Refactor metadata

.github/workflows/tests.yml

@@ -78,6 +78,7 @@ jobs:
             python-version: 3.8
         - name: Install dependencies
           run: |
+            python -m pip install --upgrade pip
             python -m pip install ".[dev]"
             python -m pip install --upgrade git+https://github.com/rstudio/vetiver-python@${{ github.sha }}
         - name: run Docker

docs/source/advancedusage/custom_handler.md

@@ -12,6 +12,7 @@ class CustomHandler(BaseHandler):
         super().__init__(model, ptype_data)
 
     model_type = staticmethod(lambda: newmodeltype)
+    pip_name = "scikit-learn" # pkg name on pip, used for tracking pkg versions
 
     def handler_predict(self, input_data, check_ptype: bool):
         """

vetiver/attach_pkgs.py

@@ -1,6 +1,7 @@
 import tempfile
 import os
-from vetiver import VetiverModel
+from .vetiver_model import VetiverModel
+from .meta import VetiverMeta
 
 
 def load_pkgs(model: VetiverModel = None, packages: list = None, path=""):
@@ -19,8 +20,12 @@ def load_pkgs(model: VetiverModel = None, packages: list = None, path=""):
     required_pkgs = ["vetiver"]
     if packages:
         required_pkgs = list(set(required_pkgs + packages))
-    if model.metadata.get("required_pkgs"):
-        required_pkgs = list(set(required_pkgs + model.metadata.get("required_pkgs")))
+
+    if isinstance(model.metadata, dict):
+        model.metadata = VetiverMeta.from_dict(model.metadata)
+
+    if model.metadata.required_pkgs:
+        required_pkgs = list(set(required_pkgs + model.metadata.required_pkgs))
 
     tmp = tempfile.NamedTemporaryFile(suffix=".in", delete=False)
     tmp.close()

vetiver/handlers/base.py

@@ -3,7 +3,7 @@
 from contextlib import suppress
 
 from ..prototype import vetiver_create_prototype
-from ..meta import _model_meta
+from ..meta import VetiverMeta
 
 
 class InvalidModelError(Exception):
@@ -43,7 +43,7 @@ def create_handler(model, prototype_data):
     >>> model = vetiver.mock.get_mock_model()
     >>> handler = vetiver.create_handler(model, X)
     >>> handler.describe()
-    "Scikit-learn <class 'sklearn.dummy.DummyRegressor'> model"
+    'A scikit-learn DummyRegressor model'
     """
 
     raise InvalidModelError(
@@ -79,19 +79,20 @@ def __init__(self, model, prototype_data):
 
     def describe(self):
         """Create description for model"""
-        desc = f"{self.model.__class__} model"
+
+        pip_name = self.pip_name if hasattr(self, "pip_name") else ""
+        obj_name = type(self.model).__qualname__
+
+        desc = f"A {pip_name} {obj_name} model"
+
         return desc
 
-    def create_meta(
-        user: list = None,
-        version: str = None,
-        url: str = None,
-        required_pkgs: list = [],
-    ):
+    def create_meta(self, metadata):
         """Create metadata for a model"""
-        meta = _model_meta(user, version, url, required_pkgs)
 
-        return meta
+        pip_name = self.pip_name if hasattr(self, "pip_name") else None
+
+        return VetiverMeta.from_dict(metadata, pip_name)
 
     def construct_prototype(self):
         """Create data prototype for a model

vetiver/handlers/sklearn.py

@@ -1,7 +1,6 @@
 import pandas as pd
 import sklearn
 
-from ..meta import _model_meta
 from .base import BaseHandler
 
 
@@ -15,23 +14,7 @@ class SKLearnHandler(BaseHandler):
     """
 
     model_class = staticmethod(lambda: sklearn.base.BaseEstimator)
-
-    def describe(self):
-        """Create description for sklearn model"""
-        desc = f"Scikit-learn {self.model.__class__} model"
-        return desc
-
-    def create_meta(
-        user: list = None,
-        version: str = None,
-        url: str = None,
-        required_pkgs: list = [],
-    ):
-        """Create metadata for sklearn model"""
-        required_pkgs = required_pkgs + ["scikit-learn"]
-        meta = _model_meta(user, version, url, required_pkgs)
-
-        return meta
+    pip_name = "scikit-learn"
 
     def handler_predict(self, input_data, check_prototype):
         """Generates method for /predict endpoint in VetiverAPI

vetiver/handlers/statsmodels.py

@@ -1,6 +1,5 @@
 import pandas as pd
 
-from ..meta import _model_meta
 from .base import BaseHandler
 
 sm_exists = True
@@ -20,23 +19,8 @@ class StatsmodelsHandler(BaseHandler):
     """
 
     model_class = staticmethod(lambda: statsmodels.base.wrapper.ResultsWrapper)
-
-    def describe(self):
-        """Create description for statsmodels model"""
-        desc = f"Statsmodels {self.model.__class__} model."
-        return desc
-
-    def create_meta(
-        user: list = None,
-        version: str = None,
-        url: str = None,
-        required_pkgs: list = [],
-    ):
-        """Create metadata for statsmodel"""
-        required_pkgs = required_pkgs + ["statsmodels"]
-        meta = _model_meta(user, version, url, required_pkgs)
-
-        return meta
+    if sm_exists:
+        pip_name = "statsmodels"
 
     def handler_predict(self, input_data, check_prototype):
         """Generates method for /predict endpoint in VetiverAPI

vetiver/handlers/torch.py

@@ -1,6 +1,5 @@
 import numpy as np
 
-from ..meta import _model_meta
 from .base import BaseHandler
 
 torch_exists = True
@@ -20,23 +19,8 @@ class TorchHandler(BaseHandler):
     """
 
     model_class = staticmethod(lambda: torch.nn.Module)
-
-    def describe(self):
-        """Create description for torch model"""
-        desc = f"Pytorch model of type {type(self.model)}"
-        return desc
-
-    def create_meta(
-        user: list = None,
-        version: str = None,
-        url: str = None,
-        required_pkgs: list = [],
-    ):
-        """Create metadata for torch model"""
-        required_pkgs = required_pkgs + ["torch"]
-        meta = _model_meta(user, version, url, required_pkgs)
-
-        return meta
+    if torch_exists:
+        pip_name = "torch"
 
     def handler_predict(self, input_data, check_prototype):
         """Generates method for /predict endpoint in VetiverAPI

vetiver/handlers/xgboost.py

@@ -1,6 +1,5 @@
 import pandas as pd
 
-from ..meta import _model_meta
 from .base import BaseHandler
 
 xgb_exists = True
@@ -20,23 +19,8 @@ class XGBoostHandler(BaseHandler):
     """
 
     model_class = staticmethod(lambda: xgboost.Booster)
-
-    def describe(self):
-        """Create description for xgboost model"""
-        desc = f"XGBoost {self.model.__class__} model."
-        return desc
-
-    def create_meta(
-        user: list = None,
-        version: str = None,
-        url: str = None,
-        required_pkgs: list = [],
-    ):
-        """Create metadata for xgboost"""
-        required_pkgs = required_pkgs + ["xgboost"]
-        meta = _model_meta(user, version, url, required_pkgs)
-
-        return meta
+    if xgb_exists:
+        pip_name = "xgboost"
 
     def handler_predict(self, input_data, check_prototype):
         """Generates method for /predict endpoint in VetiverAPI

vetiver/meta.py

@@ -1,23 +1,33 @@
-def _model_meta(
-    user: dict = None, version: str = None, url: str = None, required_pkgs: list = None
-):
-    """Populate relevant metadata for VetiverModel
-
-    Args
-    ----
-        user: dict
-           Extra user-defined information
-        version: str
-            Model version, generally populated from pins
-        url: str
-            Discoverable URL for API
-        required_pkgs: list
-            Packages necessary to make predictions
-    """
-    meta = {
-        "user": user,
-        "version": version,
-        "url": url,
-        "required_pkgs": required_pkgs,
-    }
-    return meta
+from dataclasses import dataclass, asdict, field
+from typing import Mapping
+
+
+@dataclass
+class VetiverMeta:
+    """Metadata in a VetiverModel"""
+
+    user: "dict | None" = field(default_factory=dict)
+    version: "str | None" = None
+    url: "str | None" = None
+    required_pkgs: "list | None" = field(default_factory=list)
+
+    def to_dict(self) -> Mapping:
+        data = asdict(self)
+
+        return data
+
+    @classmethod
+    def from_dict(cls, metadata, pip_name=None) -> "VetiverMeta":
+
+        metadata = {} if metadata is None else metadata
+
+        user = metadata.get("user", metadata)
+        version = metadata.get("version", None)
+        url = metadata.get("url", None)
+        required_pkgs = metadata.get("required_pkgs", [])
+
+        if pip_name:
+            if not list(filter(lambda x: pip_name in x, required_pkgs)):
+                required_pkgs = required_pkgs + [f"{pip_name}"]
+
+        return cls(user, version, url, required_pkgs)

vetiver/pin_read_write.py

@@ -1,4 +1,5 @@
 from .vetiver_model import VetiverModel
+from .meta import VetiverMeta
 from .utils import inform
 import warnings
 import logging
@@ -54,15 +55,22 @@ def vetiver_pin_write(board, model: VetiverModel, versioned: bool = True):
     # convert older model's ptype to prototype
     if hasattr(model, "ptype"):
         model.prototype = model.ptype
+        delattr(model, "ptype")
+    # metadata is dict
+    if isinstance(model.metadata, dict):
+        model.metadata = VetiverMeta.from_dict(model.metadata)
 
     board.pin_write(
         model.model,
         name=model.model_name,
         type="joblib",
         description=model.description,
         metadata={
-            "required_pkgs": model.metadata.get("required_pkgs"),
-            "prototype": None if model.prototype is None else model.prototype().json(),
+            "user": model.metadata.user,
+            "vetiver_meta": {
+                "required_pkgs": model.metadata.required_pkgs,
+                "prototype": None if not model.prototype else model.prototype().json(),
+            },
         },
         versioned=versioned,
     )

vetiver/server.py

@@ -12,6 +12,7 @@
 
 from .utils import _jupyter_nb
 from .vetiver_model import VetiverModel
+from .meta import VetiverMeta
 
 
 class VetiverAPI:
@@ -82,11 +83,14 @@ def docs_redirect():
 
             return RedirectResponse(redirect)
 
-        if self.model.metadata.get("url") is not None:
+        if isinstance(self.model.metadata, dict):
+            self.model.metadata = VetiverMeta.from_dict(self.model.metadata)
+
+        if self.model.metadata.url is not None:
 
             @app.get("/pin-url")
             def pin_url():
-                return repr(self.model.metadata.get("url"))
+                return repr(self.model.metadata.url)
 
         @app.get("/ping", include_in_schema=True)
         async def ping():

vetiver/tests/test_build_vetiver_model.py

@@ -1,6 +1,7 @@
 import sklearn
 
 import vetiver as vt
+from vetiver.meta import VetiverMeta
 from vetiver.mock import get_mock_data, get_mock_model
 
 import pandas as pd
@@ -73,9 +74,8 @@ def test_vetiver_model_basemodel_prototype():
         model=model,
         prototype_data=m,
         model_name="model",
-        versioned=None,
+        versioned=False,
         description=None,
-        metadata=None,
     )
 
     assert vt4.model == model
@@ -99,16 +99,21 @@ def test_vetiver_model_no_prototype():
 def test_vetiver_model_use_ptype():
     vt5 = vt.VetiverModel(
         model=model,
-        ptype_data=X_df,
+        prototype_data=None,
         model_name="model",
         versioned=None,
         description=None,
-        metadata=None,
+        metadata={"test": 123},
     )
 
     assert vt5.model == model
-    assert isinstance(vt5.prototype.construct(), pydantic.BaseModel)
-    assert list(vt5.prototype.__fields__.values())[0].type_ == int
+    assert vt5.prototype is None
+    assert vt5.metadata == VetiverMeta(
+        user={"test": 123},
+        version=None,
+        url=None,
+        required_pkgs=["scikit-learn"],
+    )
 
 
 def test_vetiver_model_from_pin():
@@ -119,12 +124,18 @@ def test_vetiver_model_from_pin():
         model_name="model",
         versioned=None,
         description=None,
-        metadata=None,
+        metadata={"test": 123},
     )
+
     board = pins.board_temp(allow_pickle_read=True)
     vt.vetiver_pin_write(board=board, model=v)
     v2 = vt.VetiverModel.from_pin(board, "model")
+
     assert isinstance(v2, vt.VetiverModel)
     assert isinstance(v2.model, sklearn.base.BaseEstimator)
     assert isinstance(v2.prototype.construct(), pydantic.BaseModel)
+    assert v2.metadata.user == {"test": 123}
+    assert v2.metadata.version is not None
+    assert v2.metadata.required_pkgs == ["scikit-learn"]
+
     board.pin_delete("model")