kosh 3.4__tar.gz → 3.6__tar.gz

{kosh-3.4/kosh.egg-info → kosh-3.6}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: kosh
-Version: 3.4
+Version: 3.6
 Summary: Manages Data Store with External Bulk Data
 Author: Charles Doutriaux, Renee Olson
 License: MIT
@@ -20,6 +20,9 @@ Requires-Dist: pandas
 Requires-Dist: hdbscan
 Requires-Dist: matplotlib
 Requires-Dist: tqdm
+Provides-Extra: mcp
+Requires-Dist: mada; extra == "mcp"
+Requires-Dist: mada-tools; extra == "mcp"
 Dynamic: license-file
 
 ![Kosh Logo](share/icons/png/Kosh_Logo_Blue.png)
@@ -59,6 +62,9 @@ For more details look into the [installation doc](INSTALL.md)
 
 If you want to use Kosh through Codex, see [the Codex skill install guide](docs/codex_skill.md).
 
+If you want to use Kosh from a MADA agent or another MCP-compatible client,
+see the [Kosh MCP server guide](docs/source/users/mcp_server.md).
+
 # First steps
 
 See [this file](docs/source/users/index.md)

{kosh-3.4 → kosh-3.6}/README.md

@@ -35,6 +35,9 @@ For more details look into the [installation doc](INSTALL.md)
 
 If you want to use Kosh through Codex, see [the Codex skill install guide](docs/codex_skill.md).
 
+If you want to use Kosh from a MADA agent or another MCP-compatible client,
+see the [Kosh MCP server guide](docs/source/users/mcp_server.md).
+
 # First steps
 
 See [this file](docs/source/users/index.md)

kosh-3.6/kosh/current_version.py

@@ -0,0 +1 @@
+current_version = "3.6"

{kosh-3.4 → kosh-3.6}/kosh/dataset.py

@@ -1463,7 +1463,7 @@ class KoshDataset(KoshSinaObject):
         return attributes
 
     @lock_strategies.lock_method
-    def to_dataframe(self, data_columns=[], *atts, **keys):
+    def to_dataframe(self, data_columns=[], typed_columns=None, *atts, **keys):
         """Return the find object as a Pandas DataFrame.
 
         Pass in the same arguments and keyword arguments as the find method.
@@ -1482,6 +1482,39 @@ class KoshDataset(KoshSinaObject):
         :param data_columns: Columns to extract. By default this will include ['id', 'mime_type', 'uri', 'associated'].
                             If nothing is passed, will return all data.
         :type data_columns: Union(str, list), optional
+        :param typed_columns: Dictionary of column types and replacement values.
+            If omitted, MySQL defaults are returned.
+
+            Supported types:
+                [
+                    "Int8", "Int16", "Int32", "Int64",
+                    "UInt8", "UInt16", "UInt32", "UInt64",
+                    "float16", "float32", "float64",
+                    "Float32", "Float64",
+                    "bool", "boolean", "string", "object",
+                    "category", "datetime64[ns]", "timedelta64[ns]",
+                ]
+
+            Example:
+                {
+                    "my_int": {
+                        "type": "Int64",
+                        "replace": {None: 42},
+                    },
+                    "my_float": {
+                        "type": "float64",
+                        "replace": {None: np.nan, 999: -1},
+                    },
+                    "my_boolean": {
+                        "type": "boolean",
+                        "replace": {None: False},
+                    },
+                    "my_string": {
+                        "type": "string",
+                        "replace": {"This is empty": "Nothing Here"},
+                    },
+                }
+        :type typed_columns: dict, optional
         :return: Pandas DataFrame
         :rtype: Pandas DataFrame
         """
@@ -1522,4 +1555,14 @@ class KoshDataset(KoshSinaObject):
                     attr_dict[column][i] = source['data'].get(column, {}).get('value', pd.NA)
 
         df = pd.DataFrame(attr_dict)
+
+        if isinstance(typed_columns, dict):
+            for key, config in typed_columns.items():
+                series = df[key]
+
+                if "replace" in config:
+                    series = series.replace(config["replace"]).infer_objects(copy=False)
+
+                df[key] = series.astype(config["type"])
+
         return df

{kosh-3.4 → kosh-3.6}/kosh/ensemble.py

@@ -337,7 +337,7 @@ These datasets will inherit attributes and associated sources from the ensemble.
 
     @lock_strategies.lock_method
     def to_dataframe(self, data_columns=[], include_ensemble_attributes=True, include_ensemble_tags=True,
-                     *atts, **keys):
+                     typed_columns=None, *atts, **keys):
         """Return the find_datasets object as a Pandas DataFrame.
 
         Pass in the same arguments and keyword arguments as the find method.
@@ -362,6 +362,39 @@ These datasets will inherit attributes and associated sources from the ensemble.
         :type include_ensemble_attributes: bool, optional
         :param include_ensemble_tags: Include ensemble tags in DataFrame.
         :type include_ensemble_tags: bool, optional
+        :param typed_columns: Dictionary of column types and replacement values.
+            If omitted, MySQL defaults are returned.
+
+            Supported types:
+                [
+                    "Int8", "Int16", "Int32", "Int64",
+                    "UInt8", "UInt16", "UInt32", "UInt64",
+                    "float16", "float32", "float64",
+                    "Float32", "Float64",
+                    "bool", "boolean", "string", "object",
+                    "category", "datetime64[ns]", "timedelta64[ns]",
+                ]
+
+            Example:
+                {
+                    "my_int": {
+                        "type": "Int64",
+                        "replace": {None: 42},
+                    },
+                    "my_float": {
+                        "type": "float64",
+                        "replace": {None: np.nan, 999: -1},
+                    },
+                    "my_boolean": {
+                        "type": "boolean",
+                        "replace": {None: False},
+                    },
+                    "my_string": {
+                        "type": "string",
+                        "replace": {"This is empty": "Nothing Here"},
+                    },
+                }
+        :type typed_columns: dict, optional
         :return: Pandas DataFrame
         :rtype: Pandas DataFrame
         """
@@ -425,4 +458,14 @@ These datasets will inherit attributes and associated sources from the ensemble.
                 attr_dict[f"{self.id}_ENSEMBLE_ATTRIBUTE_{key}"] = [val] * total_datasets
 
         df = pd.DataFrame(attr_dict)
+
+        if isinstance(typed_columns, dict):
+            for key, config in typed_columns.items():
+                series = df[key]
+
+                if "replace" in config:
+                    series = series.replace(config["replace"]).infer_objects(copy=False)
+
+                df[key] = series.astype(config["type"])
+
         return df

kosh-3.6/kosh/mada_extension.py

@@ -0,0 +1,33 @@
+"""MADA extension manifest for Kosh-provided MCP servers."""
+
+from kosh.current_version import current_version
+
+
+def get_extension_manifest():
+    """Return the MADA extension manifest for Kosh.
+
+    :returns: The Kosh extension manifest with its MCP server registration.
+    :raises RuntimeError: If the optional ``kosh[mcp]`` dependencies are not
+        installed.
+    """
+    try:
+        from mada_tools.extensions import ExtensionManifest, MCPServerRegistration
+    except ImportError as e:
+        raise RuntimeError(
+            "Kosh MADA extension discovery requires the optional 'mcp' extra. "
+            "Install it with: pip install 'kosh[mcp]'"
+        ) from e
+
+    return ExtensionManifest(
+        display_name="Kosh",
+        version=current_version,
+        provider_package="kosh",
+        mcp_servers=(
+            MCPServerRegistration(
+                name="kosh",
+                module_path="kosh.mcp_server.server",
+                package="kosh",
+                description="MCP server for Kosh data management.",
+            ),
+        ),
+    )