datachain 0.6.8__tar.gz → 0.6.10__tar.gz

{datachain-0.6.8 → datachain-0.6.10}/.pre-commit-config.yaml

@@ -24,7 +24,7 @@ repos:
       - id: trailing-whitespace
         exclude: '^LICENSES/'
   - repo: https://github.com/astral-sh/ruff-pre-commit
-    rev: 'v0.7.2'
+    rev: 'v0.7.3'
     hooks:
       - id: ruff
         args: [--fix, --exit-non-zero-on-fix]

{datachain-0.6.8/src/datachain.egg-info → datachain-0.6.10}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: datachain
-Version: 0.6.8
+Version: 0.6.10
 Summary: Wrangle unstructured AI data at scale
 Author-email: Dmitry Petrov <support@dvc.org>
 License: Apache-2.0
@@ -71,7 +71,7 @@ Requires-Dist: pytest<9,>=8; extra == "tests"
 Requires-Dist: pytest-sugar>=0.9.6; extra == "tests"
 Requires-Dist: pytest-cov>=4.1.0; extra == "tests"
 Requires-Dist: pytest-mock>=3.12.0; extra == "tests"
-Requires-Dist: pytest-servers[all]>=0.5.7; extra == "tests"
+Requires-Dist: pytest-servers[all]>=0.5.8; extra == "tests"
 Requires-Dist: pytest-benchmark[histogram]; extra == "tests"
 Requires-Dist: pytest-xdist>=3.3.1; extra == "tests"
 Requires-Dist: virtualenv; extra == "tests"
@@ -120,33 +120,41 @@ Requires-Dist: onnx==1.16.1; extra == "examples"
    :target: https://github.com/iterative/datachain/actions/workflows/tests.yml
    :alt: Tests
 
-DataChain is a modern Pythonic data-frame library designed for artificial intelligence.
-It is made to organize your unstructured data into datasets and wrangle it at scale on
-your local machine. Datachain does not abstract or hide the AI models and API calls, but helps to integrate them into the postmodern data stack.
+DataChain is a Python-based AI-data warehouse for transforming and analyzing unstructured
+data like images, audio, videos, text and PDFs. It integrates with external storage
+(e.g., S3) to process data efficiently without data duplication and manages metadata
+in an internal database for easy and efficient querying.
+
+
+Use Cases
+=========
+
+1. **Multimodal Dataset Preparation and Curation**: ideal for organizing and
+   refining data in pre-training, finetuning or LLM evaluating stages.
+2. **GenAI Data Analytics**: Enables advanced analytics for multimodal data and
+   ad-hoc analytics using LLMs.
 
 Key Features
 ============
 
-📂 **Storage as a Source of Truth.**
-   - Process unstructured data without redundant copies from S3, GCP, Azure, and local
-     file systems.
-   - Multimodal data support: images, video, text, PDFs, JSONs, CSVs, parquet.
+📂 **Multimodal Dataset Versioning.**
+   - Version unstructured data without redundant data copies, by supporitng
+     references to S3, GCP, Azure, and local file systems.
+   - Multimodal data support: images, video, text, PDFs, JSONs, CSVs, parquet, etc.
    - Unite files and metadata together into persistent, versioned, columnar datasets.
 
-🐍 **Python-friendly data pipelines.**
-   - Operate on Python objects and object fields.
-   - Built-in parallelization and out-of-memory compute without SQL or Spark.
+🐍 **Python-friendly.**
+   - Operate on Python objects and object fields: float scores, strings, matrixes,
+     LLM response objects.
+   - Run Python code in a high-scale, terabytes size datasets, with built-in
+     parallelization and memory-efficient computing — no SQL or Spark required.
 
 🧠 **Data Enrichment and Processing.**
    - Generate metadata using local AI models and LLM APIs.
-   - Filter, join, and group by metadata. Search by vector embeddings.
+   - Filter, join, and group datasets by metadata. Search by vector embeddings.
+   - High-performance vectorized operations on Python objects: sum, count, avg, etc.
    - Pass datasets to Pytorch and Tensorflow, or export them back into storage.
 
-🚀 **Efficiency.**
-   - Parallelization, out-of-memory workloads and data caching.
-   - Vectorized operations on Python object fields: sum, count, avg, etc.
-   - Optimized vector search.
-
 
 Quick Start
 -----------
@@ -196,7 +204,7 @@ Batch inference with a simple sentiment model using the `transformers` library:
 
     pip install transformers
 
-The code below downloads files the cloud, and applies a user-defined function
+The code below downloads files from the cloud, and applies a user-defined function
 to each one of them. All files with a positive sentiment
 detected are then copied to the local directory.
 
@@ -429,6 +437,19 @@ name suffix, the following code will do it:
     loader = DataLoader(chain, batch_size=1)
 
 
+DataChain Studio Platform
+-------------------------
+
+`DataChain Studio`_ is a proprietary solution for teams that offers:
+
+- **Centralized dataset registry** to manage data, code and dependency
+  dependencies in one place.
+- **Data Lineage** for data sources as well as direvative dataset.
+- **UI for Multimodal Data** like images, videos, and PDFs.
+- **Scalable Compute** to handle large datasets (100M+ files) and in-house
+  AI model inference.
+- **Access control** including SSO and team based collaboration.
+
 Tutorials
 ---------
 
@@ -462,6 +483,5 @@ Community and Support
 .. _Pydantic: https://github.com/pydantic/pydantic
 .. _publicly available: https://radar.kit.edu/radar/en/dataset/FdJmclKpjHzLfExE.ExpBot%2B-%2BA%2Bdataset%2Bof%2B79%2Bdialogs%2Bwith%2Ban%2Bexperimental%2Bcustomer%2Bservice%2Bchatbot
 .. _SQLite: https://www.sqlite.org/
-.. _Getting Started: https://datachain.dvc.ai/
-.. |Flowchart| image:: https://github.com/iterative/datachain/blob/main/docs/assets/flowchart.png?raw=true
-   :alt: DataChain FlowChart
+.. _Getting Started: https://docs.datachain.ai/
+.. _DataChain Studio: https://studio.datachain.ai/

{datachain-0.6.8 → datachain-0.6.10}/README.rst

@@ -19,33 +19,41 @@
    :target: https://github.com/iterative/datachain/actions/workflows/tests.yml
    :alt: Tests
 
-DataChain is a modern Pythonic data-frame library designed for artificial intelligence.
-It is made to organize your unstructured data into datasets and wrangle it at scale on
-your local machine. Datachain does not abstract or hide the AI models and API calls, but helps to integrate them into the postmodern data stack.
+DataChain is a Python-based AI-data warehouse for transforming and analyzing unstructured
+data like images, audio, videos, text and PDFs. It integrates with external storage
+(e.g., S3) to process data efficiently without data duplication and manages metadata
+in an internal database for easy and efficient querying.
+
+
+Use Cases
+=========
+
+1. **Multimodal Dataset Preparation and Curation**: ideal for organizing and
+   refining data in pre-training, finetuning or LLM evaluating stages.
+2. **GenAI Data Analytics**: Enables advanced analytics for multimodal data and
+   ad-hoc analytics using LLMs.
 
 Key Features
 ============
 
-📂 **Storage as a Source of Truth.**
-   - Process unstructured data without redundant copies from S3, GCP, Azure, and local
-     file systems.
-   - Multimodal data support: images, video, text, PDFs, JSONs, CSVs, parquet.
+📂 **Multimodal Dataset Versioning.**
+   - Version unstructured data without redundant data copies, by supporitng
+     references to S3, GCP, Azure, and local file systems.
+   - Multimodal data support: images, video, text, PDFs, JSONs, CSVs, parquet, etc.
    - Unite files and metadata together into persistent, versioned, columnar datasets.
 
-🐍 **Python-friendly data pipelines.**
-   - Operate on Python objects and object fields.
-   - Built-in parallelization and out-of-memory compute without SQL or Spark.
+🐍 **Python-friendly.**
+   - Operate on Python objects and object fields: float scores, strings, matrixes,
+     LLM response objects.
+   - Run Python code in a high-scale, terabytes size datasets, with built-in
+     parallelization and memory-efficient computing — no SQL or Spark required.
 
 🧠 **Data Enrichment and Processing.**
    - Generate metadata using local AI models and LLM APIs.
-   - Filter, join, and group by metadata. Search by vector embeddings.
+   - Filter, join, and group datasets by metadata. Search by vector embeddings.
+   - High-performance vectorized operations on Python objects: sum, count, avg, etc.
    - Pass datasets to Pytorch and Tensorflow, or export them back into storage.
 
-🚀 **Efficiency.**
-   - Parallelization, out-of-memory workloads and data caching.
-   - Vectorized operations on Python object fields: sum, count, avg, etc.
-   - Optimized vector search.
-
 
 Quick Start
 -----------
@@ -95,7 +103,7 @@ Batch inference with a simple sentiment model using the `transformers` library:
 
     pip install transformers
 
-The code below downloads files the cloud, and applies a user-defined function
+The code below downloads files from the cloud, and applies a user-defined function
 to each one of them. All files with a positive sentiment
 detected are then copied to the local directory.
 
@@ -328,6 +336,19 @@ name suffix, the following code will do it:
     loader = DataLoader(chain, batch_size=1)
 
 
+DataChain Studio Platform
+-------------------------
+
+`DataChain Studio`_ is a proprietary solution for teams that offers:
+
+- **Centralized dataset registry** to manage data, code and dependency
+  dependencies in one place.
+- **Data Lineage** for data sources as well as direvative dataset.
+- **UI for Multimodal Data** like images, videos, and PDFs.
+- **Scalable Compute** to handle large datasets (100M+ files) and in-house
+  AI model inference.
+- **Access control** including SSO and team based collaboration.
+
 Tutorials
 ---------
 
@@ -361,6 +382,5 @@ Community and Support
 .. _Pydantic: https://github.com/pydantic/pydantic
 .. _publicly available: https://radar.kit.edu/radar/en/dataset/FdJmclKpjHzLfExE.ExpBot%2B-%2BA%2Bdataset%2Bof%2B79%2Bdialogs%2Bwith%2Ban%2Bexperimental%2Bcustomer%2Bservice%2Bchatbot
 .. _SQLite: https://www.sqlite.org/
-.. _Getting Started: https://datachain.dvc.ai/
-.. |Flowchart| image:: https://github.com/iterative/datachain/blob/main/docs/assets/flowchart.png?raw=true
-   :alt: DataChain FlowChart
+.. _Getting Started: https://docs.datachain.ai/
+.. _DataChain Studio: https://studio.datachain.ai/

{datachain-0.6.8 → datachain-0.6.10}/mkdocs.yml

@@ -15,7 +15,7 @@ validation:
 
 theme:
   name: material
-  custom_dir: overrides
+  custom_dir: docs/overrides
   logo: assets/datachain-white.svg
   favicon: assets/datachain.svg
   icon:

{datachain-0.6.8 → datachain-0.6.10}/pyproject.toml

@@ -82,7 +82,7 @@ tests = [
   "pytest-sugar>=0.9.6",
   "pytest-cov>=4.1.0",
   "pytest-mock>=3.12.0",
-  "pytest-servers[all]>=0.5.7",
+  "pytest-servers[all]>=0.5.8",
   "pytest-benchmark[histogram]",
   "pytest-xdist>=3.3.1",
   "virtualenv",

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/catalog/catalog.py

@@ -603,9 +603,10 @@ class Catalog:
         )
 
         lst = Listing(
+            self.metastore.clone(),
             self.warehouse.clone(),
             Client.get_client(list_uri, self.cache, **self.client_config),
-            self.get_dataset(list_ds_name),
+            dataset_name=list_ds_name,
             object_name=object_name,
         )
 
@@ -698,9 +699,13 @@ class Catalog:
 
                     client = self.get_client(source, **client_config)
                     uri = client.uri
-                    st = self.warehouse.clone()
                     dataset_name, _, _, _ = DataChain.parse_uri(uri, self.session)
-                    listing = Listing(st, client, self.get_dataset(dataset_name))
+                    listing = Listing(
+                        self.metastore.clone(),
+                        self.warehouse.clone(),
+                        client,
+                        dataset_name=dataset_name,
+                    )
                     rows = DatasetQuery(
                         name=dataset.name, version=ds_version, catalog=self
                     ).to_db_records()
@@ -769,6 +774,7 @@ class Catalog:
         create_rows: Optional[bool] = True,
         validate_version: Optional[bool] = True,
         listing: Optional[bool] = False,
+        uuid: Optional[str] = None,
     ) -> "DatasetRecord":
         """
         Creates new dataset of a specific version.
@@ -816,6 +822,7 @@ class Catalog:
             query_script=query_script,
             create_rows_table=create_rows,
             columns=columns,
+            uuid=uuid,
         )
 
     def create_new_dataset_version(
@@ -832,6 +839,7 @@ class Catalog:
         script_output="",
         create_rows_table=True,
         job_id: Optional[str] = None,
+        uuid: Optional[str] = None,
     ) -> DatasetRecord:
         """
         Creates dataset version if it doesn't exist.
@@ -855,6 +863,7 @@ class Catalog:
             schema=schema,
             job_id=job_id,
             ignore_if_exists=True,
+            uuid=uuid,
         )
 
         if create_rows_table:
@@ -1350,6 +1359,13 @@ class Catalog:
             # we will create new one if it doesn't exist
             pass
 
+        if dataset and version and dataset.has_version(version):
+            """No need to communicate with Studio at all"""
+            dataset_uri = create_dataset_uri(remote_dataset_name, version)
+            print(f"Local copy of dataset {dataset_uri} already present")
+            _instantiate_dataset()
+            return
+
         remote_dataset = self.get_remote_dataset(remote_dataset_name)
         # if version is not specified in uri, take the latest one
         if not version:
@@ -1400,6 +1416,7 @@ class Catalog:
             columns=columns,
             feature_schema=remote_dataset_version.feature_schema,
             validate_version=False,
+            uuid=remote_dataset_version.uuid,
         )
 
         # asking remote to export dataset rows table to s3 and to return signed

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/client/fsspec.py

@@ -358,7 +358,7 @@ class Client(ABC):
     ) -> BinaryIO:
         """Open a file, including files in tar archives."""
         if use_cache and (cache_path := self.cache.get_path(file)):
-            return open(cache_path, mode="rb")  # noqa: SIM115
+            return open(cache_path, mode="rb")
         assert not file.location
         return FileWrapper(self.fs.open(self.get_full_path(file.path)), cb)  # type: ignore[return-value]
 

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/data_storage/metastore.py

@@ -138,6 +138,7 @@ class AbstractMetastore(ABC, Serializable):
         size: Optional[int] = None,
         preview: Optional[list[dict]] = None,
         job_id: Optional[str] = None,
+        uuid: Optional[str] = None,
     ) -> DatasetRecord:
         """Creates new dataset version."""
 
@@ -352,6 +353,7 @@ class AbstractDBMetastore(AbstractMetastore):
         """Datasets versions table columns."""
         return [
             Column("id", Integer, primary_key=True),
+            Column("uuid", Text, nullable=False, default=uuid4()),
             Column(
                 "dataset_id",
                 Integer,
@@ -545,6 +547,7 @@ class AbstractDBMetastore(AbstractMetastore):
         size: Optional[int] = None,
         preview: Optional[list[dict]] = None,
         job_id: Optional[str] = None,
+        uuid: Optional[str] = None,
         conn=None,
     ) -> DatasetRecord:
         """Creates new dataset version."""
@@ -555,6 +558,7 @@ class AbstractDBMetastore(AbstractMetastore):
 
         query = self._datasets_versions_insert().values(
             dataset_id=dataset.id,
+            uuid=uuid or str(uuid4()),
             version=version,
             status=status,
             feature_schema=json.dumps(feature_schema or {}),

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/data_storage/sqlite.py

@@ -747,8 +747,12 @@ class SQLiteWarehouse(AbstractWarehouse):
 
         ids = self.db.execute(select_ids).fetchall()
 
-        select_q = query.with_only_columns(
-            *[c for c in query.selected_columns if c.name != "sys__id"]
+        select_q = (
+            query.with_only_columns(
+                *[c for c in query.selected_columns if c.name != "sys__id"]
+            )
+            .offset(None)
+            .limit(None)
         )
 
         for batch in batched_it(ids, 10_000):

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/dataset.py

@@ -163,6 +163,7 @@ class DatasetStatus:
 @dataclass
 class DatasetVersion:
     id: int
+    uuid: str
     dataset_id: int
     version: int
     status: int
@@ -184,6 +185,7 @@ class DatasetVersion:
     def parse(  # noqa: PLR0913
         cls: type[V],
         id: int,
+        uuid: str,
         dataset_id: int,
         version: int,
         status: int,
@@ -203,6 +205,7 @@ class DatasetVersion:
     ):
         return cls(
             id,
+            uuid,
             dataset_id,
             version,
             status,
@@ -306,6 +309,7 @@ class DatasetRecord:
         query_script: str,
         schema: str,
         version_id: int,
+        version_uuid: str,
         version_dataset_id: int,
         version: int,
         version_status: int,
@@ -331,6 +335,7 @@ class DatasetRecord:
 
         dataset_version = DatasetVersion.parse(
             version_id,
+            version_uuid,
             version_dataset_id,
             version,
             version_status,

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/lib/dataset_info.py

@@ -1,6 +1,7 @@
 import json
 from datetime import datetime
 from typing import TYPE_CHECKING, Any, Optional, Union
+from uuid import uuid4
 
 from pydantic import Field, field_validator
 
@@ -15,6 +16,7 @@ if TYPE_CHECKING:
 
 class DatasetInfo(DataModel):
     name: str
+    uuid: str = Field(default=str(uuid4()))
     version: int = Field(default=1)
     status: int = Field(default=DatasetStatus.CREATED)
     created_at: datetime = Field(default=TIME_ZERO)
@@ -60,6 +62,7 @@ class DatasetInfo(DataModel):
         job: Optional[Job],
     ) -> "Self":
         return cls(
+            uuid=version.uuid,
             name=dataset.name,
             version=version.version,
             status=version.status,

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/lib/dc.py

@@ -30,7 +30,7 @@ from datachain.client.local import FileClient
 from datachain.dataset import DatasetRecord
 from datachain.lib.convert.python_to_sql import python_to_sql
 from datachain.lib.convert.values_to_tuples import values_to_tuples
-from datachain.lib.data_model import DataModel, DataType, dict_to_data_model
+from datachain.lib.data_model import DataModel, DataType, DataValue, dict_to_data_model
 from datachain.lib.dataset_info import DatasetInfo
 from datachain.lib.file import ArrowRow, File, get_file_type
 from datachain.lib.file import ExportPlacement as FileExportPlacement
@@ -642,6 +642,59 @@ class DataChain:
         }
         return chain.gen(**signal_dict)  # type: ignore[misc, arg-type]
 
+    def explode(
+        self,
+        col: str,
+        model_name: Optional[str] = None,
+        object_name: Optional[str] = None,
+    ) -> "DataChain":
+        """Explodes a column containing JSON objects (dict or str DataChain type) into
+           individual columns based on the schema of the JSON. Schema is inferred from
+           the first row of the column.
+
+        Args:
+            col: the name of the column containing JSON to be exploded.
+            model_name: optional generated model name.  By default generates the name
+                automatically.
+            object_name: optional generated object column name. By default generates the
+                name automatically.
+
+        Returns:
+            DataChain: A new DataChain instance with the new set of columns.
+        """
+        import json
+
+        import pyarrow as pa
+
+        from datachain.lib.arrow import schema_to_output
+
+        json_value = next(self.limit(1).collect(col))
+        json_dict = (
+            json.loads(json_value) if isinstance(json_value, str) else json_value
+        )
+
+        if not isinstance(json_dict, dict):
+            raise TypeError(f"Column {col} should be a string or dict type with JSON")
+
+        schema = pa.Table.from_pylist([json_dict]).schema
+        output = schema_to_output(schema, None)
+
+        if not model_name:
+            model_name = f"{col.title()}ExplodedModel"
+
+        model = dict_to_data_model(model_name, output)
+
+        def json_to_model(json_value: Union[str, dict]):
+            json_dict = (
+                json.loads(json_value) if isinstance(json_value, str) else json_value
+            )
+            return model.model_validate(json_dict)
+
+        if not object_name:
+            object_name = f"{col}_expl"
+
+        return self.map(json_to_model, params=col, output={object_name: model})
+
     @classmethod
     def datasets(
         cls,
@@ -895,7 +948,7 @@ class DataChain:
         2. Group-based UDF function input: Instead of individual rows, the function
            receives a list all rows within each group defined by `partition_by`.
 
-        Example:
+        Examples:
             ```py
             chain = chain.agg(
                 total=lambda category, amount: [sum(amount)],
@@ -904,6 +957,26 @@ class DataChain:
             )
             chain.save("new_dataset")
             ```
+
+            An alternative syntax, when you need to specify a more complex function:
+
+            ```py
+            # It automatically resolves which columns to pass to the function
+            # by looking at the function signature.
+            def agg_sum(
+                file: list[File], amount: list[float]
+            ) -> Iterator[tuple[File, float]]:
+                yield file[0], sum(amount)
+
+            chain = chain.agg(
+                agg_sum,
+                output={"file": File, "total": float},
+                # Alternative syntax is to use `C` (short for Column) to specify
+                # a column name or a nested column, e.g. C("file.path").
+                partition_by=C("category"),
+            )
+            chain.save("new_dataset")
+            ```
         """
         udf_obj = self._udf_to_obj(Aggregator, func, params, output, signal_map)
         return self._evolve(
@@ -1242,15 +1315,15 @@ class DataChain:
         return self.results(row_factory=to_dict)
 
     @overload
-    def collect(self) -> Iterator[tuple[DataType, ...]]: ...
+    def collect(self) -> Iterator[tuple[DataValue, ...]]: ...
 
     @overload
-    def collect(self, col: str) -> Iterator[DataType]: ...  # type: ignore[overload-overlap]
+    def collect(self, col: str) -> Iterator[DataValue]: ...
 
     @overload
-    def collect(self, *cols: str) -> Iterator[tuple[DataType, ...]]: ...
+    def collect(self, *cols: str) -> Iterator[tuple[DataValue, ...]]: ...
 
-    def collect(self, *cols: str) -> Iterator[Union[DataType, tuple[DataType, ...]]]:  # type: ignore[overload-overlap,misc]
+    def collect(self, *cols: str) -> Iterator[Union[DataValue, tuple[DataValue, ...]]]:  # type: ignore[overload-overlap,misc]
         """Yields rows of values, optionally limited to the specified columns.
 
         Args:

{datachain-0.6.8 → datachain-0.6.10}/src/datachain/lib/meta_formats.py

@@ -114,6 +114,7 @@ def read_meta(  # noqa: C901
             )
         )
         (model_output,) = chain.collect("meta_schema")
+        assert isinstance(model_output, str)
         if print_schema:
             print(f"{model_output}")
         # Below 'spec' should be a dynamically converted DataModel from Pydantic

datachain-0.6.10/src/datachain/lib/models/__init__.py

@@ -0,0 +1,6 @@
+from . import ultralytics
+from .bbox import BBox, OBBox
+from .pose import Pose, Pose3D
+from .segment import Segments
+
+__all__ = ["BBox", "OBBox", "Pose", "Pose3D", "Segments", "ultralytics"]