singlestoredb 1.13.0__tar.gz → 1.14.0__tar.gz

{singlestoredb-1.13.0/singlestoredb.egg-info → singlestoredb-1.14.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: singlestoredb
-Version: 1.13.0
+Version: 1.14.0
 Summary: Interface to the SingleStoreDB database and workspace management APIs
 Home-page: https://github.com/singlestore-labs/singlestoredb-python
 Author: SingleStore
@@ -11,7 +11,7 @@ Classifier: License :: OSI Approved :: Apache Software License
 Classifier: Programming Language :: Python :: 3
 Classifier: Programming Language :: Python :: 3 :: Only
 Classifier: Topic :: Database
-Requires-Python: >=3.8
+Requires-Python: >=3.9
 Description-Content-Type: text/markdown
 Provides-Extra: dataframe
 Provides-Extra: dbt

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/setup.cfg

@@ -1,6 +1,6 @@
 [metadata]
 name = singlestoredb
-version = 1.13.0
+version = 1.14.0
 description = Interface to the SingleStoreDB database and workspace management APIs
 long_description = file: README.md
 long_description_content_type = text/markdown
@@ -24,11 +24,12 @@ install_requires =
 	parsimonious
 	requests
 	setuptools
+	singlestore-vectorstore>=0.1.2
 	sqlparams
 	wheel
 	tomli>=1.1.0;python_version < '3.11'
 	typing-extensions<=4.13.2;python_version < '3.11'
-python_requires = >=3.8
+python_requires = >=3.9
 include_package_data = True
 tests_require = 
 	coverage

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/singlestoredb/__init__.py

@@ -13,7 +13,7 @@ Examples
 
 """
 
-__version__ = '1.13.0'
+__version__ = '1.14.0'
 
 from typing import Any
 
@@ -31,6 +31,14 @@ from .types import (
     Date, Time, Timestamp, DateFromTicks, TimeFromTicks, TimestampFromTicks,
     Binary, STRING, BINARY, NUMBER, DATETIME, ROWID,
 )
+from .vectorstore import (
+    vector_db, IndexInterface, IndexList, IndexModel, MatchTypedDict,
+    Metric, IndexStatsTypedDict, NamespaceStatsTypedDict, Vector,
+    VectorDictMetadataValue, VectorMetadataTypedDict, VectorTuple,
+    VectorTupleWithMetadata, DeletionProtection, AndFilter, EqFilter,
+    ExactMatchFilter, FilterTypedDict, GteFilter, GtFilter, InFilter,
+    LteFilter, LtFilter, NeFilter, NinFilter, OrFilter, SimpleFilter,
+)
 
 
 #

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/singlestoredb/ai/__init__.py

@@ -1 +1,2 @@
+from .chat import SingleStoreChatOpenAI  # noqa: F401
 from .embeddings import SingleStoreEmbeddings  # noqa: F401

singlestoredb-1.14.0/singlestoredb/ai/chat.py

@@ -0,0 +1,26 @@
+import os
+from typing import Any
+
+from singlestoredb.fusion.handlers.utils import get_workspace_manager
+
+try:
+    from langchain_openai import ChatOpenAI
+except ImportError:
+    raise ImportError(
+        'Could not import langchain_openai python package. '
+        'Please install it with `pip install langchain_openai`.',
+    )
+
+
+class SingleStoreChatOpenAI(ChatOpenAI):
+    def __init__(self, model_name: str, **kwargs: Any):
+        inference_api_manger = (
+            get_workspace_manager().organizations.current.inference_apis
+        )
+        info = inference_api_manger.get(model_name=model_name)
+        super().__init__(
+            base_url=info.connection_url,
+            api_key=os.environ.get('SINGLESTOREDB_USER_TOKEN'),
+            model=model_name,
+            **kwargs,
+        )

singlestoredb-1.14.0/singlestoredb/ai/embeddings.py

@@ -0,0 +1,27 @@
+import os
+from typing import Any
+
+from singlestoredb.fusion.handlers.utils import get_workspace_manager
+
+try:
+    from langchain_openai import OpenAIEmbeddings
+except ImportError:
+    raise ImportError(
+        'Could not import langchain_openai python package. '
+        'Please install it with `pip install langchain_openai`.',
+    )
+
+
+class SingleStoreEmbeddings(OpenAIEmbeddings):
+
+    def __init__(self, model_name: str, **kwargs: Any):
+        inference_api_manger = (
+            get_workspace_manager().organizations.current.inference_apis
+        )
+        info = inference_api_manger.get(model_name=model_name)
+        super().__init__(
+            base_url=info.connection_url,
+            api_key=os.environ.get('SINGLESTOREDB_USER_TOKEN'),
+            model=model_name,
+            **kwargs,
+        )

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/singlestoredb/connection.py

@@ -1,6 +1,7 @@
 #!/usr/bin/env python
 """SingleStoreDB connections and cursors."""
 import abc
+import functools
 import inspect
 import io
 import queue
@@ -1288,6 +1289,14 @@ class Connection(metaclass=abc.ABCMeta):
         """Access server properties managed by the SHOW statement."""
         return ShowAccessor(self)
 
+    @functools.cached_property
+    def vector_db(self) -> Any:
+        """
+        Get vectorstore API accessor
+        """
+        from vectorstore import VectorDB
+        return VectorDB(connection=self)
+
 
 #
 # NOTE: When adding parameters to this function, you should always

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/singlestoredb/functions/__init__.py

@@ -1,6 +1,10 @@
 from .decorator import udf  # noqa: F401
 from .typing import Masked  # noqa: F401
 from .typing import Table  # noqa: F401
+from .utils import pack_vector  # noqa: F401
+from .utils import pack_vectors  # noqa: F401
+from .utils import unpack_vector  # noqa: F401
+from .utils import unpack_vectors  # noqa: F401
 from .utils import VectorTypes
 
 

{singlestoredb-1.13.0 → singlestoredb-1.14.0}/singlestoredb/functions/ext/asgi.py

@@ -246,6 +246,192 @@ def get_masked_params(func: Callable[..., Any]) -> List[bool]:
     return [typing.get_origin(x.annotation) is Masked for x in params.values()]
 
 
+def build_tuple(x: Any) -> Any:
+    """Convert object to tuple."""
+    return tuple(x) if isinstance(x, Masked) else (x, None)
+
+
+def build_udf_endpoint(
+    func: Callable[..., Any],
+    returns_data_format: str,
+) -> Callable[..., Any]:
+    """
+    Build a UDF endpoint for scalar / list types (row-based).
+
+    Parameters
+    ----------
+    func : Callable
+        The function to call as the endpoint
+    returns_data_format : str
+        The format of the return values
+
+    Returns
+    -------
+    Callable
+        The function endpoint
+
+    """
+    if returns_data_format in ['scalar', 'list']:
+
+        async def do_func(
+            row_ids: Sequence[int],
+            rows: Sequence[Sequence[Any]],
+        ) -> Tuple[Sequence[int], List[Tuple[Any, ...]]]:
+            '''Call function on given rows of data.'''
+            return row_ids, [as_tuple(x) for x in zip(func_map(func, rows))]
+
+        return do_func
+
+    return build_vector_udf_endpoint(func, returns_data_format)
+
+
+def build_vector_udf_endpoint(
+    func: Callable[..., Any],
+    returns_data_format: str,
+) -> Callable[..., Any]:
+    """
+    Build a UDF endpoint for vector formats (column-based).
+
+    Parameters
+    ----------
+    func : Callable
+        The function to call as the endpoint
+    returns_data_format : str
+        The format of the return values
+
+    Returns
+    -------
+    Callable
+        The function endpoint
+
+    """
+    masks = get_masked_params(func)
+    array_cls = get_array_class(returns_data_format)
+
+    async def do_func(
+        row_ids: Sequence[int],
+        cols: Sequence[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
+    ) -> Tuple[
+        Sequence[int],
+        List[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
+    ]:
+        '''Call function on given columns of data.'''
+        row_ids = array_cls(row_ids)
+
+        # Call the function with `cols` as the function parameters
+        if cols and cols[0]:
+            out = func(*[x if m else x[0] for x, m in zip(cols, masks)])
+        else:
+            out = func()
+
+        # Single masked value
+        if isinstance(out, Masked):
+            return row_ids, [tuple(out)]
+
+        # Multiple return values
+        if isinstance(out, tuple):
+            return row_ids, [build_tuple(x) for x in out]
+
+        # Single return value
+        return row_ids, [(out, None)]
+
+    return do_func
+
+
+def build_tvf_endpoint(
+    func: Callable[..., Any],
+    returns_data_format: str,
+) -> Callable[..., Any]:
+    """
+    Build a TVF endpoint for scalar / list types (row-based).
+
+    Parameters
+    ----------
+    func : Callable
+        The function to call as the endpoint
+    returns_data_format : str
+        The format of the return values
+
+    Returns
+    -------
+    Callable
+        The function endpoint
+
+    """
+    if returns_data_format in ['scalar', 'list']:
+
+        async def do_func(
+            row_ids: Sequence[int],
+            rows: Sequence[Sequence[Any]],
+        ) -> Tuple[Sequence[int], List[Tuple[Any, ...]]]:
+            '''Call function on given rows of data.'''
+            out_ids: List[int] = []
+            out = []
+            # Call function on each row of data
+            for i, res in zip(row_ids, func_map(func, rows)):
+                out.extend(as_list_of_tuples(res))
+                out_ids.extend([row_ids[i]] * (len(out)-len(out_ids)))
+            return out_ids, out
+
+        return do_func
+
+    return build_vector_tvf_endpoint(func, returns_data_format)
+
+
+def build_vector_tvf_endpoint(
+    func: Callable[..., Any],
+    returns_data_format: str,
+) -> Callable[..., Any]:
+    """
+    Build a TVF endpoint for vector formats (column-based).
+
+    Parameters
+    ----------
+    func : Callable
+        The function to call as the endpoint
+    returns_data_format : str
+        The format of the return values
+
+    Returns
+    -------
+    Callable
+        The function endpoint
+
+    """
+    masks = get_masked_params(func)
+    array_cls = get_array_class(returns_data_format)
+
+    async def do_func(
+        row_ids: Sequence[int],
+        cols: Sequence[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
+    ) -> Tuple[
+        Sequence[int],
+        List[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
+    ]:
+        '''Call function on given columns of data.'''
+        # NOTE: There is no way to determine which row ID belongs to
+        #        each result row, so we just have to use the same
+        #        row ID for all rows in the result.
+
+        # Call function on each column of data
+        if cols and cols[0]:
+            res = get_dataframe_columns(
+                func(*[x if m else x[0] for x, m in zip(cols, masks)]),
+            )
+        else:
+            res = get_dataframe_columns(func())
+
+        # Generate row IDs
+        if isinstance(res[0], Masked):
+            row_ids = array_cls([row_ids[0]] * len(res[0][0]))
+        else:
+            row_ids = array_cls([row_ids[0]] * len(res[0]))
+
+        return row_ids, [build_tuple(x) for x in res]
+
+    return do_func
+
+
 def make_func(
     name: str,
     func: Callable[..., Any],
@@ -273,102 +459,10 @@ def make_func(
     args_data_format = sig.get('args_data_format', 'scalar')
     returns_data_format = sig.get('returns_data_format', 'scalar')
 
-    masks = get_masked_params(func)
-
     if function_type == 'tvf':
-        # Scalar / list types (row-based)
-        if returns_data_format in ['scalar', 'list']:
-            async def do_func(
-                row_ids: Sequence[int],
-                rows: Sequence[Sequence[Any]],
-            ) -> Tuple[Sequence[int], List[Tuple[Any, ...]]]:
-                '''Call function on given rows of data.'''
-                out_ids: List[int] = []
-                out = []
-                # Call function on each row of data
-                for i, res in zip(row_ids, func_map(func, rows)):
-                    out.extend(as_list_of_tuples(res))
-                    out_ids.extend([row_ids[i]] * (len(out)-len(out_ids)))
-                return out_ids, out
-
-        # Vector formats (column-based)
-        else:
-            array_cls = get_array_class(returns_data_format)
-
-            async def do_func(  # type: ignore
-                row_ids: Sequence[int],
-                cols: Sequence[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
-            ) -> Tuple[
-                Sequence[int],
-                List[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
-            ]:
-                '''Call function on given cols of data.'''
-                # NOTE: There is no way to determine which row ID belongs to
-                #        each result row, so we just have to use the same
-                #        row ID for all rows in the result.
-
-                def build_tuple(x: Any) -> Any:
-                    return tuple(x) if isinstance(x, Masked) else (x, None)
-
-                # Call function on each column of data
-                if cols and cols[0]:
-                    res = get_dataframe_columns(
-                        func(*[x if m else x[0] for x, m in zip(cols, masks)]),
-                    )
-                else:
-                    res = get_dataframe_columns(func())
-
-                # Generate row IDs
-                if isinstance(res[0], Masked):
-                    row_ids = array_cls([row_ids[0]] * len(res[0][0]))
-                else:
-                    row_ids = array_cls([row_ids[0]] * len(res[0]))
-
-                return row_ids, [build_tuple(x) for x in res]
-
+        do_func = build_tvf_endpoint(func, returns_data_format)
     else:
-        # Scalar / list types (row-based)
-        if returns_data_format in ['scalar', 'list']:
-            async def do_func(
-                row_ids: Sequence[int],
-                rows: Sequence[Sequence[Any]],
-            ) -> Tuple[Sequence[int], List[Tuple[Any, ...]]]:
-                '''Call function on given rows of data.'''
-                return row_ids, [as_tuple(x) for x in zip(func_map(func, rows))]
-
-        # Vector formats (column-based)
-        else:
-            array_cls = get_array_class(returns_data_format)
-
-            async def do_func(  # type: ignore
-                row_ids: Sequence[int],
-                cols: Sequence[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
-            ) -> Tuple[
-                Sequence[int],
-                List[Tuple[Sequence[Any], Optional[Sequence[bool]]]],
-            ]:
-                '''Call function on given cols of data.'''
-                row_ids = array_cls(row_ids)
-
-                def build_tuple(x: Any) -> Any:
-                    return tuple(x) if isinstance(x, Masked) else (x, None)
-
-                # Call the function with `cols` as the function parameters
-                if cols and cols[0]:
-                    out = func(*[x if m else x[0] for x, m in zip(cols, masks)])
-                else:
-                    out = func()
-
-                # Single masked value
-                if isinstance(out, Masked):
-                    return row_ids, [tuple(out)]
-
-                # Multiple return values
-                if isinstance(out, tuple):
-                    return row_ids, [build_tuple(x) for x in out]
-
-                # Single return value
-                return row_ids, [(out, None)]
+        do_func = build_udf_endpoint(func, returns_data_format)
 
     do_func.__name__ = name
     do_func.__doc__ = func.__doc__