without
+        # skipped rows positions equal sample counts, so falling back to
+        # _resume_offset (the .get default in __iter__) is exact.
+        positions = state.get("positions_consumed_per_split")
+        if positions is None:
+            self._resume_positions = {}
+        else:
+            self._resume_positions = {
+                split: int(pos) for split, pos in enumerate(positions)
+            }
+
+    @staticmethod
+    def merge_state_dicts(states: list[dict]) -> dict:
+        """Merge state dicts saved by different ranks into one exact state.
+
+        Only needed when ``on_transform_error`` skips rows in multi-rank
+        training: each rank then knows the exact permutation position only for
+        its own splits, and records a lower bound for the rest.  Because
+        exactly one rank owns each split, the elementwise maximum across all
+        ranks' ``positions_consumed_per_split`` recovers the exact position of
+        every split.  Without skipped rows every rank's state is already
+        identical and merging is a no-op.
+
+        Raises ``ValueError`` if the states are empty or were not produced by
+        the same run (mismatched seed, split count, epoch, or sample counts).
+
+        The merge is always all-to-all and topology-agnostic: collect the
+        ``state_dict()`` from every rank of the *previous* run into one list,
+        merge that whole list, and hand the identical merged result to every
+        rank of the *next* run — regardless of whether the rank count grew,
+        shrank, or stayed the same. There is no pairwise or subset merging
+        step, because each split's exact position is only known to whichever
+        rank owned that split, and the elementwise maximum needs every rank's
+        contribution to be correct.
+
+        For example, checkpointing 8 ranks and resuming on 4 (the same
+        pattern applies when growing, e.g. 4 ranks resuming on 8)::
+
+            states = [ds.state_dict() for ds in previous_run_datasets]  # 8
+            merged = StreamingDataset.merge_state_dicts(states)
+            for ds in resumed_datasets:  # now only 4 ranks
+                ds.load_state_dict(merged)  # same dict on every rank
+
+        The rank count on either side never affects the merge itself, since
+        ``merge_state_dicts`` only cares about the list of states it is
+        given.  Each split's position is recovered by elementwise maximum;
+        here rank 0 owned split 0 (and skipped two rows there) while rank 1
+        owned split 1 (and skipped one row):
+
+        >>> rank0 = {
+        ...     "shuffle_seed": 0, "num_splits": 2, "epoch": 0,
+        ...     "samples_consumed_per_split": [3, 3],
+        ...     "positions_consumed_per_split": [5, 3],
+        ... }
+        >>> rank1 = {
+        ...     "shuffle_seed": 0, "num_splits": 2, "epoch": 0,
+        ...     "samples_consumed_per_split": [3, 3],
+        ...     "positions_consumed_per_split": [3, 4],
+        ... }
+        >>> merged = StreamingDataset.merge_state_dicts([rank0, rank1])
+        >>> merged["positions_consumed_per_split"]
+        [5, 4]
+        """
+        if not states:
+            raise ValueError("merge_state_dicts requires at least one state dict")
+        first = states[0]
+        for state in states[1:]:
+            for key in ("shuffle_seed", "num_splits", "epoch"):
+                if state[key] != first[key]:
+                    raise ValueError(
+                        f"{key} mismatch across state dicts: "
+                        f"{state[key]} != {first[key]}"
+                    )
+            if (
+                state["samples_consumed_per_split"]
+                != first["samples_consumed_per_split"]
+            ):
+                raise ValueError(
+                    "samples_consumed_per_split mismatch across state dicts; "
+                    "state_dict() must be called at the same global step "
+                    "boundary on every rank"
+                )
+        merged = dict(first)
+        all_positions = [
+            state.get(
+                "positions_consumed_per_split", state["samples_consumed_per_split"]
+            )
+            for state in states
+        ]
+        merged["positions_consumed_per_split"] = [
+            max(per_split) for per_split in zip(*all_positions)
+        ]
+        return merged

python/python/lancedb/table.py

@@ -72,6 +72,7 @@ from .index import (
     FTS,
 )
 from .expr import Expr
+from .functions import FunctionApplication
 from .merge import LanceMergeInsertBuilder
 from .pydantic import LanceModel, model_to_dict
 from .query import (
@@ -108,6 +109,11 @@ def _should_push_down_query_table(
     return namespace_client is not None and "QueryTable" in pushdown_operations
 
 
+def _polars_predicate_pushdown_barrier(frame: Any) -> Any:
+    """Return a Polars frame unchanged while blocking predicate pushdown."""
+    return frame
+
+
 _MODEL_BACKED_TOKENIZER_PREFIXES = ("jieba", "lindera")
 _MODEL_BACKED_TOKENIZER_ERRORS = (
     "unknown base tokenizer",
@@ -171,6 +177,7 @@ if TYPE_CHECKING:
         CompactionStats,
         Tag,
         AddColumnsResult,
+        RefreshColumnResult,
         AddResult,
         AlterColumnsResult,
         UpdateFieldMetadataResult,
@@ -427,6 +434,20 @@ def _cast_to_target_schema(
     return pa.RecordBatchReader.from_batches(reordered_schema, gen())
 
 
+def _field_extension_name(field: pa.Field) -> Optional[str]:
+    extension_name = getattr(field.type, "extension_name", None)
+    if extension_name is not None:
+        return extension_name
+
+    metadata = field.metadata or {}
+    extension_name = metadata.get(b"ARROW:extension:name") or metadata.get(
+        "ARROW:extension:name"
+    )
+    if isinstance(extension_name, bytes):
+        return extension_name.decode()
+    return extension_name
+
+
 def _align_field_types(
     fields: List[pa.Field],
     target_fields: List[pa.Field],
@@ -439,6 +460,16 @@ def _align_field_types(
         target_field = next((f for f in target_fields if f.name == field.name), None)
         if target_field is None:
             raise ValueError(f"Field '{field.name}' not found in target schema")
+        # Preserve arrow.json input until it reaches Lance. LanceDB exposes stored
+        # JSON columns as lance.json (JSONB-backed LargeBinary), but casting the
+        # input to that storage type here merely relabels the raw JSON bytes as
+        # JSONB. Lance must see arrow.json so it can perform the JSONB encoding.
+        if (
+            _field_extension_name(field) == "arrow.json"
+            and _field_extension_name(target_field) == "lance.json"
+        ):
+            new_fields.append(field)
+            continue
         if pa.types.is_struct(target_field.type):
             if pa.types.is_struct(field.type):
                 new_type = pa.struct(
@@ -864,12 +895,18 @@ class Table(ABC):
         """
         raise NotImplementedError
 
-    def to_polars(self, **kwargs) -> "pl.DataFrame":
-        """Return the table as a polars.DataFrame.
+    def to_polars(self, **kwargs) -> "pl.LazyFrame":
+        """Return the table as a Polars LazyFrame.
+
+        Note
+        ----
+        The Polars streaming engine is not supported because it does not currently
+        implement Python PyArrow dataset scans. Use the default engine when collecting
+        this LazyFrame.
 
         Returns
         -------
-        polars.DataFrame
+        polars.LazyFrame
         """
         raise NotImplementedError
 
@@ -1905,14 +1942,23 @@ class Table(ABC):
 
     @abstractmethod
     def add_columns(
-        self, transforms: Dict[str, str] | pa.Field | List[pa.Field] | pa.Schema
+        self,
+        transforms: Dict[str, str | FunctionApplication]
+        | FunctionApplication
+        | pa.Field
+        | List[pa.Field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
     ):
         """
         Add new columns with defined values.
 
         Parameters
         ----------
-        transforms: Dict[str, str], pa.Field, List[pa.Field], pa.Schema
+        transforms: Dict[str, str | FunctionApplication], FunctionApplication,
+            pa.Field, List[pa.Field], pa.Schema
             A map of column name to a SQL expression to use to calculate the
             value of the new column. These expressions will be evaluated for
             each row in the table, and can reference existing columns.
@@ -1920,10 +1966,101 @@ class Table(ABC):
             new columns with the specified data types. The new columns will
             be initialized with null values.
 
+            A mapping with one ``FunctionApplication`` value keeps its scalar
+            or named-struct result in the named table column. A bare
+            named-struct application expands its ordered result fields as one
+            atomic sibling group; aliases come from ``rename(columns=...)``.
+            Function columns are supported only on LanceDB Cloud and
+            Enterprise.
+        computed: Dict[str, str], optional
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression, so no
+            data type is supplied.
+
+            Unlike ``transforms``, the expression is stored rather than
+            evaluated now: the column is committed with no values, and rows get
+            them from [`refresh_column`][lancedb.table.Table.refresh_column].
+            Declaring one therefore costs the same on a large table as on an
+            empty one.
+
+            A refresh does not revisit rows it has already filled, so mutating
+            an input leaves the value computed at fill time; recomputing means
+            dropping the column and declaring it again. While a declaration
+            reads a column, that column cannot be renamed, retyped or dropped.
+
+            On LanceDB Cloud and Enterprise the expression is planned by the
+            server, and the refresh runs as a server job -- see
+            [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+            Cannot be combined with ``transforms``.
+
         Returns
         -------
         AddColumnsResult
             version: the new version number of the table after adding columns.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("computed_demo", [{"x": 1}, {"x": 2}])
+        >>> table.add_columns(computed={"doubled": "x * 2"})
+        AddColumnsResult(version=2)
+        >>> table.refresh_column("doubled")
+        RefreshColumnResult(rows_filled=2, version=3)
+        >>> table.to_arrow().sort_by("x").to_pandas()
+           x  doubled
+        0  1        2
+        1  2        4
+        """
+
+    @abstractmethod
+    def refresh_column(self, column: str) -> "RefreshColumnResult":
+        """
+        Fill the rows of a computed column that hold no value yet.
+
+        Declared with ``add_columns(computed=...)``, a column starts empty and
+        gets its values here. Rows appended since the last refresh are filled
+        by the next one; rows already filled are left as they are, so the call
+        is idempotent and does not observe a mutated input.
+
+        Local tables only: a remote refresh runs as a server job, through
+        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+
+        Parameters
+        ----------
+        column: str
+            The name of the computed column to fill.
+
+        Returns
+        -------
+        RefreshColumnResult
+            rows_filled: the number of rows given a value.
+            version: the new version number of the table.
+        """
+
+    @abstractmethod
+    def refresh_column_async(self, column: str) -> Job:
+        """
+        Like :meth:`refresh_column`, but returns a handle to the refresh job
+        instead of blocking until it completes.
+
+        The job may already be complete when returned; callers must not assume
+        the column is filled until :meth:`Job.wait` returns. Invalid input --
+        an unknown column, or one that is not computed -- raises here rather
+        than failing the job. On local tables the job runs in-process; on
+        LanceDB Cloud and Enterprise it is the server's backfill job.
+
+        Examples
+        --------
+        >>> import lancedb
+        >>> db = lancedb.connect("./.lancedb")
+        >>> table = db.create_table("computed_job_demo", [{"x": 1}, {"x": 2}])
+        >>> table.add_columns(computed={"doubled": "x * 2"})
+        AddColumnsResult(version=2)
+        >>> job = table.refresh_column_async("doubled")
+        >>> job.wait()
+        >>> job.status()
+        'finished'
         """
 
     @abstractmethod
@@ -2569,6 +2706,9 @@ class LanceTable(Table):
         2. Currently we've disabled push-down of the filters from polars
            because polars pushdown into pyarrow uses pyarrow compute
            expressions rather than SQl strings (which LanceDB supports)
+        3. The Polars streaming engine is not supported because it does not
+           currently implement Python PyArrow dataset scans. Use the default
+           engine when collecting this LazyFrame.
 
         Returns
         -------
@@ -2577,8 +2717,12 @@ class LanceTable(Table):
         from lancedb.integrations.pyarrow import PyarrowDatasetAdapter
 
         dataset = PyarrowDatasetAdapter(self)
-        return pl.scan_pyarrow_dataset(
-            dataset, allow_pyarrow_filter=False, batch_size=batch_size
+        # Polars 1.32's non-PyArrow callback path passes batch_size twice.  Keep
+        # the compatible PyArrow path, but block predicates because this adapter
+        # cannot translate PyArrow expressions into LanceDB filters.
+        return pl.scan_pyarrow_dataset(dataset, batch_size=batch_size).map_batches(
+            _polars_predicate_pushdown_barrier,
+            predicate_pushdown=False,
         )
 
     # New unified API overload
@@ -3921,9 +4065,29 @@ class LanceTable(Table):
         return LOOP.run(self._table.index_stats(index_name))
 
     def add_columns(
-        self, transforms: Dict[str, str] | pa.field | List[pa.field] | pa.Schema
+        self,
+        transforms: Dict[str, str | FunctionApplication]
+        | FunctionApplication
+        | pa.Field
+        | List[pa.Field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: Dict[str, str] | None = None,
     ) -> AddColumnsResult:
-        return LOOP.run(self._table.add_columns(transforms))
+        return LOOP.run(self._table.add_columns(transforms, computed=computed))
+
+    def refresh_column(self, column: str) -> "RefreshColumnResult":
+        """Fill a computed column's unfilled rows. See
+        [`AsyncTable.refresh_column`][lancedb.AsyncTable.refresh_column]."""
+        return LOOP.run(self._table.refresh_column(column))
+
+    def refresh_column_async(self, column: str) -> Job:
+        """Fill a computed column's unfilled rows, returning a handle to the
+        refresh job. See
+        [`Table.refresh_column_async`][lancedb.table.Table.refresh_column_async].
+        """
+        return Job(LOOP.run(self._table.refresh_column_async(column)))
 
     def alter_columns(
         self, *alterations: Iterable[Dict[str, str]]
@@ -3958,6 +4122,28 @@ class LanceTable(Table):
         [`AsyncTable.get_lsm_write_spec`][lancedb.AsyncTable.get_lsm_write_spec]."""
         return LOOP.run(self._table.get_lsm_write_spec())
 
+    def checkpoint_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.checkpoint_lsm`][lancedb.AsyncTable.checkpoint_lsm]."""
+        return LOOP.run(self._table.checkpoint_lsm())
+
+    def flush_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.flush_lsm`][lancedb.AsyncTable.flush_lsm]."""
+        return LOOP.run(self._table.flush_lsm())
+
+    def compact_lsm(self) -> None:
+        """Synchronous version of
+        [`AsyncTable.compact_lsm`][lancedb.AsyncTable.compact_lsm]."""
+        return LOOP.run(self._table.compact_lsm())
+
+    def get_lsm_stats(self, *, include_generation_rows: bool = False) -> Optional[dict]:
+        """Synchronous version of
+        [`AsyncTable.get_lsm_stats`][lancedb.AsyncTable.get_lsm_stats]."""
+        return LOOP.run(
+            self._table.get_lsm_stats(include_generation_rows=include_generation_rows)
+        )
+
     def close_lsm_writers(self) -> None:
         """Close cached MemWAL shard writers. See
         [`AsyncTable.close_lsm_writers`][lancedb.AsyncTable.close_lsm_writers]."""
@@ -4636,6 +4822,13 @@ class AsyncTable:
         via [`set_unenforced_primary_key`]; bucket sharding additionally
         requires it to be the single column being bucketed.
 
+        By default the MemWAL maintains every index on the table, resolved
+        here — a snapshot, so an index created afterwards needs the spec unset
+        and set again. This fails if one cannot be maintained; name the set
+        with ``with_maintained_indexes`` to install anyway. That pins an exact
+        set (a still-building index is rejected, not omitted); ``[]`` maintains
+        none.
+
         Parameters
         ----------
         spec : LsmWriteSpec
@@ -4643,7 +4836,7 @@ class AsyncTable:
 
         Examples
         --------
-        >>> from lancedb._lancedb import LsmWriteSpec
+        >>> from lancedb import LsmWriteSpec
         >>> # table.set_unenforced_primary_key("id")
         >>> # table.set_lsm_write_spec(LsmWriteSpec.bucket("id", 16))
         """
@@ -4662,12 +4855,73 @@ class AsyncTable:
 
         Returns ``None`` when the MemWAL LSM write path is not enabled (no
         spec has been set, or it was removed with `unset_lsm_write_spec`).
-        The returned spec — including its ``maintained_indexes`` and
-        ``writer_config_defaults`` — mirrors what was passed to
-        `set_lsm_write_spec`.
+        The returned spec mirrors what was passed to `set_lsm_write_spec`,
+        except that ``maintained_indexes`` always reports the concrete list
+        resolved when the spec was set — ``None`` never round-trips.
         """
         return await self._inner.get_lsm_write_spec()
 
+    async def checkpoint_lsm(self) -> None:
+        """Converge this table's LSM write path into its base table.
+
+        One flush, sealing every memtable into L0, then compaction triggers
+        until every generation that existed at that moment has reached base.
+        The loop runs client-side, reading progress from ``get_lsm_stats``.
+
+        Best-effort: generations created *while* it runs are deliberately not
+        waited on, which is what lets it terminate on a table taking writes.
+        Idempotent and safe on a cadence.
+
+        There is no deadline, and the caller owns that. It returns when the
+        target generations are gone, raises on a terminal server fault, and
+        otherwise waits however long the server takes. A slow table and a
+        stuck one are the same picture from the client: the compactor pool is
+        shared across every table on the node, so a checkpoint queued behind
+        unrelated work looks exactly like one that is merging. Wrap this in
+        ``asyncio.wait_for`` for a wall-clock bound; abandoning it partway
+        costs nothing.
+        """
+        await self._inner.checkpoint_lsm()
+
+    async def flush_lsm(self) -> None:
+        """Seal every bucket's active memtable into L0.
+
+        Does not touch the base table — moving L0 into base is
+        `compact_lsm`. On a node that has not claimed this table, this claims
+        it and replays its WAL log first.
+        """
+        await self._inner.flush_lsm()
+
+    async def compact_lsm(self) -> None:
+        """Trigger a background L0 to base compaction pass per bucket.
+
+        Returns once the passes are dispatched, not once they finish: watch
+        ``get_lsm_stats`` for progress, or use ``checkpoint_lsm`` to loop
+        until the current L0 has reached base.
+        """
+        await self._inner.compact_lsm()
+
+    async def get_lsm_stats(
+        self, *, include_generation_rows: bool = False
+    ) -> Optional[dict]:
+        """Read live per-bucket LSM state.
+
+        Answers "how far behind is my fresh tier", "which bucket is hot", and
+        "why is my fresh-tier vector search brute-force". Mutates no table
+        state, though on a node that has not claimed this table it claims it,
+        exactly as a read would.
+
+        Returns ``None`` only when the LSM write path is not enabled.
+
+        Parameters
+        ----------
+        include_generation_rows
+            Report a row count per L0 generation. Off by default: each count
+            opens an uncached Lance dataset, and ``checkpoint_lsm`` polls this
+            needing only generation numbers.
+        """
+        return await self._inner.get_lsm_stats(include_generation_rows)
+
     async def close_lsm_writers(self) -> None:
         """Drain and close any cached MemWAL shard writers for this table.
 
@@ -5748,37 +6002,154 @@ class AsyncTable:
         return await self._inner.update(updates_sql, where)
 
     async def add_columns(
-        self, transforms: dict[str, str] | pa.field | List[pa.field] | pa.Schema
+        self,
+        transforms: dict[str, str | FunctionApplication]
+        | FunctionApplication
+        | pa.Field
+        | List[pa.Field]
+        | pa.Schema
+        | None = None,
+        *,
+        computed: dict[str, str] | None = None,
     ) -> AddColumnsResult:
         """
         Add new columns with defined values.
 
         Parameters
         ----------
-        transforms: Dict[str, str]
+        transforms: Dict[str, str | FunctionApplication] or FunctionApplication
             A map of column name to a SQL expression to use to calculate the
             value of the new column. These expressions will be evaluated for
             each row in the table, and can reference existing columns.
             Alternatively, you can pass a pyarrow field or schema to add
             new columns with NULLs.
 
+            A mapping with one ``FunctionApplication`` value keeps its scalar
+            or named-struct result in the named table column. A bare
+            named-struct application expands its ordered result fields as one
+            atomic sibling group; aliases come from ``rename(columns=...)``.
+            Function columns are supported only on LanceDB Cloud and
+            Enterprise.
+        computed: Dict[str, str], optional
+            A map of column name to a SQL expression defining the column. The
+            column's type and inputs are derived from the expression.
+
+            Unlike ``transforms``, the expression is stored rather than
+            evaluated now: the column is committed with no values, and rows get
+            them from
+            [`refresh_column`][lancedb.table.AsyncTable.refresh_column].
+
+            A refresh does not revisit rows it has already filled, so mutating
+            an input leaves the value computed at fill time. While a
+            declaration reads a column, that column cannot be renamed, retyped
+            or dropped.
+
+            On LanceDB Cloud and Enterprise the expression is planned by
+            the server. Cannot be combined with ``transforms``.
+
         Returns
         -------
         AddColumnsResult
             version: the new version number of the table after adding columns.
 
         """
+        function_application = None
+        function_output_name = None
+        if isinstance(transforms, FunctionApplication):
+            function_application = transforms
+        elif isinstance(transforms, dict) and any(
+            isinstance(value, FunctionApplication) for value in transforms.values()
+        ):
+            if len(transforms) != 1 or not all(
+                isinstance(value, FunctionApplication) for value in transforms.values()
+            ):
+                raise ValueError(
+                    "one add_columns call declares exactly one Function sibling group"
+                )
+            function_output_name, function_application = next(iter(transforms.items()))
+
+        if function_application is not None:
+            if computed:
+                raise ValueError(
+                    "add_columns cannot mix a Function application with SQL "
+                    "computed columns"
+                )
+            function_application._ensure_declarable()
+            return await self._inner.add_function_columns(
+                function_application.to_canonical_json(), function_output_name
+            )
+
         if isinstance(transforms, pa.Field):
             transforms = [transforms]
         if isinstance(transforms, list) and all(
             {isinstance(f, pa.Field) for f in transforms}
         ):
             transforms = pa.schema(transforms)
+        if computed:
+            if transforms:
+                raise ValueError(
+                    "add_columns cannot take both transforms and computed columns"
+                )
+            return await self._inner.add_computed_columns(list(computed.items()))
+        if transforms is None:
+            raise ValueError("add_columns requires transforms or computed columns")
         if isinstance(transforms, pa.Schema):
             return await self._inner.add_columns_with_schema(transforms)
         else:
             return await self._inner.add_columns(list(transforms.items()))
 
+    async def refresh_column(self, column: str) -> RefreshColumnResult:
+        """
+        Fill the rows of a computed column that hold no value yet.
+
+        Declared with ``add_columns(computed=...)``, a column starts empty and
+        gets its values here. Rows appended since the last refresh are filled
+        by the next one; rows already filled are left as they are, so the call
+        is idempotent and does not observe a mutated input.
+
+        Local tables only: a remote refresh runs as a server job, through
+        [`refresh_column_async`][lancedb.table.Table.refresh_column_async].
+
+        Parameters
+        ----------
+        column: str
+            The name of the computed column to fill.
+
+        Returns
+        -------
+        RefreshColumnResult
+            The number of rows filled and the new version of the table.
+        """
+        return await self._inner.refresh_column(column)
+
+    async def refresh_column_async(self, column: str) -> AsyncJob:
+        """
+        Like :meth:`refresh_column`, but returns a handle to the refresh job
+        instead of blocking until it completes.
+
+        The job may already be complete when returned; callers must not assume
+        the column is filled until :meth:`AsyncJob.wait` resolves. Invalid
+        input -- an unknown column, or one that is not computed -- raises here
+        rather than failing the job. On local tables the job runs
+        in-process; on LanceDB Cloud and Enterprise it is the server's
+        backfill job.
+
+        Examples
+        --------
+        >>> import asyncio
+        >>> import lancedb
+        >>> async def refresh_in_background():
+        ...     db = await lancedb.connect_async("./.lancedb")
+        ...     table = await db.create_table("computed_job_async_demo", [{"x": 1}])
+        ...     await table.add_columns(computed={"doubled": "x * 2"})
+        ...     job = await table.refresh_column_async("doubled")
+        ...     await job.wait()
+        ...     return await job.status()
+        >>> asyncio.run(refresh_in_background())
+        'finished'
+        """
+        return AsyncJob(await self._inner.refresh_column_async(column))
+
     async def alter_columns(
         self, *alterations: Iterable[dict[str, Any]]
     ) -> AlterColumnsResult:
@@ -6233,7 +6604,9 @@ class TableStatistics:
     Attributes
     ----------
     total_bytes: int
-        The total number of bytes in the table.
+        The total size, in bytes, of the table's data files, index files, and
+        overlay files. Read from the manifest, so this excludes deletion files
+        and manifests.
     num_rows: int
         The total number of rows in the table.
     num_indices: int

python/python/lancedb/util.py

@@ -395,6 +395,11 @@ def _(value: dict):
     )
 
 
+@value_to_sql.register(pa.Scalar)
+def _(value: pa.Scalar):
+    return value_to_sql(value.as_py())
+
+
 @value_to_sql.register(np.ndarray)
 def _(value: np.ndarray):
     return value_to_sql(value.tolist())

python/python/tests/test_db.py

@@ -2,9 +2,11 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 
+import inspect
 import re
 import sys
 from datetime import timedelta
+from importlib import resources
 import os
 from types import SimpleNamespace
 
@@ -17,6 +19,10 @@ from lance_namespace.errors import NamespaceNotEmptyError, TableNotFoundError
 from lancedb.pydantic import LanceModel, Vector
 
 
+def test_package_includes_pep_561_marker():
+    assert resources.files(lancedb).joinpath("py.typed").is_file()
+
+
 def test_basic(tmp_path):
     db = lancedb.connect(tmp_path)
 
@@ -62,17 +68,23 @@ def test_basic(tmp_path):
     assert db.open_table("test").name == db["test"].name
 
 
-def test_sync_repr_does_not_use_background_loop(tmp_path, monkeypatch):
+def test_sync_debugger_inspection_does_not_use_background_loop(tmp_path, monkeypatch):
     from lancedb.background_loop import LOOP
 
     db = lancedb.connect(tmp_path)
     table = db.create_table("test", data=[{"id": 1}])
 
     def fail_run(*args, **kwargs):
-        raise AssertionError("repr should not use the Python background loop")
+        raise AssertionError("debugger inspection should not use the background loop")
 
     monkeypatch.setattr(LOOP, "run", fail_run)
 
+    # Debuggers enumerate and evaluate every exposed attribute when expanding a
+    # variable. This must remain safe while their breakpoint suspends LOOP's thread.
+    members = dict(inspect.getmembers(db))
+
+    assert members["uri"] == str(tmp_path)
+    assert members["read_consistency_interval"] is None
     assert repr(db) == f"LanceDBConnection(uri={str(tmp_path)!r})"
     assert repr(table) == f"LanceTable(name='test', _conn={db!r})"
 
@@ -743,8 +755,7 @@ def test_delete_table(tmp_db: lancedb.DBConnection):
     assert tmp_db.table_names() == []
 
 
-@pytest.mark.asyncio
-async def test_delete_table_async(tmp_db: lancedb.DBConnection):
+def test_drop_table_async(tmp_db: lancedb.DBConnection):
     data = pd.DataFrame(
         {
             "vector": [[3.1, 4.1], [5.9, 26.5]],
@@ -760,7 +771,10 @@ async def test_delete_table_async(tmp_db: lancedb.DBConnection):
 
     assert tmp_db.table_names() == ["test"]
 
-    tmp_db.drop_table("test")
+    job = tmp_db.drop_table_async("test")
+    assert job.id is None
+    assert job.status() == "finished"
+    job.wait()
     assert tmp_db.table_names() == []
 
     tmp_db.create_table("test", data=data)
@@ -769,6 +783,17 @@ async def test_delete_table_async(tmp_db: lancedb.DBConnection):
     tmp_db.drop_table("does_not_exist", ignore_missing=True)
 
 
+@pytest.mark.asyncio
+async def test_drop_table_async_connection(tmp_db_async: lancedb.AsyncConnection):
+    await tmp_db_async.create_table("test", data=pa.table({"id": [1, 2]}))
+
+    job = await tmp_db_async.drop_table_async("test")
+    assert job.id is None
+    assert await job.status() == "finished"
+    await job.wait()
+    assert await tmp_db_async.table_names() == []
+
+
 def test_drop_database(tmp_db: lancedb.DBConnection):
     data = pd.DataFrame(
         {

python/python/tests/test_elastic_dataloader.py

@@ -1456,6 +1456,408 @@ def test_shuffle_clump_size_yields_all_rows(lance_table):
     )
 
 
+# ---------------------------------------------------------------------------
+# on_transform_error tests
+# ---------------------------------------------------------------------------
+
+
+class BadRowError(ValueError):
+    """Raised by the failing transforms below when a batch contains a bad id."""
+
+
+def _failing_transform(bad_ids: set):
+    """A transform that raises BadRowError whenever the batch has a bad id.
+
+    Raises on the full batch and on any single-row slice containing a bad id,
+    so per-row isolation drops exactly the bad rows.
+    """
+
+    def transform(batch: pa.RecordBatch) -> list:
+        ids = batch.column("id").to_pylist()
+        bad = sorted(set(ids) & bad_ids)
+        if bad:
+            raise BadRowError(f"bad ids in batch: {bad}")
+        return [{"id": i} for i in ids]
+
+    return transform
+
+
+def _sequential_split_members(table) -> list[list[int]]:
+    """Return each split's ids in yield order for shuffle=False.
+
+    With a single rank and no workers the round-robin yields one row per split
+    per cycle, so item k of a clean run belongs to split k % NUM_SPLITS.
+    """
+    ds = StreamingDataset(table, num_splits=NUM_SPLITS, shuffle=False)
+    members: list[list[int]] = [[] for _ in range(NUM_SPLITS)]
+    for k, row in enumerate(ds):
+        members[k % NUM_SPLITS].append(row["id"])
+    return members
+
+
+def test_on_transform_error_default_raises(lance_table):
+    """By default a transform exception propagates and aborts iteration."""
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle_seed=SHUFFLE_SEED,
+        transform=_failing_transform({7}),
+    )
+    with pytest.raises(BadRowError):
+        list(ds)
+
+
+def test_on_transform_error_invalid_value(lance_table):
+    with pytest.raises(ValueError, match="on_transform_error"):
+        StreamingDataset(lance_table, num_splits=NUM_SPLITS, on_transform_error="bogus")
+
+
+def test_on_transform_error_skip_drops_bad_rows(lance_table):
+    """With one bad row per split, 'skip' yields every good row exactly once
+    and counts the dropped rows in rows_skipped."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][4] for i in range(NUM_SPLITS)}
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    assert ds.rows_skipped == 0
+
+    ids = [row["id"] for row in ds]
+
+    assert sorted(ids) == sorted(set(range(NUM_ROWS)) - bad_ids)
+    assert ds.rows_skipped == NUM_SPLITS
+
+
+def test_on_transform_error_skip_uneven_ends_at_last_complete_cycle(lance_table):
+    """When one split loses more rows than the others, the epoch ends at the
+    last cycle where every split still has a row — no crash, no bad rows, and
+    every step remains one sample per split."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = set(members[0][:3])  # all 3 bad rows in split 0
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    items = [row["id"] for row in ds]
+
+    rows_per_split = NUM_ROWS // NUM_SPLITS
+    expected_cycles = rows_per_split - len(bad_ids)
+    assert len(items) == expected_cycles * NUM_SPLITS
+    assert len(set(items)) == len(items), "duplicate samples yielded"
+    assert not set(items) & bad_ids, "a bad row was yielded"
+    # Split 0 contributed exactly its surviving rows, in order, one per cycle.
+    survivors = [i for i in members[0] if i not in bad_ids]
+    assert items[0::NUM_SPLITS] == survivors[:expected_cycles]
+
+
+def test_on_transform_error_warn_logs(lance_table, caplog):
+    """'warn' skips like 'skip' but logs a warning for the failing batch."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][3] for i in range(NUM_SPLITS)}
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="warn",
+    )
+    with caplog.at_level(logging.WARNING, logger="lancedb.streaming"):
+        items = list(ds)
+
+    assert len(items) == NUM_ROWS - NUM_SPLITS
+    assert ds.rows_skipped == NUM_SPLITS
+    assert "Skipped" in caplog.text
+    assert "BadRowError" in caplog.text
+
+
+def test_on_transform_error_callable_selective(lance_table):
+    """A callable handler can skip expected errors and re-raise the rest."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][0] for i in range(NUM_SPLITS)}
+
+    handled: list[Exception] = []
+
+    def handler(exc: Exception) -> bool:
+        handled.append(exc)
+        return isinstance(exc, BadRowError)
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error=handler,
+    )
+    items = list(ds)
+    assert len(items) == NUM_ROWS - NUM_SPLITS
+    assert handled and all(isinstance(exc, BadRowError) for exc in handled)
+
+    def broken_transform(batch: pa.RecordBatch) -> list:
+        raise TypeError("boom")
+
+    ds2 = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=broken_transform,
+        on_transform_error=handler,
+    )
+    with pytest.raises(TypeError, match="boom"):
+        list(ds2)
+
+
+def test_transform_wrong_row_count_raises(lance_table):
+    """A transform that returns the wrong number of rows is an error even with
+    on_transform_error='skip' — silent shrinkage would corrupt accounting."""
+
+    def drops_rows(batch: pa.RecordBatch) -> list:
+        return batch.column("id").to_pylist()[:-1]
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle_seed=SHUFFLE_SEED,
+        transform=drops_rows,
+        on_transform_error="skip",
+    )
+    with pytest.raises(ValueError, match="one output row per input row"):
+        list(ds)
+
+
+def test_skip_deterministic_across_runs(lance_table):
+    """With a fixed seed, skipping produces the identical sample sequence on
+    every run — skips are data-dependent, not run-dependent."""
+    bad_ids = {5, 17, 46}
+
+    def run() -> tuple[list[int], int]:
+        ds = StreamingDataset(
+            lance_table,
+            num_splits=NUM_SPLITS,
+            shuffle_seed=SHUFFLE_SEED,
+            transform=_failing_transform(bad_ids),
+            on_transform_error="skip",
+        )
+        return [row["id"] for row in ds], ds.rows_skipped
+
+    ids_a, skipped_a = run()
+    ids_b, skipped_b = run()
+    assert ids_a == ids_b
+    assert skipped_a == skipped_b
+    assert not set(ids_a) & bad_ids
+
+
+def test_skip_elastic_det_across_world_sizes(lance_table):
+    """With equal bad-row counts per split, skipping preserves the full
+    elastic-determinism guarantee: identical global batches at every step for
+    every compatible world_size."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = {members[i][6] for i in range(NUM_SPLITS)}
+
+    def collect(world_size: int) -> list[frozenset[int]]:
+        micro = GLOBAL_BATCH_SIZE // world_size
+        iters = [
+            iter(
+                StreamingDataset(
+                    lance_table,
+                    num_splits=NUM_SPLITS,
+                    shuffle=False,
+                    rank=rank,
+                    world_size=world_size,
+                    transform=_failing_transform(bad_ids),
+                    on_transform_error="skip",
+                )
+            )
+            for rank in range(world_size)
+        ]
+        _STOP = object()
+        batches: list[frozenset[int]] = []
+        while True:
+            step_samples: set[int] = set()
+            exhausted = 0
+            for it in iters:
+                for _ in range(micro):
+                    val = next(it, _STOP)
+                    if val is _STOP:
+                        exhausted += 1
+                        break
+                    step_samples.add(val["id"])
+            if exhausted == len(iters):
+                break
+            assert exhausted == 0, (
+                "Rank iterators exhausted at different steps despite equal "
+                "bad-row counts per split"
+            )
+            batches.append(frozenset(step_samples))
+        return batches
+
+    reference = collect(1)
+    assert len(reference) == NUM_ROWS // NUM_SPLITS - 1
+    for ws in (2, 3, 4):
+        assert collect(ws) == reference, f"world_size={ws} diverged"
+
+
+def test_resumability_with_skips_same_topology(lance_table):
+    """Checkpointing mid-epoch with skipped rows resumes exactly: no sample
+    repeated, no sample lost, skipped rows stay skipped."""
+    members = _sequential_split_members(lance_table)
+    # Uneven skips: positions diverge across splits (2 bad in split 0, 1 in
+    # split 5), which only a position-based checkpoint can resume exactly.
+    bad_ids = {members[0][2], members[0][3], members[5][7]}
+    kwargs = dict(
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+
+    reference = [row["id"] for row in StreamingDataset(lance_table, **kwargs)]
+    rows_per_split = NUM_ROWS // NUM_SPLITS
+    assert len(reference) == (rows_per_split - 2) * NUM_SPLITS
+
+    steps = 3
+    ds = StreamingDataset(lance_table, **kwargs)
+    it = iter(ds)
+    consumed = [next(it)["id"] for _ in range(steps * NUM_SPLITS)]
+    checkpoint = ds.state_dict()
+    it.close()
+
+    # Split 0 skipped positions 2 and 3 within its first 3 yields; split 5's
+    # bad row is beyond the checkpoint.  Everything else is at 3 = the sample
+    # count.
+    positions = checkpoint["positions_consumed_per_split"]
+    assert positions[0] == 5
+    assert positions[1:] == [3] * (NUM_SPLITS - 1)
+    assert checkpoint["samples_consumed_per_split"] == [3] * NUM_SPLITS
+
+    ds2 = StreamingDataset(lance_table, **kwargs)
+    ds2.load_state_dict(checkpoint)
+    resumed = [row["id"] for row in ds2]
+
+    assert consumed == reference[: steps * NUM_SPLITS]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
+def test_resumability_with_skips_elastic_merge(lance_table):
+    """Elastic resume with skips: each rank's checkpoint knows exact positions
+    only for its own splits; merge_state_dicts recovers the global state, and
+    a run on a different world_size continues exactly."""
+    members = _sequential_split_members(lance_table)
+    # Bad rows early in split 0 (rank 0) and split 6 (rank 1 of a ws=2 run) so
+    # both ranks' position vectors diverge before the checkpoint.
+    bad_ids = {members[0][0], members[0][2], members[6][1]}
+    kwargs = dict(
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+
+    reference = [row["id"] for row in StreamingDataset(lance_table, **kwargs)]
+
+    steps = 3
+    world_size = 2
+    micro = GLOBAL_BATCH_SIZE // world_size
+    datasets = [
+        StreamingDataset(lance_table, rank=rank, world_size=world_size, **kwargs)
+        for rank in range(world_size)
+    ]
+    iters = [iter(ds) for ds in datasets]
+    seen: list[frozenset[int]] = []
+    for _ in range(steps):
+        step_samples = set()
+        for it in iters:
+            for _ in range(micro):
+                step_samples.add(next(it)["id"])
+        seen.append(frozenset(step_samples))
+    states = [ds.state_dict() for ds in datasets]
+    for it in iters:
+        it.close()
+
+    merged = StreamingDataset.merge_state_dicts(states)
+    expected_positions = [3] * NUM_SPLITS
+    expected_positions[0] = 5  # skipped positions 0 and 2
+    expected_positions[6] = 4  # skipped position 1
+    assert merged["positions_consumed_per_split"] == expected_positions
+
+    # The first 3 global batches match the world_size=1 reference.
+    ref_batches = [
+        frozenset(reference[s * NUM_SPLITS : (s + 1) * NUM_SPLITS])
+        for s in range(len(reference) // NUM_SPLITS)
+    ]
+    assert seen == ref_batches[:steps]
+
+    # Resume on world_size=1 from the merged state.
+    ds_resume = StreamingDataset(lance_table, **kwargs)
+    ds_resume.load_state_dict(merged)
+    resumed = [row["id"] for row in ds_resume]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
+def test_rows_skipped_flushed_when_split_entirely_bad(lance_table):
+    """A split whose rows all fail never completes a cycle, so the epoch ends
+    immediately — but rows_skipped must still report the drops after the
+    iterator exits (the shared-memory counter is flushed on exhaustion)."""
+    members = _sequential_split_members(lance_table)
+    bad_ids = set(members[0])  # every row of split 0 is bad
+
+    ds = StreamingDataset(
+        lance_table,
+        num_splits=NUM_SPLITS,
+        shuffle=False,
+        transform=_failing_transform(bad_ids),
+        on_transform_error="skip",
+    )
+    assert list(ds) == []
+    assert ds.rows_skipped == len(bad_ids)
+
+
+def test_merge_state_dicts_validates_consistency(lance_table):
+    ds = StreamingDataset(lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED)
+    state = ds.state_dict()
+    other = dict(state, shuffle_seed=SHUFFLE_SEED + 1)
+    with pytest.raises(ValueError, match="shuffle_seed mismatch"):
+        StreamingDataset.merge_state_dicts([state, other])
+    with pytest.raises(ValueError, match="at least one"):
+        StreamingDataset.merge_state_dicts([])
+
+
+def test_load_state_dict_without_positions_key(lance_table):
+    """Checkpoints from before positions_consumed_per_split existed still
+    resume exactly (positions equal sample counts when nothing is skipped)."""
+    reference = [
+        row["id"]
+        for row in StreamingDataset(
+            lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED
+        )
+    ]
+
+    steps = 4
+    ds = StreamingDataset(lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED)
+    it = iter(ds)
+    for _ in range(steps * NUM_SPLITS):
+        next(it)
+    checkpoint = ds.state_dict()
+    it.close()
+    del checkpoint["positions_consumed_per_split"]
+
+    ds2 = StreamingDataset(
+        lance_table, num_splits=NUM_SPLITS, shuffle_seed=SHUFFLE_SEED
+    )
+    ds2.load_state_dict(checkpoint)
+    resumed = [row["id"] for row in ds2]
+    assert resumed == reference[steps * NUM_SPLITS :]
+
+
 def test_num_splits_defaults_to_world_size(lance_table):
     """Omitting num_splits gives world_size splits (one per rank)."""
     ds = StreamingDataset(

python/python/tests/test_embeddings.py

@@ -64,6 +64,23 @@ def test_embedding_function(tmp_path):
     assert np.allclose(actual, expected)
 
 
+def test_instructor_ndims_uses_instruction():
+    instructor = get_registry().get("instructor").create()
+    model = MagicMock()
+    model.encode.return_value = np.zeros((1, 384))
+
+    with patch.object(type(instructor), "get_model", return_value=model):
+        assert instructor.ndims() == 384
+
+    model.encode.assert_called_once_with(
+        [[instructor.source_instruction, "foo"]],
+        batch_size=instructor.batch_size,
+        show_progress_bar=instructor.show_progress_bar,
+        normalize_embeddings=instructor.normalize_embeddings,
+        device=instructor.device,
+    )
+
+
 def test_embedding_function_variables():
     @register("variable-testing")
     class VariableTestingFunction(TextEmbeddingFunction):
@@ -115,34 +132,16 @@ def test_embedding_function_variables():
     assert func.safe_model_dump()["secret_key"] == "$var:secret"
 
 
-def test_parse_functions_with_variables():
-    @register("variable-parsing-test")
-    class VariableParsingFunction(TextEmbeddingFunction):
-        api_key: str
-        base_url: Optional[str] = None
-
-        @staticmethod
-        def sensitive_keys():
-            return ["api_key"]
-
-        def ndims(self):
-            return 10
-
-        def generate_embeddings(self, texts):
-            # Mock implementation that just returns random embeddings
-            # In real usage, this would use the api_key to call an API
-            return [np.random.rand(self.ndims()).tolist() for _ in texts]
-
+def test_openai_variables_survive_metadata_round_trip():
     registry = EmbeddingFunctionRegistry.get_instance()
 
     registry.set_var("test_api_key", "sk-test-key-12345")
-    registry.set_var("test_base_url", "https://api.example.com")
 
     conf = EmbeddingFunctionConfig(
         source_column="text",
         vector_column="vector",
-        function=registry.get("variable-parsing-test").create(
-            api_key="$var:test_api_key", base_url="$var:test_base_url"
+        function=registry.get("openai").create(
+            api_key="$var:test_api_key", base_url="https://api.example.com"
         ),
     )
 
@@ -150,7 +149,10 @@ def test_parse_functions_with_variables():
 
     # Create a mock arrow table with the metadata
     schema = pa.schema(
-        [pa.field("text", pa.string()), pa.field("vector", pa.list_(pa.float32(), 10))]
+        [
+            pa.field("text", pa.string()),
+            pa.field("vector", pa.list_(pa.float32(), 1536)),
+        ]
     )
     table = pa.table({"text": [], "vector": []}, schema=schema)
     table = table.replace_schema_metadata(metadata)
@@ -164,13 +166,15 @@ def test_parse_functions_with_variables():
 
     assert parsed_func.api_key == "sk-test-key-12345"
     assert parsed_func.base_url == "https://api.example.com"
-
-    embeddings = parsed_func.generate_embeddings(["test text"])
-    assert len(embeddings) == 1
-    assert len(embeddings[0]) == 10
-
     assert parsed_func.safe_model_dump()["api_key"] == "$var:test_api_key"
 
+    with patch("lancedb.embeddings.openai.attempt_import_or_raise") as import_openai:
+        parsed_func._openai_client
+
+    import_openai.return_value.OpenAI.assert_called_once_with(
+        api_key="sk-test-key-12345", base_url="https://api.example.com"
+    )
+
 
 def test_embedding_with_bad_results(tmp_path):
     @register("null-embedding")
@@ -627,3 +631,23 @@ def test_url_retrieve_downloads_image():
     image_bytes = url_retrieve(image_url)
     img = Image.open(io.BytesIO(image_bytes))
     assert img.size[0] > 0 and img.size[1] > 0
+
+
+def test_jina_generate_image_input_dict_local_path(tmp_path):
+    """
+    JinaEmbeddings._generate_image_input_dict must accept a local image path
+    (str or Path), not just bytes. Previously it crashed with
+    `AttributeError: 'function' object has no attribute 'urlparse'` on any
+    str/Path input because it called `urlparse.urlparse(image)` instead of
+    `urlparse(image)` (urlparse was imported as a function, not a module).
+    """
+    Image = pytest.importorskip("PIL.Image")
+    from lancedb.embeddings.jinaai import JinaEmbeddings
+
+    image_path = tmp_path / "test.png"
+    Image.new("RGB", (4, 4), color="red").save(image_path, format="PNG")
+
+    for image in (str(image_path), image_path):
+        image_dict = JinaEmbeddings._generate_image_input_dict(image)
+        assert "image" in image_dict
+        assert isinstance(image_dict["image"], str) and len(image_dict["image"]) > 0

python/python/tests/test_expr.py

@@ -632,3 +632,101 @@ class TestExprBytesIntegration:
             .to_arrow()
         )
         assert result.num_rows == 2
+
+
+# ── datetime / timezone integration for lit() (issue #3262) ──────────────────
+
+
+class TestExprDatetimeTimezoneIntegration:
+    """Integration coverage for lit(datetime) against table timestamp columns.
+
+    PyArrow stores naive timestamps as UTC wall-clock microseconds. Python's
+    datetime.timestamp() treats naive values as *local* time, which used to
+    shift lit(naive) by the host UTC offset and break equality filters on
+    non-UTC machines. These cases lock the expected semantics.
+    """
+
+    def test_both_naive_match(self, tmp_path):
+        """Table naive + lit naive with the same wall clock must match."""
+        db = lancedb.connect(str(tmp_path / "naive"))
+        ts = datetime(2024, 7, 1, 10, 0, 0)
+        table = db.create_table(
+            "t", [{"id": 1, "ts": ts}, {"id": 2, "ts": datetime(2024, 7, 2, 10, 0, 0)}]
+        )
+        result = table.search().where(col("ts") == lit(ts)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_both_same_timezone_match(self, tmp_path):
+        """Table UTC + lit UTC for the same instant must match."""
+        db = lancedb.connect(str(tmp_path / "utc"))
+        ts = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1, 2],
+                    "ts": pa.array(
+                        [ts, datetime(2024, 7, 2, 10, 0, 0, tzinfo=timezone.utc)],
+                        type=pa.timestamp("us", tz="UTC"),
+                    ),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_different_timezones_same_instant(self, tmp_path):
+        """UTC table row equals lit of the same instant in a different zone."""
+        db = lancedb.connect(str(tmp_path / "diff_tz"))
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        # Same instant as 06:00 in UTC-4
+        ts_est = datetime(2024, 7, 1, 6, 0, 0, tzinfo=timezone(timedelta(hours=-4)))
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1],
+                    "ts": pa.array([ts_utc], type=pa.timestamp("us", tz="UTC")),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts_est)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_table_tz_literal_naive(self, tmp_path):
+        """UTC table + naive lit uses wall-clock equality (10:00 == 10:00 UTC)."""
+        db = lancedb.connect(str(tmp_path / "tz_naive"))
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        ts_naive = datetime(2024, 7, 1, 10, 0, 0)
+        table = db.create_table(
+            "t",
+            pa.table(
+                {
+                    "id": [1],
+                    "ts": pa.array([ts_utc], type=pa.timestamp("us", tz="UTC")),
+                }
+            ),
+        )
+        result = table.search().where(col("ts") == lit(ts_naive)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_table_naive_literal_aware(self, tmp_path):
+        """Naive table + UTC lit with the same wall clock must match."""
+        db = lancedb.connect(str(tmp_path / "naive_aware"))
+        ts_naive = datetime(2024, 7, 1, 10, 0, 0)
+        ts_utc = datetime(2024, 7, 1, 10, 0, 0, tzinfo=timezone.utc)
+        table = db.create_table("t", [{"id": 1, "ts": ts_naive}])
+        result = table.search().where(col("ts") == lit(ts_utc)).to_list()
+        assert len(result) == 1
+        assert result[0]["id"] == 1
+
+    def test_naive_lit_sql_is_wall_clock_not_local_shifted(self):
+        """Regression: naive lit must not apply the host local UTC offset."""
+        ts = datetime(2024, 7, 1, 10, 0, 0)
+        sql = lit(ts).to_sql()
+        # Must encode 10:00 wall clock, not 10:00+local_offset.
+        assert "2024-07-01 10:00:00" in sql

python/python/tests/test_first_class_function_slice1.py

@@ -0,0 +1,311 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import json
+from pathlib import Path
+
+import pytest
+
+import lancedb.functions as functions
+from lancedb.functions import (
+    FunctionApplication,
+    FunctionBinding,
+    FunctionVersion,
+    PythonRuntimeSpec,
+    RefreshColumnResult,
+)
+from lancedb.table import AsyncTable
+
+
+FIXTURES = (
+    Path(__file__).parents[3]
+    / "rust"
+    / "lancedb"
+    / "tests"
+    / "fixtures"
+    / "first_class_functions"
+    / "v1"
+)
+
+
+def fixture(name: str) -> str:
+    return (FIXTURES / name).read_text()
+
+
+def job_result(name: str) -> dict:
+    return json.loads(fixture(name))["result"]
+
+
+def assert_no_secret_values(value):
+    if isinstance(value, dict):
+        for key, child in value.items():
+            assert key not in {
+                "secret_value",
+                "secret_values",
+                "resolved_secret",
+                "resolved_secrets",
+            }
+            assert_no_secret_values(child)
+    elif isinstance(value, list):
+        for child in value:
+            assert_no_secret_values(child)
+
+
+def test_public_function_values_are_in_api_reference():
+    docs = Path(__file__).parents[3] / "docs" / "src" / "python" / "python.md"
+    rendered = docs.read_text()
+    for name in functions.__all__:
+        assert f"::: lancedb.functions.{name}" in rendered
+
+
+@pytest.mark.parametrize(
+    ("fixture_name", "canonical_name", "model", "nested_result"),
+    [
+        (
+            "remote_function_job.json",
+            "remote_function_version.canonical.json",
+            FunctionVersion,
+            True,
+        ),
+        (
+            "remote_function_application.json",
+            "remote_function_application.canonical.json",
+            FunctionApplication,
+            False,
+        ),
+        (
+            "remote_function_binding.json",
+            "remote_function_binding.canonical.json",
+            FunctionBinding,
+            False,
+        ),
+        (
+            "remote_refresh_job.json",
+            "remote_refresh_result.canonical.json",
+            RefreshColumnResult,
+            True,
+        ),
+        (
+            "remote_refresh_result_without_published_version.json",
+            "remote_refresh_result_without_published_version.canonical.json",
+            RefreshColumnResult,
+            False,
+        ),
+    ],
+)
+def test_python_and_rust_share_remote_canonical_goldens(
+    fixture_name, canonical_name, model, nested_result
+):
+    value = json.loads(fixture(fixture_name))
+    if nested_result:
+        value = value["result"]
+    decoded = model.from_json(json.dumps(value))
+    assert decoded.to_canonical_json() == fixture(canonical_name).strip()
+
+
+def test_function_version_identity_is_immutable_and_exact():
+    value = job_result("remote_function_job.json")
+    version = FunctionVersion.from_json(json.dumps(value))
+    assert version.name == "embed"
+    assert version.version == "fv_01K3EXACT"
+    assert version.required_secrets == ("HF_TOKEN",)
+
+    with pytest.raises((TypeError, ValueError)):
+        version.version = "fv_changed"
+    with pytest.raises(TypeError, match="immutable"):
+        version.runtime.env["TOKENIZERS_PARALLELISM"] = "true"
+
+    changed = dict(value)
+    changed["version"] = "fv_changed"
+    assert FunctionVersion(**changed) != version
+
+
+def test_unknown_fields_and_discriminators_are_forward_decodable():
+    value = job_result("remote_function_job.json")
+    value["future_version_metadata"] = {"retention_class": "catalog"}
+    value["runtime"] = {"kind": "wasm", "module_digest": "sha256:wasm"}
+    value["signature"]["output"]["kind"] = "future_output_shape"
+
+    version = FunctionVersion.from_json(json.dumps(value))
+    assert version.runtime.kind == "wasm"
+    assert version.runtime.python_version is None
+    assert version.runtime.environment is None
+    assert json.loads(version.to_canonical_json())["runtime"] == {"kind": "wasm"}
+    assert version.signature.output.kind == "future_output_shape"
+
+
+def test_function_application_uses_rename_columns_only():
+    application = FunctionApplication.from_json(
+        fixture("remote_function_application.json")
+    )
+    renamed = application.rename(columns={"normalized_text": "body_normalized"})
+
+    assert application.columns["normalized_text"] == "search_text"
+    assert renamed.columns["normalized_text"] == "body_normalized"
+    assert renamed.function == application.function
+    assert renamed.group_id == application.group_id
+    assert not hasattr(application, "rename_outputs")
+    with pytest.raises(TypeError, match="immutable"):
+        renamed.columns["normalized_text"] = "changed"
+    with pytest.raises(TypeError, match="immutable"):
+        application.inputs[0].value["path"] = "changed"
+
+    with pytest.raises(ValueError, match="unknown Function result fields"):
+        application.rename(columns={"missing": "search_text"})
+    with pytest.raises(ValueError, match="destinations must be unique"):
+        application.rename(columns={"normalized_text": "same", "token_count": "same"})
+
+    bare_value = json.loads(fixture("remote_function_application.json"))
+    bare_value.pop("columns")
+    bare = FunctionApplication(**bare_value)
+    with pytest.raises(ValueError, match="destinations must be unique"):
+        bare.rename(columns={"normalized_text": "token_count"})
+
+
+def test_binding_and_refresh_result_keep_stable_remote_fields():
+    binding = FunctionBinding.from_json(fixture("remote_function_binding.json"))
+    assert binding.revision == 3
+    assert binding.function.version == "fv_01K3TEXT"
+    assert [output.output_ordinal for output in binding.outputs] == [0, 1]
+    assert binding.input_schema is not None
+    assert binding.output_schema is not None
+
+    result = RefreshColumnResult.from_json(
+        json.dumps(job_result("remote_refresh_job.json"))
+    )
+    assert result.rows_filled == result.rows_assigned
+    assert result.version == result.published_version
+
+    result = RefreshColumnResult.from_json(
+        fixture("remote_refresh_result_without_published_version.json")
+    )
+    assert result.published_version is None
+    assert RefreshColumnResult.from_json(result.to_canonical_json()) == result
+
+
+def test_function_literal_numeric_domain_matches_rust():
+    with pytest.raises(ValueError, match="floating-point Function literals"):
+        FunctionApplication.from_json(fixture("remote_function_application_float.json"))
+
+    value = json.loads(fixture("remote_function_application_float.json"))
+    value["inputs"][0]["value"] = 2**64
+    with pytest.raises(ValueError, match="outside the canonical JSON range"):
+        FunctionApplication.from_json(json.dumps(value))
+
+
+def test_empty_default_maps_have_stable_canonical_bytes():
+    runtime = PythonRuntimeSpec(
+        kind="python", python_version="3.12", environment={"kind": "pip"}
+    )
+    assert runtime.to_canonical_json() == (
+        '{"environment":{"kind":"pip"},"kind":"python","python_version":"3.12"}'
+    )
+
+    value = json.loads(fixture("remote_function_application.json"))
+    value.pop("columns")
+    application = FunctionApplication.from_json(json.dumps(value))
+    assert "columns" not in json.loads(application.to_canonical_json())
+
+
+@pytest.mark.parametrize("field", ["rows_assigned", "source_version"])
+def test_refresh_result_rejects_non_u64_values(field):
+    value = job_result("remote_refresh_job.json")
+    value[field] = -1
+    with pytest.raises(ValueError):
+        RefreshColumnResult.from_json(json.dumps(value))
+
+    value[field] = "1"
+    with pytest.raises(ValueError):
+        RefreshColumnResult.from_json(json.dumps(value))
+
+
+def test_canonical_client_values_contain_secret_names_only():
+    version = FunctionVersion.from_json(
+        json.dumps(job_result("remote_function_job.json"))
+    )
+    canonical = json.loads(version.to_canonical_json())
+    assert canonical["required_secrets"] == ["HF_TOKEN"]
+    assert_no_secret_values(canonical)
+
+
+class _FunctionDeclarationInner:
+    def __init__(self):
+        self.calls = []
+
+    async def add_function_columns(self, application_json, output_name):
+        self.calls.append((json.loads(application_json), output_name))
+        return "declared"
+
+
+def known_application() -> FunctionApplication:
+    value = json.loads(fixture("remote_function_application.json"))
+    value.pop("future_application")
+    return FunctionApplication(**value)
+
+
+@pytest.mark.asyncio
+async def test_add_columns_routes_struct_as_one_and_grouped_expansion_atomically():
+    inner = _FunctionDeclarationInner()
+    table = AsyncTable(inner)
+    application = known_application()
+
+    result = await table.add_columns(
+        {"features": application._copy(update={"columns": {}})}
+    )
+    assert result == "declared"
+    assert inner.calls[-1][1] == "features"
+
+    bare = application._copy(update={"columns": {}}).rename(
+        columns={"normalized_text": "search_text"}
+    )
+    result = await table.add_columns(bare)
+    assert result == "declared"
+    assert inner.calls[-1][1] is None
+    assert inner.calls[-1][0]["columns"] == {"normalized_text": "search_text"}
+
+
+@pytest.mark.asyncio
+async def test_add_columns_rejects_mixed_groups_and_unknown_newer_application():
+    inner = _FunctionDeclarationInner()
+    table = AsyncTable(inner)
+    application = known_application()
+
+    with pytest.raises(ValueError, match="exactly one Function sibling group"):
+        await table.add_columns({"a": application, "b": application})
+
+    future = json.loads(fixture("remote_function_application.json"))
+    application = FunctionApplication(**future)
+    with pytest.raises(ValueError, match="newer contract"):
+        await table.add_columns(application)
+
+    future.pop("future_application")
+    future["output"]["assignment"] = "cell_flag"
+    application = FunctionApplication(**future)
+    assert "assignment" not in json.loads(application.to_canonical_json())["output"]
+    with pytest.raises(ValueError, match="output.assignment"):
+        await table.add_columns(application)
+    assert inner.calls == []
+
+
+def test_rename_requires_named_struct_and_keeps_partial_mapping_immutable():
+    scalar = FunctionApplication.from_json(
+        json.dumps(
+            {
+                "function": {"name": "embed", "version": "fv_exact"},
+                "inputs": [],
+                "output": {
+                    "kind": "scalar",
+                    "arrow_type": "list<float32>",
+                    "nullable": False,
+                },
+                "group_id": "fg_scalar",
+            }
+        )
+    )
+    with pytest.raises(ValueError, match="named-struct"):
+        scalar.rename(columns={"value": "embedding"})
+
+    application = known_application()._copy(update={"columns": {}})
+    renamed = application.rename(columns={"normalized_text": "search_text"})
+    assert dict(application.columns) == {}
+    assert dict(renamed.columns) == {"normalized_text": "search_text"}

python/python/tests/test_hybrid_query.py

@@ -12,7 +12,7 @@ import pyarrow.compute as pc
 import pytest
 import pytest_asyncio
 
-from lancedb.index import FTS
+from lancedb.index import BTree, FTS, IvfPq
 from lancedb.table import AsyncTable, Table
 
 
@@ -99,6 +99,86 @@ async def test_async_hybrid_query_filters(table: AsyncTable):
     assert result["text"].to_pylist() == ["cat", "b"]
 
 
+@pytest.mark.asyncio
+async def test_hybrid_query_with_stale_fixed_size_binary_prefilter(
+    tmpdir_factory,
+):
+    tmp_path = str(tmpdir_factory.mktemp("stale_scalar_prefilter"))
+    db = await lancedb.connect_async(tmp_path)
+
+    def fixed_size_binary(value: int) -> bytes:
+        return value.to_bytes(16, byteorder="big")
+
+    num_rows = 1000
+    data = pa.table(
+        {
+            "space_id": pa.array(
+                [fixed_size_binary(i) for i in range(num_rows)],
+                type=pa.binary(16),
+            ),
+            "text": ["book"] * num_rows,
+            "vector": pa.array(
+                [[float(i), float(i)] for i in range(num_rows)],
+                type=pa.list_(pa.float32(), 2),
+            ),
+        }
+    )
+    table = await db.create_table("test", data)
+    await table.create_index(
+        "vector", config=IvfPq(num_partitions=4, num_sub_vectors=2)
+    )
+    await table.create_index("space_id", config=BTree())
+    await table.create_index("text", config=FTS(with_position=False))
+
+    # Advance the search indices without advancing the scalar index. This is the
+    # state that previously let hybrid search use an incomplete scalar prefilter.
+    await table.add(data)
+    lance_dataset = await table.to_lance()
+    lance_dataset.optimize.optimize_indices(index_names=["vector_idx", "text_idx"])
+    await table.checkout_latest()
+
+    scalar_stats = await table.index_stats("space_id_idx")
+    assert scalar_stats is not None
+    assert scalar_stats.num_indexed_rows == num_rows
+    assert scalar_stats.num_unindexed_rows == num_rows
+
+    for index_name in ["vector_idx", "text_idx"]:
+        search_stats = await table.index_stats(index_name)
+        assert search_stats is not None
+        assert search_stats.num_indexed_rows == num_rows * 2
+        assert search_stats.num_unindexed_rows == 0
+
+    matching_ids = [5, 10, 15, 20, 25, 30]
+    literals = [
+        f"arrow_cast(0x{fixed_size_binary(i).hex()}, 'FixedSizeBinary(16)')"
+        for i in matching_ids
+    ]
+    predicate = f"space_id IN ({', '.join(literals)})"
+    expected_ids = sorted(fixed_size_binary(i) for i in matching_ids for _ in range(2))
+
+    vector_query = (
+        table.query().where(predicate).nearest_to([5.0, 5.0]).limit(num_rows * 2)
+    )
+    vector_results = await vector_query.to_arrow()
+    assert sorted(vector_results["space_id"].to_pylist()) == expected_ids
+
+    fts_query = (
+        table.query().where(predicate).nearest_to_text("book").limit(num_rows * 2)
+    )
+    fts_results = await fts_query.to_arrow()
+    assert sorted(fts_results["space_id"].to_pylist()) == expected_ids
+
+    hybrid_results = await (
+        table.query()
+        .where(predicate)
+        .nearest_to([5.0, 5.0])
+        .nearest_to_text("book")
+        .limit(num_rows * 2)
+        .to_arrow()
+    )
+    assert sorted(hybrid_results["space_id"].to_pylist()) == expected_ids
+
+
 @pytest.mark.asyncio
 async def test_async_hybrid_query_default_limit(table: AsyncTable):
     # add 10 new rows
@@ -123,6 +203,31 @@ async def test_async_hybrid_query_default_limit(table: AsyncTable):
     assert texts.count("a") == 1
 
 
+def test_hybrid_query_offset(sync_table: Table):
+    # The offset window of a hybrid query must be a suffix of the same query
+    # run without an offset -- it must not be silently ignored.
+    full = (
+        sync_table.search(query_type="hybrid")
+        .vector([0.0, 0.4])
+        .text("dog")
+        .limit(4)
+        .with_row_id(True)
+        .to_arrow()
+    )
+    assert len(full) == 4
+
+    offset_result = (
+        sync_table.search(query_type="hybrid")
+        .vector([0.0, 0.4])
+        .text("dog")
+        .offset(2)
+        .limit(2)
+        .with_row_id(True)
+        .to_arrow()
+    )
+    assert offset_result["_rowid"].to_pylist() == full["_rowid"].to_pylist()[2:]
+
+
 def test_hybrid_query_minimum_nprobes_zero_raises(sync_table: Table):
     # minimum_nprobes(0) must raise the same validation error a plain vector
     # query raises, not silently no-op because 0 is falsy.

python/python/tests/test_import.py

@@ -0,0 +1,33 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import re
+import shutil
+import subprocess
+import sys
+
+import lancedb._lancedb as _lancedb
+import pytest
+
+
+@pytest.mark.skipif(sys.platform != "linux", reason="ldd is Linux-specific")
+def test_native_extension_does_not_link_openssl():
+    """OpenSSL-linked wheels abort when imported on RHEL hosts in FIPS mode."""
+    ldd = shutil.which("ldd")
+    if ldd is None:
+        pytest.skip("ldd is not installed")
+
+    result = subprocess.run(
+        [ldd, _lancedb.__file__],
+        check=True,
+        capture_output=True,
+        text=True,
+    )
+    openssl_libraries = re.findall(
+        r"^\s*(lib(?:crypto|ssl)\S*)\s+=>", result.stdout, flags=re.MULTILINE
+    )
+
+    assert not openssl_libraries, (
+        "the LanceDB native extension must use rustls instead of linking OpenSSL: "
+        f"{openssl_libraries}"
+    )

python/python/tests/test_index.py

@@ -372,6 +372,31 @@ async def test_create_vector_index(some_table: AsyncTable):
     assert stats.num_indices == 1
 
 
+@pytest.mark.asyncio
+async def test_create_ivf_index_reports_unsplittable_partitions(db_async):
+    dim = 8
+    num_partitions = 300  # More than 256 selects hierarchical k-means.
+    base_vectors = [[float(row == column) for column in range(dim)] for row in range(5)]
+    vectors = pa.array(base_vectors * 200, pa.list_(pa.float32(), dim))
+    table = await db_async.create_table(
+        "unsplittable_partitions",
+        pa.table({"vector": vectors}),
+    )
+
+    error_pattern = (
+        rf"Cannot create {num_partitions} IVF partitions: k-means could only form"
+    )
+    with pytest.raises(RuntimeError, match=error_pattern):
+        await table.create_index(
+            "vector",
+            config=IvfFlat(
+                distance_type="dot",
+                num_partitions=num_partitions,
+                max_iterations=10,
+            ),
+        )
+
+
 @pytest.mark.asyncio
 async def test_create_4bit_ivfpq_index(some_table: AsyncTable):
     # Can create

python/python/tests/test_lsm_write_spec.py

@@ -21,6 +21,11 @@ SCHEMA = pa.schema(
 )
 
 
+def test_lsm_write_spec_module_metadata():
+    assert lancedb.LsmWriteSpec is LsmWriteSpec
+    assert LsmWriteSpec.__module__ == "lancedb._lancedb"
+
+
 def _batch(ids, vs):
     return pa.RecordBatch.from_arrays(
         [pa.array(ids, type=pa.utf8()), pa.array(vs, type=pa.int32())],
@@ -83,7 +88,9 @@ def test_lsm_write_spec_repr():
     assert s.spec_type == "bucket"
     assert s.column == "id"
     assert s.num_buckets == 4
-    assert s.maintained_indexes == []
+    # A fresh spec defers its maintained set to install time.
+    assert s.maintained_indexes is None
+    assert s.with_maintained_indexes([]).maintained_indexes == []
     assert "bucket" in repr(s)
     assert "id" in repr(s)
     assert "4" in repr(s)
@@ -169,18 +176,23 @@ def test_get_lsm_write_spec(tmp_path):
     table.unset_lsm_write_spec()
     assert table.get_lsm_write_spec() is None
 
-    # Identity round-trips (column recovered from the schema).
+    # Identity round-trips (column recovered from the schema). Leaving the
+    # maintained set to be inferred picks up the index on the table, so the
+    # spec reads back naming it rather than as "infer".
     table.set_lsm_write_spec(LsmWriteSpec.identity("id"))
     spec = table.get_lsm_write_spec()
     assert spec.spec_type == "identity"
     assert spec.column == "id"
+    assert spec.maintained_indexes == [idx_name]
     table.unset_lsm_write_spec()
 
-    # Unsharded round-trips (no routing column).
-    table.set_lsm_write_spec(LsmWriteSpec.unsharded())
+    # Unsharded round-trips (no routing column). Opting out is distinct from
+    # the inferred default.
+    table.set_lsm_write_spec(LsmWriteSpec.unsharded().with_maintained_indexes([]))
     spec = table.get_lsm_write_spec()
     assert spec.spec_type == "unsharded"
     assert spec.column is None
+    assert spec.maintained_indexes == []
 
 
 @pytest.mark.asyncio

python/python/tests/test_merge_insert_lsm.py

@@ -544,7 +544,7 @@ def test_lsm_read_fts_unmaintained_index_errors(tmp_path):
     table.create_index("text", config=FTS())
     # No maintained indexes: the active memtable FTS arm cannot serve un-compacted
     # docs, so the search would silently omit them — reject instead.
-    table.set_lsm_write_spec(LsmWriteSpec.unsharded())
+    table.set_lsm_write_spec(LsmWriteSpec.unsharded().with_maintained_indexes([]))
     with pytest.raises(Exception, match="maintained"):
         table.search("fox", query_type="fts", fts_columns="text").to_arrow()
 
@@ -631,7 +631,7 @@ def test_lsm_read_vector_unmaintained_index_errors(tmp_path):
     )
     # Spec with NO maintained indexes: the base vector index's catch-up is untracked,
     # so the scanner rejects rather than risk dropping compacted-but-unindexed rows.
-    table.set_lsm_write_spec(LsmWriteSpec.unsharded())
+    table.set_lsm_write_spec(LsmWriteSpec.unsharded().with_maintained_indexes([]))
     with pytest.raises(Exception, match="maintained"):
         table.search([1.0] * VECTOR_DIM).to_arrow()
 

python/python/tests/test_package_metadata.py

@@ -0,0 +1,42 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+import importlib
+import re
+import sys
+from pathlib import Path
+
+import pytest
+
+
+def test_pyo3_abi_matches_minimum_supported_python():
+    project_dir = Path(__file__).parents[2]
+    pyproject = (project_dir / "pyproject.toml").read_text()
+    cargo_manifest = (project_dir / "Cargo.toml").read_text()
+
+    minimum_python = re.search(
+        r'^requires-python\s*=\s*">=(\d+)\.(\d+)"$', pyproject, re.MULTILINE
+    )
+    assert minimum_python is not None
+
+    major, minor = minimum_python.groups()
+    expected_abi = f"abi3-py{major}{minor}"
+    configured_abis = re.findall(r'"(abi3-py\d+)"', cargo_manifest)
+
+    assert configured_abis == [expected_abi, expected_abi], (
+        "the pyo3 runtime and build ABI features must both match requires-python"
+    )
+
+
+@pytest.mark.skipif(sys.platform != "win32", reason="Windows wheel regression test")
+def test_windows_wheel_tag_and_native_import():
+    project_dir = Path(__file__).parents[2]
+    wheels = list((project_dir.parent / "target" / "wheels").glob("lancedb-*.whl"))
+    if not wheels:
+        pytest.skip("no wheel artifact is available in this development environment")
+
+    assert len(wheels) == 1
+    assert wheels[0].name.endswith("-cp310-abi3-win_amd64.whl")
+
+    native_module = importlib.import_module("lancedb._lancedb")
+    assert Path(native_module.__file__).suffix == ".pyd"

python/python/tests/test_pydantic.py

@@ -10,11 +10,10 @@ import pyarrow as pa
 import pydantic
 import pytest
 from lancedb.pydantic import (
-    PYDANTIC_VERSION,
     LanceModel,
+    MultiVector,
     Vector,
     pydantic_to_schema,
-    MultiVector,
 )
 from pydantic import BaseModel
 from pydantic import Field
@@ -415,22 +414,27 @@ def test_nullable_vector():
     assert schema == pa.schema([pa.field("vec", pa.list_(pa.float32(), 16), True)])
 
 
+def test_bare_vector_raises_clear_error():
+    namespace = {
+        "__name__": "test_model_without_pyarrow",
+        "LanceModel": LanceModel,
+        "Vector": Vector,
+    }
+
+    with pytest.raises(TypeError, match=r"Vector must be parameterized.*Vector\(128\)"):
+        exec("class TestModel(LanceModel):\n    vector: Vector", namespace)
+
+
 def test_fixed_size_list_field():
     class TestModel(pydantic.BaseModel):
         vec: Vector(16)
         li: List[int]
 
     data = TestModel(vec=list(range(16)), li=[1, 2, 3])
-    if PYDANTIC_VERSION.major >= 2:
-        assert json.loads(data.model_dump_json()) == {
-            "vec": list(range(16)),
-            "li": [1, 2, 3],
-        }
-    else:
-        assert data.dict() == {
-            "vec": list(range(16)),
-            "li": [1, 2, 3],
-        }
+    assert json.loads(data.model_dump_json()) == {
+        "vec": list(range(16)),
+        "li": [1, 2, 3],
+    }
 
     schema = pydantic_to_schema(TestModel)
     assert schema == pa.schema(
@@ -440,10 +444,7 @@ def test_fixed_size_list_field():
         ]
     )
 
-    if PYDANTIC_VERSION.major >= 2:
-        json_schema = TestModel.model_json_schema()
-    else:
-        json_schema = TestModel.schema()
+    json_schema = TestModel.model_json_schema()
 
     assert json_schema == {
         "properties": {

python/python/tests/test_query.py

@@ -570,6 +570,15 @@ def test_query_builder(table):
     assert all(np.array(rs[0]["vector"]) == [1, 2])
 
 
+def test_query_multiple_vectors(table):
+    results = table.search([np.array([1, 2]), np.array([4, 5])]).limit(1).to_list()
+
+    assert len(results) == 2
+    results_by_query = {result["query_index"]: result for result in results}
+    assert results_by_query[0]["id"] == 1
+    assert results_by_query[1]["id"] == 2
+
+
 def test_with_row_id(table: lancedb.table.Table):
     rs = table.search().with_row_id(True).to_arrow()
     assert "_rowid" in rs.column_names

python/python/tests/test_remote_db.py

@@ -35,6 +35,12 @@ def make_mock_http_handler(handler):
     return MockLanceDBHandler
 
 
+@pytest.mark.parametrize("db_name", ["a" * 64, "invalid..database"])
+def test_connect_rejects_invalid_cloud_dns_hostname(db_name):
+    with pytest.raises(ValueError, match="DNS labels must contain 1 to 63 bytes"):
+        lancedb.connect(f"db://{db_name}", api_key="fake")
+
+
 @contextlib.contextmanager
 def mock_lancedb_connection(handler):
     with http.server.HTTPServer(
@@ -1127,6 +1133,131 @@ def test_stats():
         assert res == stats
 
 
+@contextlib.contextmanager
+def lsm_test_table(lsm_handler):
+    """A remote table whose LSM routes are served by ``lsm_handler``.
+
+    ``lsm_handler(request, route)`` is called for ``/v1/table/test/<route>/``
+    where route is one of flush_lsm, compact_lsm, get_lsm_stats, and is
+    responsible for writing the response.
+    """
+    routes = ("flush_lsm", "compact_lsm", "get_lsm_stats")
+
+    def handler(request):
+        match = re.fullmatch(r"/v1/table/test/(\w+)/", request.path)
+        route = match.group(1) if match else None
+        if route in routes:
+            lsm_handler(request, route)
+        elif route == "describe":
+            request.send_response(200)
+            request.send_header("Content-Type", "application/json")
+            request.end_headers()
+            request.wfile.write(b'{"version": 1, "schema": {"fields": []}}')
+        else:
+            request.send_response(404)
+            request.end_headers()
+
+    with mock_lancedb_connection(handler) as db:
+        yield db.open_table("test")
+
+
+def read_json_body(request):
+    content_len = int(request.headers.get("Content-Length"))
+    return json.loads(request.rfile.read(content_len))
+
+
+def send_json(request, payload, status=200):
+    request.send_response(status)
+    request.send_header("Content-Type", "application/json")
+    request.end_headers()
+    request.wfile.write(json.dumps(payload).encode())
+
+
+def test_get_lsm_stats_sync():
+    """The sync wrapper round-trips the server payload into a dict."""
+    bucket = {
+        "shard_id": "b0",
+        "status": "Active",
+        "writer_epoch": 3,
+        "manifest_version": 12,
+        "current_generation": 6,
+        "replay_after_wal_entry_position": 40,
+        "wal_entry_position_last_seen": 42,
+        "generations": [{"generation": 5, "bytes": 1024, "rows": 7}],
+        "compacting": False,
+        "memtables": [
+            {
+                "generation": 6,
+                "rows": 2,
+                "bytes": 64,
+                "batches": 1,
+                "indexes": ["vec_idx"],
+            }
+        ],
+    }
+    seen_bodies = []
+
+    def lsm_handler(request, route):
+        assert route == "get_lsm_stats"
+        seen_bodies.append(read_json_body(request))
+        send_json(request, {"lsm_stats": {"buckets": [bucket]}})
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.get_lsm_stats() == {"buckets": [bucket]}
+        # Off by default, and forwarded when asked for.
+        assert seen_bodies == [{"include_generation_rows": False}]
+        table.get_lsm_stats(include_generation_rows=True)
+        assert seen_bodies[-1] == {"include_generation_rows": True}
+
+
+def test_get_lsm_stats_sync_returns_none_when_lsm_disabled():
+    """A null envelope means the LSM write path is not enabled, not an error."""
+
+    def lsm_handler(request, route):
+        send_json(request, {"lsm_stats": None})
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.get_lsm_stats() is None
+
+
+def test_flush_and_compact_lsm_sync():
+    """Both are one-shot POSTs answered 202 with no body."""
+    called = []
+
+    def lsm_handler(request, route):
+        called.append(route)
+        request.send_response(202)
+        request.end_headers()
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.flush_lsm() is None
+        assert table.compact_lsm() is None
+        assert called == ["flush_lsm", "compact_lsm"]
+
+
+def test_checkpoint_lsm_sync():
+    """Seal, read the watermark, and return once L0 holds nothing.
+
+    The convergence loop itself is covered in Rust; this pins the sync
+    binding to the endpoints it drives.
+    """
+    called = []
+
+    def lsm_handler(request, route):
+        called.append(route)
+        if route == "get_lsm_stats":
+            # An empty L0 yields no target watermark, so the loop is done
+            # after the seal without ever polling compaction.
+            send_json(request, {"lsm_stats": {"buckets": []}})
+        else:
+            request.send_response(202)
+            request.end_headers()
+
+    with lsm_test_table(lsm_handler) as table:
+        assert table.checkpoint_lsm() is None
+        assert called == ["flush_lsm", "get_lsm_stats"]
+
+
 @contextlib.contextmanager
 def query_test_table(query_handler, *, server_version=Version("0.1.0")):
     def handler(request):

python/python/tests/test_table.py

@@ -2,10 +2,13 @@
 # SPDX-FileCopyrightText: Copyright The LanceDB Authors
 
 
+import ctypes
+import gc
 import os
 import sys
 import threading
 import warnings
+import weakref
 from concurrent.futures import ThreadPoolExecutor
 from datetime import date, datetime, timedelta
 from time import sleep
@@ -99,6 +102,30 @@ def test_basic(mem_db: DBConnection):
     assert table.to_arrow() == expected_data
 
 
+def test_search_preserves_nulls_from_sliced_arrow_table(mem_db: DBConnection):
+    data = pa.table(
+        {
+            "id": [0, 1, 2, 3, 4],
+            "score_cn": [None, 22, None, 5, 8],
+            "score_mt": [None, 42, None, 5, 8],
+            "vector": [
+                [20, 19, -1, -1],
+                [41, 38, 22, 42],
+                [10, 10, -1, -1],
+                [5, 5, 5, 5],
+                [8, 8, 8, 8],
+            ],
+        }
+    ).slice(1)
+
+    table = mem_db.create_table("sliced_nullable", data=data)
+    result = table.search([41, 38, 22, 42]).limit(1).to_arrow()
+
+    assert result["id"].to_pylist() == [1]
+    assert result["score_cn"].to_pylist() == [22]
+    assert result["score_mt"].to_pylist() == [42]
+
+
 def test_table_to_pandas_default_matches_arrow(tmp_db: DBConnection):
     pd = pytest.importorskip("pandas")
     data = pa.table({"id": [1, 2], "text": ["one", "two"]})
@@ -435,6 +462,38 @@ def test_add(mem_db: DBConnection):
     _add(table, schema)
 
 
+def test_add_releases_arrow_buffers_without_gc(mem_db: DBConnection):
+    """Regression test for https://github.com/lancedb/lancedb/issues/2512."""
+    schema = pa.schema([pa.field("x", pa.int64())])
+    table = mem_db.create_table("test_add_releases_arrow_buffers", schema=schema)
+
+    class BufferOwner:
+        def __init__(self, size: int):
+            self.memory = ctypes.create_string_buffer(size)
+
+    owner_refs = []
+    gc_was_enabled = gc.isenabled()
+    gc.disable()
+    try:
+        for _ in range(3):
+            size = 8 * 1024
+            owner = BufferOwner(size)
+            arrow_buffer = pa.foreign_buffer(
+                ctypes.addressof(owner.memory), size, owner
+            )
+            array = pa.Array.from_buffers(pa.int64(), 1024, [None, arrow_buffer])
+            batch = pa.RecordBatch.from_arrays([array], schema=schema)
+            owner_refs.append(weakref.ref(owner))
+
+            table.add(batch)
+            del batch, array, arrow_buffer, owner
+
+        assert all(owner_ref() is None for owner_ref in owner_refs)
+    finally:
+        if gc_was_enabled:
+            gc.enable()
+
+
 def test_add_write_parallelism(mem_db: DBConnection):
     schema = pa.schema([pa.field("id", pa.int64())])
     table = mem_db.create_table("test", schema=schema)
@@ -870,6 +929,7 @@ def test_polars(mem_db: DBConnection):
 
     # enter table to polars dataframe
     result = table.to_polars()
+    assert isinstance(result, pl.LazyFrame)
     assert np.allclose(result.collect()["vector"].to_list(), data["vector"])
 
     # make sure filtering isn't broken
@@ -1786,6 +1846,27 @@ def test_add_with_empty_fixed_size_list_drops_bad_rows(mem_db: DBConnection):
     assert np.allclose(data["embedding"].to_pylist()[0], np.array([0.1] * 16))
 
 
+def test_add_nullable_fixed_size_list_with_none(mem_db: DBConnection):
+    """Regression test for issue #2340."""
+    table = mem_db.create_table(
+        "test_nullable_fixed_size_list",
+        schema=pa.schema(
+            [
+                pa.field("id", pa.string()),
+                pa.field("feature", pa.list_(pa.float32(), 256)),
+                pa.field("tags", pa.list_(pa.string())),
+            ]
+        ),
+    )
+
+    table.add([{"id": "1", "feature": None, "tags": ["tag1", "tag2"]}])
+
+    result = table.to_arrow()
+    assert result.to_pylist() == [
+        {"id": "1", "feature": None, "tags": ["tag1", "tag2"]}
+    ]
+
+
 def test_add_nullable_struct_with_none(mem_db: DBConnection):
     """Regression test for issue #2654: a nullable struct column whose
     first batch contains only None values must not crash in
@@ -1825,6 +1906,33 @@ def test_add_nullable_struct_with_none(mem_db: DBConnection):
     assert result.column("data").to_pylist() == [{"x": 1.0}, None]
 
 
+def test_read_mostly_null_list_v2_2_page_boundary(tmp_path):
+    # Regression test for #3194. This row/value count crosses a v2.2 structural
+    # encoding page boundary where Lance 3.0.0 sliced repetition/definition
+    # levels by row offset and decoded child arrays at different lengths.
+    num_rows = 64_885
+    num_values = 217
+    list_type = pa.list_(pa.float32())
+    source = pa.table(
+        {
+            "id": np.arange(num_rows, dtype=np.int64),
+            "coords": pa.array(
+                [[1.0, 2.0, 3.0, 4.0]] * num_values + [None] * (num_rows - num_values),
+                type=list_type,
+            ),
+        }
+    )
+    db = lancedb.connect(
+        tmp_path,
+        storage_options={"new_table_data_storage_version": "2.2"},
+    )
+    table = db.create_table("test_sparse_nullable_list", data=source)
+
+    result = table.search().select(["id", "coords"]).limit(num_rows).to_arrow()
+
+    assert result.equals(source)
+
+
 def test_add_with_integer_embeddings_preserves_casting(mem_db: DBConnection):
     class Schema(LanceModel):
         text: str
@@ -2110,6 +2218,45 @@ def test_merge(tmp_db: DBConnection, tmp_path):
     table.merge(other_dataset, left_on="id")
 
 
+@pytest.mark.parametrize("storage_version", ["legacy", "stable"])
+def test_search_after_merge(tmp_path, storage_version):
+    pytest.importorskip("lance")
+    pd = pytest.importorskip("pandas")
+
+    db = lancedb.connect(
+        tmp_path,
+        storage_options={"new_table_data_storage_version": storage_version},
+    )
+    rng = np.random.default_rng(42)
+    row_count = 512
+    vectors = rng.standard_normal((row_count, 8)).astype(np.float32)
+    table = db.create_table(
+        "search_after_merge",
+        data=pd.DataFrame(
+            {
+                "id": [str(i) for i in range(row_count)],
+                "vector": list(vectors),
+            }
+        ),
+    )
+    table.create_index("vector", config=IvfPq(num_partitions=1, num_sub_vectors=2))
+
+    links = pd.DataFrame(
+        {
+            "id": [str(i) for i in range(row_count // 2)],
+            "link": [f"https://example.com/{i}" for i in range(row_count // 2)],
+        }
+    )
+    table.merge(links, left_on="id")
+
+    query = table.search(vectors[-1]).refine_factor(50).limit(10)
+    assert "ANN" in query.explain_plan(verbose=True)
+
+    result = query.to_arrow()
+    links_by_id = dict(zip(result["id"].to_pylist(), result["link"].to_pylist()))
+    assert links_by_id[str(row_count - 1)] is None
+
+
 def test_delete(mem_db: DBConnection):
     table = mem_db.create_table(
         "my_table",
@@ -2196,6 +2343,20 @@ def test_update(mem_db: DBConnection):
     assert np.allclose(v, np.array([[1.2, 1.9], [1.1, 1.1]]))
 
 
+def test_update_with_arrow_scalar(mem_db: DBConnection):
+    schema = pa.schema({"id": pa.int64(), "vector": pa.list_(pa.float32(), 4)})
+    table = mem_db.create_table("my_table", schema=schema)
+    table.add([{"id": 1, "vector": [1.0, 2.0, 3.0, 4.0]}])
+
+    value = table.search().select(["vector"]).limit(1).to_arrow()["vector"][0]
+    assert isinstance(value, pa.FixedSizeListScalar)
+
+    result = table.update(where="id == 1", values={"vector": value})
+
+    assert result.rows_updated == 1
+    assert table.to_arrow()["vector"].to_pylist() == [[1.0, 2.0, 3.0, 4.0]]
+
+
 def test_update_types(mem_db: DBConnection):
     table = mem_db.create_table(
         "my_table",
@@ -2363,6 +2524,55 @@ def test_merge_insert(mem_db: DBConnection):
         )
 
 
+def test_merge_insert_nullable_pandas_into_pydantic_schema(mem_db: DBConnection):
+    # Regression test for https://github.com/lancedb/lancedb/issues/2366
+    pd = pytest.importorskip("pandas")
+
+    class Document(LanceModel):
+        id: int
+        title: str
+        content: str
+
+    table = mem_db.create_table("documents", schema=Document)
+    table.add(
+        pd.DataFrame(
+            {
+                "title": ["Old title", "Unchanged"],
+                "id": [2, 3],
+                "content": ["Old content", "Keep this"],
+            }
+        )
+    )
+
+    # Pandas produces nullable Arrow fields, in an order that differs from the
+    # non-nullable Pydantic schema. This is valid as long as the data has no nulls.
+    new_data = pd.DataFrame(
+        {
+            "title": ["Inserted", "Updated"],
+            "id": [1, 2],
+            "content": ["New row", "New content"],
+        }
+    )
+    result = (
+        table.merge_insert("id")
+        .when_matched_update_all()
+        .when_not_matched_insert_all()
+        .execute(new_data)
+    )
+
+    assert result.num_inserted_rows == 1
+    assert result.num_updated_rows == 1
+    expected = pa.Table.from_pylist(
+        [
+            {"id": 1, "title": "Inserted", "content": "New row"},
+            {"id": 2, "title": "Updated", "content": "New content"},
+            {"id": 3, "title": "Unchanged", "content": "Keep this"},
+        ],
+        schema=Document.to_arrow_schema(),
+    )
+    assert table.to_arrow().sort_by("id") == expected
+
+
 def test_merge_insert_by_source_delete_expr(mem_db: DBConnection):
     table = mem_db.create_table(
         "my_table",
@@ -2463,6 +2673,36 @@ def test_merge_insert_subschema(mem_db: DBConnection, data_format):
     assert table.to_arrow().sort_by("id") == expected
 
 
+def test_repeated_partial_merge_insert_with_scalar_index(mem_db: DBConnection):
+    def make_batch(start: int) -> pa.Table:
+        return pa.table(
+            {
+                "id": [f"id-{i:04}" for i in range(start, start + 100)],
+                "category": ["A"] * 100,
+                "value_a": [float(i) for i in range(start, start + 100)],
+                "value_b": [float(i) / 10 for i in range(100)],
+            }
+        )
+
+    table = mem_db.create_table("my_table", data=make_batch(0))
+    table.add(make_batch(100))
+    table.add(make_batch(200))
+    table.create_index("id", config=BTree())
+
+    ids = [f"id-{i:04}" for i in range(100, 200)]
+    for value in (999.0, 888.0):
+        result = (
+            table.merge_insert("id")
+            .when_matched_update_all()
+            .execute(pa.table({"id": ids, "value_a": [value] * 100}))
+        )
+        assert result.num_updated_rows == 100
+
+    actual = table.to_arrow().sort_by("id")
+    assert actual.num_rows == 300
+    assert actual["value_a"].to_pylist()[100:200] == [888.0] * 100
+
+
 @pytest.mark.asyncio
 async def test_merge_insert_async(mem_db_async: AsyncConnection):
     data = pa.table({"a": [1, 2, 3], "b": ["a", "b", "c"]})
@@ -2532,6 +2772,56 @@ async def test_merge_insert_async(mem_db_async: AsyncConnection):
     assert (await table.to_arrow()).sort_by("a") == expected
 
 
+@pytest.mark.skipif(not hasattr(pa, "json_"), reason="requires PyArrow JSON type")
+@pytest.mark.asyncio
+async def test_merge_insert_encodes_json(mem_db_async: AsyncConnection):
+    json_type = pa.json_()
+    schema = pa.schema([pa.field("id", pa.string()), pa.field("j", json_type)])
+
+    def json_table(rows):
+        json_values = pa.ExtensionArray.from_storage(
+            json_type,
+            pa.array([value for _, value in rows], type=json_type.storage_type),
+        )
+        return pa.Table.from_arrays(
+            [pa.array([row_id for row_id, _ in rows]), json_values], schema=schema
+        )
+
+    table = await mem_db_async.create_table("json_merge", schema=schema)
+    await table.add(json_table([("a", '{"k": 1}'), ("b", '{"k": 9}')]))
+
+    await (
+        table.merge_insert("id")
+        .when_matched_update_all()
+        .execute(json_table([("a", '{"k": 2}')]))
+    )
+
+    rows = sorted(await table.query().to_list(), key=lambda row: row["id"])
+    assert rows == [
+        {"id": "a", "j": '{"k":2}'},
+        {"id": "b", "j": '{"k":9}'},
+    ]
+    filtered = await table.query().where("json_extract(j, '$.k') = '2'").to_list()
+    assert filtered == [{"id": "a", "j": '{"k":2}'}]
+
+
+@pytest.mark.skipif(not hasattr(pa, "json_"), reason="requires PyArrow JSON type")
+@pytest.mark.asyncio
+async def test_add_sanitization_encodes_json(mem_db_async: AsyncConnection):
+    json_type = pa.json_()
+    schema = pa.schema([pa.field("id", pa.string()), pa.field("j", json_type)])
+    json_values = pa.ExtensionArray.from_storage(
+        json_type, pa.array(['{"k": 3}'], type=json_type.storage_type)
+    )
+    data = pa.Table.from_arrays([pa.array(["c"]), json_values], schema=schema)
+
+    table = await mem_db_async.create_table("json_add", schema=schema)
+    await table.add(data, on_bad_vectors="fill")
+
+    rows = await table.query().where("json_extract(j, '$.k') = '3'").to_list()
+    assert rows == [{"id": "c", "j": '{"k":3}'}]
+
+
 def test_create_with_embedding_function(mem_db: DBConnection):
     class MyTable(LanceModel):
         text: str
@@ -2559,15 +2849,40 @@ def test_create_with_embedding_function(mem_db: DBConnection):
     assert actual == expected
 
 
+def test_create_f16_table_from_arrow_data(mem_db: DBConnection):
+    dimension = 32
+    num_rows = 512
+    values = pa.array(
+        np.random.default_rng(42)
+        .standard_normal(num_rows * dimension)
+        .astype(np.float16)
+    )
+    df = pa.table(
+        {
+            "text": [f"s-{i}" for i in range(num_rows)],
+            "vector": pa.FixedSizeListArray.from_arrays(values, dimension),
+        }
+    )
+    table = mem_db.create_table("f16_tbl", data=df)
+    assert table.schema.field("vector").type == pa.list_(pa.float16(), dimension)
+    table.create_index(num_partitions=2, num_sub_vectors=2)
+
+    query = df["vector"][2].as_py()
+    expected = table.search(query).limit(2).to_arrow()
+
+    assert "s-2" in expected["text"].to_pylist()
+
+
 def test_create_f16_table(mem_db: DBConnection):
     class MyTable(LanceModel):
         text: str
         vector: Vector(32, value_type=pa.float16())
 
+    rng = np.random.default_rng(42)
     df = pa.table(
         {
             "text": [f"s-{i}" for i in range(512)],
-            "vector": [np.random.randn(32).astype(np.float16) for _ in range(512)],
+            "vector": [rng.standard_normal(32).astype(np.float16) for _ in range(512)],
         }
     )
     table = mem_db.create_table(
@@ -3448,7 +3763,8 @@ def test_stats(mem_db: DBConnection):
     stats = table.stats()
     print(f"{stats=}")
     assert stats == {
-        "total_bytes": 60,
+        # Full on-disk size of the data file, footer and metadata included.
+        "total_bytes": 633,
         "num_rows": 2,
         "num_indices": 0,
         "fragment_stats": {
@@ -3466,6 +3782,13 @@ def test_stats(mem_db: DBConnection):
         },
     }
 
+    # Index files count toward total_bytes too (only deletion files and
+    # manifests are excluded).
+    table.create_index("id", config=BTree())
+    stats_with_index = table.stats()
+    assert stats_with_index["num_indices"] == 1
+    assert stats_with_index["total_bytes"] > stats["total_bytes"]
+
 
 def test_create_table_empty_list_with_schema(mem_db: DBConnection):
     """Test creating table with empty list data and schema
@@ -3489,8 +3812,8 @@ def test_create_table_empty_list_no_schema_error(mem_db: DBConnection):
         mem_db.create_table("test_empty_no_schema", data=[])
 
 
-def test_add_table_with_empty_embeddings(tmp_path):
-    """Test exact scenario from issue #1968
+def test_create_table_without_data_with_vector_schema(tmp_path):
+    """Test exact scenario from issue #1968.
 
     Regression test for issue #1968:
     https://github.com/lancedb/lancedb/issues/1968
@@ -3502,6 +3825,9 @@ def test_add_table_with_empty_embeddings(tmp_path):
         embedding: Vector(16)
 
     table = db.create_table("test", schema=MySchema)
+    assert table.count_rows() == 0
+    assert table.schema == MySchema.to_arrow_schema()
+
     table.add(
         [{"text": "bar", "embedding": [0.1] * 16}],
         on_bad_vectors="drop",
@@ -3578,3 +3904,65 @@ async def test_async_search_runs_embedding_on_dedicated_executor(
     assert all(name.startswith("lancedb-embedding") for name in captured_threads), (
         f"embedding ran off the dedicated executor: {captured_threads}"
     )
+
+
+def test_computed_column_declare_and_refresh(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed", [{"x": 1}, {"x": 2}])
+
+    table.add_columns(computed={"doubled": "x * 2"})
+    assert table.to_arrow()["doubled"].to_pylist() == [None, None]
+
+    result = table.refresh_column("doubled")
+    assert result.rows_filled == 2
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4]
+
+    table.add([{"x": 5}])
+    assert table.refresh_column("doubled").rows_filled == 1
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4, 10]
+
+
+def test_computed_column_rejects_transforms_and_computed_together(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed_mixed", [{"x": 1}])
+    with pytest.raises(ValueError):
+        table.add_columns({"a": "x + 1"}, computed={"b": "x * 2"})
+
+
+@pytest.mark.asyncio
+async def test_computed_column_async(tmp_path):
+    db = await lancedb.connect_async(tmp_path)
+    table = await db.create_table("computed_async", [{"x": 3}])
+
+    await table.add_columns(computed={"tripled": "x * 3"})
+    await table.refresh_column("tripled")
+
+    assert (await table.to_arrow())["tripled"].to_pylist() == [9]
+
+
+def test_refresh_column_async_returns_job(tmp_path):
+    db = lancedb.connect(tmp_path)
+    table = db.create_table("computed_job", [{"x": 1}, {"x": 2}])
+    table.add_columns(computed={"doubled": "x * 2"})
+
+    job = table.refresh_column_async("doubled")
+    assert job.id is None  # in-process jobs have no server id
+    job.wait()
+    assert job.status() == "finished"
+    assert sorted(table.to_arrow()["doubled"].to_pylist()) == [2, 4]
+
+    # Bad input raises at the call, not through the job.
+    with pytest.raises(Exception, match="not a computed column"):
+        table.refresh_column_async("x")
+
+
+@pytest.mark.asyncio
+async def test_refresh_column_async_job_async_table(tmp_path):
+    db = await lancedb.connect_async(tmp_path)
+    table = await db.create_table("computed_job_async", [{"x": 3}])
+    await table.add_columns(computed={"tripled": "x * 3"})
+
+    job = await table.refresh_column_async("tripled")
+    await job.wait()
+    assert await job.status() == "finished"
+    assert (await table.to_arrow())["tripled"].to_pylist() == [9]

python/python/tests/test_voyageai_embeddings.py

@@ -75,6 +75,22 @@ class TestVoyageAIModelRegistration:
         with pytest.raises(ValueError, match="not supported"):
             func.ndims()
 
+    def test_voyage3_source_embeddings_use_text_api(self, mock_voyageai_client):
+        """Regression test for text table data being sent to the multimodal API."""
+        mock_voyageai_client.tokenize.return_value = [["hello", "world"]]
+        mock_voyageai_client.embed.return_value.embeddings = [[0.1] * 1024]
+
+        registry = get_registry()
+        func = registry.get("voyageai").create(name="voyage-3")
+
+        embeddings = func.compute_source_embeddings("hello world")
+
+        assert embeddings == [[0.1] * 1024]
+        mock_voyageai_client.embed.assert_called_once_with(
+            texts=["hello world"], model="voyage-3", input_type="document"
+        )
+        mock_voyageai_client.multimodal_embed.assert_not_called()
+
     @pytest.mark.parametrize(
         "model_name",
         [

python/python/type_tests/connect.py

@@ -0,0 +1,15 @@
+# SPDX-License-Identifier: Apache-2.0
+# SPDX-FileCopyrightText: Copyright The LanceDB Authors
+
+from typing import assert_type
+
+import lancedb
+from lancedb import AsyncConnection, DBConnection
+
+
+def check_connect_type() -> None:
+    assert_type(lancedb.connect("memory://"), DBConnection)
+
+
+async def check_connect_async_type() -> None:
+    assert_type(await lancedb.connect_async("memory://"), AsyncConnection)

python/src/connection.rs

@@ -346,6 +346,23 @@ impl Connection {
         })
     }
 
+    #[pyo3(signature = (name, namespace_path=None))]
+    pub fn drop_table_async(
+        self_: PyRef<'_, Self>,
+        name: String,
+        namespace_path: Option<Vec<String>>,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.get_inner()?.clone();
+        let ns_path = namespace_path.unwrap_or_default();
+        future_into_py(self_.py(), async move {
+            inner
+                .drop_table_async(name, &ns_path)
+                .await
+                .infer_error()
+                .map(crate::job::Job::new)
+        })
+    }
+
     #[pyo3(signature = (namespace_path=None,))]
     pub fn drop_all_tables(
         self_: PyRef<'_, Self>,

python/src/expr.rs

@@ -191,8 +191,27 @@ pub fn expr_lit(value: Bound<'_, PyAny>) -> PyResult<PyExpr> {
     }
 
     // datetime.datetime is a subclass of datetime.date, so it must be checked first.
+    //
+    // Python's datetime.timestamp() treats *naive* datetimes as local wall time.
+    // PyArrow (and therefore Lance table storage) encodes naive timestamps as
+    // UTC wall-clock microseconds. Using .timestamp() for naive values therefore
+    // shifts the literal by the local UTC offset on non-UTC machines, so
+    // `col("ts") == lit(naive_dt)` fails against a table that holds the same
+    // naive value. Fix: treat naive datetimes as UTC wall clock (match Arrow);
+    // keep aware datetimes on the real .timestamp() path (correct epoch).
     if let Ok(dt) = value.cast::<PyDateTime>() {
-        let ts: f64 = dt.call_method0("timestamp")?.extract()?;
+        let ts: f64 = if dt.getattr("tzinfo")?.is_none() {
+            // Force UTC interpretation of the naive wall clock.
+            let utc = pyo3::types::PyModule::import(value.py(), "datetime")?
+                .getattr("timezone")?
+                .getattr("utc")?;
+            let kwargs = pyo3::types::PyDict::new(value.py());
+            kwargs.set_item("tzinfo", utc)?;
+            let aware = dt.call_method("replace", (), Some(&kwargs))?;
+            aware.call_method0("timestamp")?.extract()?
+        } else {
+            dt.call_method0("timestamp")?.extract()?
+        };
         let micros = (ts * 1_000_000.0).round() as i64;
         return Ok(PyExpr(df_lit(ScalarValue::TimestampMicrosecond(
             Some(micros),

python/src/index.rs

@@ -289,7 +289,7 @@ struct IvfHnswFlatParams {
     target_partition_size: Option<u32>,
 }
 
-#[pyclass(get_all)]
+#[pyclass(module = "lancedb._lancedb", get_all)]
 /// A description of an index currently configured on a column
 pub struct IndexConfig {
     /// The type of the index

python/src/lib.rs

@@ -16,7 +16,8 @@ use query::{FTSQuery, HybridQuery, Query, VectorQuery};
 use session::Session;
 use table::{
     AddColumnsResult, AddResult, AlterColumnsResult, DeleteResult, DropColumnsResult, FtsToken,
-    LsmWriteSpec, MergeResult, PyBlobFile, Table, UpdateFieldMetadataResult, UpdateResult,
+    LsmWriteSpec, MergeResult, PyBlobFile, RefreshColumnResult, Table, UpdateFieldMetadataResult,
+    UpdateResult,
 };
 
 pub mod arrow;
@@ -57,6 +58,7 @@ pub fn _lancedb(_py: Python, m: &Bound<'_, PyModule>) -> PyResult<()> {
     m.add_class::<VectorQuery>()?;
     m.add_class::<RecordBatchStream>()?;
     m.add_class::<AddColumnsResult>()?;
+    m.add_class::<RefreshColumnResult>()?;
     m.add_class::<AlterColumnsResult>()?;
     m.add_class::<UpdateFieldMetadataResult>()?;
     m.add_class::<AddResult>()?;

python/src/session.rs

@@ -11,7 +11,7 @@ use pyo3::{PyResult, pyclass, pymethods};
 /// Sessions allow you to configure cache sizes for index and metadata caches,
 /// which can significantly impact memory use and performance. They can
 /// also be re-used across multiple connections to share the same cache state.
-#[pyclass(from_py_object)]
+#[pyclass(module = "lancedb._lancedb", from_py_object)]
 #[derive(Clone)]
 pub struct Session {
     pub(crate) inner: Arc<LanceSession>,

python/src/table.rs

@@ -28,11 +28,72 @@ use pyo3::{
     Bound, FromPyObject, Py, PyAny, PyRef, PyResult, Python,
     exceptions::{PyRuntimeError, PyValueError},
     pyclass, pyfunction, pymethods,
-    types::{IntoPyDict, PyAnyMethods, PyBytes, PyDict, PyDictMethods},
+    types::{IntoPyDict, PyAnyMethods, PyBytes, PyDict, PyDictMethods, PyList, PyListMethods},
 };
 
 mod scannable;
 
+/// Convert `LsmStats` to a Python dict, preserving the per-bucket list.
+///
+/// Deliberately not flattened to a table-level summary: a table is N
+/// buckets on one node, and the per-bucket detail is the reason the
+/// endpoint exists — flattening hides the single hot bucket someone opened
+/// it to find.
+fn lsm_stats_to_py(py: Python<'_>, stats: &lancedb::table::LsmStats) -> PyResult<Py<PyDict>> {
+    let out = PyDict::new(py);
+    let buckets = PyList::empty(py);
+    for b in &stats.buckets {
+        let e = PyDict::new(py);
+        e.set_item("shard_id", &b.shard_id)?;
+        e.set_item("status", &b.status)?;
+        e.set_item("writer_epoch", b.writer_epoch)?;
+        e.set_item("manifest_version", b.manifest_version)?;
+        e.set_item("current_generation", b.current_generation)?;
+        e.set_item(
+            "replay_after_wal_entry_position",
+            b.replay_after_wal_entry_position,
+        )?;
+        e.set_item(
+            "wal_entry_position_last_seen",
+            b.wal_entry_position_last_seen,
+        )?;
+
+        let generations = PyList::empty(py);
+        for g in &b.generations {
+            let ge = PyDict::new(py);
+            ge.set_item("generation", g.generation)?;
+            ge.set_item("bytes", g.bytes)?;
+            ge.set_item("rows", g.rows)?;
+            generations.append(ge)?;
+        }
+        e.set_item("generations", generations)?;
+        e.set_item("compacting", b.compacting)?;
+
+        e.set_item(
+            "memtables",
+            b.memtables
+                .as_ref()
+                .map(|ms| {
+                    let l = PyList::empty(py);
+                    for m in ms {
+                        let d = PyDict::new(py);
+                        d.set_item("generation", m.generation)?;
+                        d.set_item("rows", m.rows)?;
+                        d.set_item("bytes", m.bytes)?;
+                        d.set_item("batches", m.batches)?;
+                        d.set_item("indexes", m.indexes.clone())?;
+                        l.append(d)?;
+                    }
+                    PyResult::Ok(l.unbind())
+                })
+                .transpose()?,
+        )?;
+        buckets.append(e)?;
+    }
+    out.set_item("buckets", buckets)?;
+    Ok(out.unbind())
+}
+
 #[derive(FromPyObject)]
 enum PredicateArg {
     Expr(PyExpr),
@@ -185,13 +246,23 @@ impl From<lancedb::table::MergeResult> for MergeResult {
     }
 }
 
+/// Render for `__repr__`, so the default reads as Python's `None` rather than
+/// Rust's `Some([..])`.
+fn fmt_maintained(maintained: &Option<Vec<String>>) -> String {
+    match maintained {
+        Some(names) => format!("{:?}", names),
+        None => "None".to_string(),
+    }
+}
+
 /// Specification selecting Lance's MemWAL LSM-style write path for
 /// `merge_insert`.
 ///
 /// Constructed via the `bucket(...)`, `identity(...)`, or `unsharded()`
 /// classmethods, then optionally chain `with_maintained_indexes(...)` and
-/// `with_writer_config_defaults(...)`.
-#[pyclass(from_py_object)]
+/// `with_writer_config_defaults(...)`. A fresh spec maintains every index the
+/// MemWAL supports, resolved on install.
+#[pyclass(module = "lancedb._lancedb", from_py_object)]
 #[derive(Clone, Debug)]
 pub struct LsmWriteSpec {
     inner: lancedb::table::LsmWriteSpec,
@@ -230,11 +301,11 @@ impl LsmWriteSpec {
         }
     }
 
-    /// Replace the list of indexes the MemWAL should keep up to date as
-    /// rows are appended. Each name must reference an index that
-    /// already exists on the table at the time `set_lsm_write_spec`
-    /// is called.
-    pub fn with_maintained_indexes(&self, indexes: Vec<String>) -> Self {
+    /// Set which indexes the MemWAL maintains. `None` (the default)
+    /// resolves every supported index on install; a list is verbatim,
+    /// and an empty list maintains nothing.
+    #[pyo3(signature = (indexes))]
+    pub fn with_maintained_indexes(&self, indexes: Option<Vec<String>>) -> Self {
         Self {
             inner: self.inner.clone().with_maintained_indexes(indexes),
         }
@@ -256,23 +327,29 @@ impl LsmWriteSpec {
                 maintained_indexes,
                 writer_config_defaults,
             } => format!(
-                "LsmWriteSpec.bucket(column={:?}, num_buckets={}, maintained_indexes={:?}, writer_config_defaults={:?})",
-                column, num_buckets, maintained_indexes, writer_config_defaults,
+                "LsmWriteSpec.bucket(column={:?}, num_buckets={}, maintained_indexes={}, writer_config_defaults={:?})",
+                column,
+                num_buckets,
+                fmt_maintained(maintained_indexes),
+                writer_config_defaults,
             ),
             lancedb::table::LsmWriteSpec::Identity {
                 column,
                 maintained_indexes,
                 writer_config_defaults,
             } => format!(
-                "LsmWriteSpec.identity(column={:?}, maintained_indexes={:?}, writer_config_defaults={:?})",
-                column, maintained_indexes, writer_config_defaults,
+                "LsmWriteSpec.identity(column={:?}, maintained_indexes={}, writer_config_defaults={:?})",
+                column,
+                fmt_maintained(maintained_indexes),
+                writer_config_defaults,
             ),
             lancedb::table::LsmWriteSpec::Unsharded {
                 maintained_indexes,
                 writer_config_defaults,
             } => format!(
-                "LsmWriteSpec.unsharded(maintained_indexes={:?}, writer_config_defaults={:?})",
-                maintained_indexes, writer_config_defaults,
+                "LsmWriteSpec.unsharded(maintained_indexes={}, writer_config_defaults={:?})",
+                fmt_maintained(maintained_indexes),
+                writer_config_defaults,
             ),
         }
     }
@@ -307,10 +384,10 @@ impl LsmWriteSpec {
         }
     }
 
-    /// Names of indexes the MemWAL should keep up to date during writes.
+    /// Indexes the MemWAL keeps up to date, or `None` for every supported one.
     #[getter]
-    pub fn maintained_indexes(&self) -> Vec<String> {
-        self.inner.maintained_indexes().to_vec()
+    pub fn maintained_indexes(&self) -> Option<Vec<String>> {
+        self.inner.maintained_indexes().map(<[String]>::to_vec)
     }
 
     /// Default `ShardWriter` configuration recorded by this spec.
@@ -338,6 +415,32 @@ pub struct AddColumnsResult {
     pub version: u64,
 }
 
+#[pyclass(get_all, from_py_object)]
+#[derive(Clone, Debug)]
+pub struct RefreshColumnResult {
+    pub rows_filled: u64,
+    pub version: u64,
+}
+
+#[pymethods]
+impl RefreshColumnResult {
+    pub fn __repr__(&self) -> String {
+        format!(
+            "RefreshColumnResult(rows_filled={}, version={})",
+            self.rows_filled, self.version
+        )
+    }
+}
+
+impl From<lancedb::table::RefreshColumnResult> for RefreshColumnResult {
+    fn from(result: lancedb::table::RefreshColumnResult) -> Self {
+        Self {
+            rows_filled: result.rows_filled,
+            version: result.version,
+        }
+    }
+}
+
 #[pymethods]
 impl AddColumnsResult {
     pub fn __repr__(&self) -> String {
@@ -502,7 +605,7 @@ impl PyBlobFile {
     }
 }
 
-#[pyclass(get_all, from_py_object)]
+#[pyclass(module = "lancedb._lancedb", get_all, from_py_object)]
 #[derive(Clone, Debug)]
 pub struct FtsToken {
     pub text: String,
@@ -1339,6 +1442,51 @@ impl Table {
         })
     }
 
+    /// Converge the table's LSM write path into its base table.
+    ///
+    /// Best-effort: with writes flowing, new rows may land after the last
+    /// pass. Errors if the table stops making progress.
+    pub fn checkpoint_lsm(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            inner.checkpoint_lsm().await.infer_error()
+        })
+    }
+
+    /// Seal every bucket's active memtable into L0.
+    pub fn flush_lsm(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(
+            self_.py(),
+            async move { inner.flush_lsm().await.infer_error() },
+        )
+    }
+
+    /// Trigger a background L0 → base pass per bucket. Returns once the
+    /// passes are dispatched, not once they finish — watch `get_lsm_stats`.
+    pub fn compact_lsm(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            inner.compact_lsm().await.infer_error()
+        })
+    }
+
+    /// Live LSM state, or `None` when the LSM write path is not enabled.
+    #[pyo3(signature = (include_generation_rows=false))]
+    pub fn get_lsm_stats(
+        self_: PyRef<'_, Self>,
+        include_generation_rows: bool,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let stats = inner
+                .get_lsm_stats(include_generation_rows)
+                .await
+                .infer_error()?;
+            Python::attach(|py| stats.map(|s| lsm_stats_to_py(py, &s)).transpose())
+        })
+    }
+
     pub fn close_lsm_writers(self_: PyRef<'_, Self>) -> PyResult<Bound<'_, PyAny>> {
         let inner = self_.inner_ref()?.clone();
         future_into_py(self_.py(), async move {
@@ -1388,6 +1536,58 @@ impl Table {
         })
     }
 
+    pub fn add_computed_columns(
+        self_: PyRef<'_, Self>,
+        columns: Vec<(String, String)>,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let mut builder = inner.add_columns();
+            for (name, expression) in columns {
+                builder = builder.computed(name, expression);
+            }
+            let result = builder.execute().await.infer_error()?;
+            Ok(AddColumnsResult::from(result))
+        })
+    }
+
+    pub fn add_function_columns(
+        self_: PyRef<'_, Self>,
+        application_json: String,
+        output_name: Option<String>,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let application =
+            lancedb::function::FunctionApplication::from_json(&application_json).infer_error()?;
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let builder = match output_name {
+                Some(name) => inner.add_columns().function_as(name, application),
+                None => inner.add_columns().function(application),
+            };
+            let result = builder.execute().await.infer_error()?;
+            Ok(AddColumnsResult::from(result))
+        })
+    }
+
+    pub fn refresh_column(self_: PyRef<'_, Self>, column: String) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let result = inner.refresh_column(column).await.infer_error()?;
+            Ok(RefreshColumnResult::from(result))
+        })
+    }
+
+    pub fn refresh_column_async(
+        self_: PyRef<'_, Self>,
+        column: String,
+    ) -> PyResult<Bound<'_, PyAny>> {
+        let inner = self_.inner_ref()?.clone();
+        future_into_py(self_.py(), async move {
+            let job = inner.refresh_column_async(column).await.infer_error()?;
+            Ok(crate::job::Job::new(job))
+        })
+    }
+
     pub fn add_columns_with_schema(
         self_: PyRef<'_, Self>,
         schema: PyArrowType<Schema>,

python/uv.lock

@@ -1998,12 +1998,12 @@ requires-dist = [
     { name = "pillow", marker = "extra == 'clip'", specifier = ">=12.1.1" },
     { name = "pillow", marker = "extra == 'embeddings'", specifier = ">=12.1.1" },
     { name = "pillow", marker = "extra == 'siglip'", specifier = ">=12.1.1" },
-    { name = "polars", marker = "extra == 'tests'", specifier = ">=0.19,<=1.3.0" },
+    { name = "polars", marker = "extra == 'tests'", specifier = ">=0.19,<=1.32.3" },
     { name = "pre-commit", marker = "extra == 'dev'", specifier = ">=3.5.0" },
     { name = "pyarrow", specifier = ">=16" },
     { name = "pyarrow", marker = "extra == 'tests'", specifier = "<25" },
     { name = "pyarrow-stubs", marker = "extra == 'tests'", specifier = ">=16.0" },
-    { name = "pydantic", specifier = ">=1.10" },
+    { name = "pydantic", specifier = ">=2.7.4,<3" },
     { name = "pylance", marker = "extra == 'pylance'", specifier = ">=5.0.0b5" },
     { name = "pylance", marker = "extra == 'tests'", specifier = "==9.0.0rc1" },
     { name = "pyright", marker = "extra == 'dev'", specifier = ">=1.1.350" },