Merge branch 'master' into griffin/get_calls-feature-parity

docs/docusaurus.config.ts

@@ -125,7 +125,7 @@ const config: Config = {
     // Replace with your project's social card
     image: "img/logo-large-padded.png",
     navbar: {
-      title: "Weave",
+      title: "W&B Weave",
       logo: {
         alt: "My Site Logo",
         src: "img/logo.svg",
@@ -231,7 +231,7 @@ const config: Config = {
           ],
         },
       ],
-      copyright: `Weave by W&B`,
+      copyright: `Made with ❤️ by Weights & Biases`,
     },
     prism: {
       // theme: prismThemes.nightOwl,

tests/trace/test_client_trace.py

@@ -1380,7 +1380,7 @@ def test_dataset_row_ref(client):
     d2 = weave.ref(ref.uri()).get()
 
     inner = d2.rows[0]["a"]
-    exp_ref = "weave:///shawn/test-project/object/Dataset:0xTDJ6hEmsx8Wg9H75y42bL2WgvW5l4IXjuhHcrMh7A/attr/rows/id/XfhC9dNA5D4taMvhKT4MKN2uce7F56Krsyv4Q6mvVMA/key/a"
+    exp_ref = "weave:///shawn/test-project/object/Dataset:tiRVKBWTP7LOwjBEqe79WFS7HEibm1WG8nfe94VWZBo/attr/rows/id/XfhC9dNA5D4taMvhKT4MKN2uce7F56Krsyv4Q6mvVMA/key/a"
     assert inner == 5
     assert inner.ref.uri() == exp_ref
     gotten = weave.ref(exp_ref).get()
@@ -2747,7 +2747,7 @@ def test_objects_and_keys_with_special_characters(client):
         exp_key
         == "n-a_m.e%3A%20%2F%2B_%28%29%7B%7D%7C%22%27%3C%3E%21%40%24%5E%26%2A%23%3A%2C.%5B%5D-%3D%3B~%60100"
     )
-    exp_digest = "O66Mk7g91rlAUtcGYOFR1Y2Wk94YyPXJy2UEAzDQcYM"
+    exp_digest = "iVLhViJ3vm8vMMo3Qj35mK7GiyP8jv3OJqasIGXjN0s"
 
     exp_obj_ref = f"{ref_base}/object/{exp_name}:{exp_digest}"
     assert obj.ref.uri() == exp_obj_ref

tests/trace/test_deepcopy.py

@@ -28,6 +28,7 @@ class Example(weave.Object):
         attrs={
             "name": None,
             "description": None,
+            "ref": None,
             "_class_name": "Example",
             "_bases": ["Object", "BaseModel"],
             "a": 1,

tests/trace/test_op_decorator_behaviour.py

@@ -128,7 +128,7 @@ def test_sync_method_call(client, weave_obj, py_obj):
             entity="shawn",
             project="test-project",
             name="A",
-            _digest="tGCIGNe9xznnkoJvn2i75TOocSfV7ui1vldSrIP3ZZo",
+            _digest="nzAe1JtLJFEVeEo3yX0TOYYGhh7vAOFYRentYI9ik6U",
             _extra=(),
         ),
         "a": 1,
@@ -163,7 +163,7 @@ async def test_async_method_call(client, weave_obj, py_obj):
             entity="shawn",
             project="test-project",
             name="A",
-            _digest="tGCIGNe9xznnkoJvn2i75TOocSfV7ui1vldSrIP3ZZo",
+            _digest="nzAe1JtLJFEVeEo3yX0TOYYGhh7vAOFYRentYI9ik6U",
             _extra=(),
         ),
         "a": 1,

tests/trace/test_prompt_easy.py

@@ -240,6 +240,7 @@ def test_prompt_as_pydantic_dict():
     assert prompt.as_pydantic_dict() == {
         "name": None,
         "description": None,
+        "ref": None,
         "config": {
             "model": "gpt-4o",
             "temperature": 0.8,

tests/trace/test_uri_get.py

@@ -0,0 +1,96 @@
+import pytest
+
+import weave
+from weave.flow.prompt.prompt import EasyPrompt
+from weave.trace_server.trace_server_interface import ObjectVersionFilter, ObjQueryReq
+
+
+@pytest.fixture(
+    params=[
+        "dataset",
+        "evaluation",
+        "string_prompt",
+        "messages_prompt",
+        "easy_prompt",
+    ]
+)
+def obj(request):
+    examples = [
+        {"question": "What is 2+2?", "expected": "4"},
+        {"question": "What is 3+3?", "expected": "6"},
+    ]
+
+    if request.param == "dataset":
+        return weave.Dataset(rows=examples)
+    elif request.param == "evaluation":
+        return weave.Evaluation(dataset=examples)
+    elif request.param == "string_prompt":
+        return weave.StringPrompt("Hello, world!")
+    elif request.param == "messages_prompt":
+        return weave.MessagesPrompt([{"role": "user", "content": "Hello, world!"}])
+    elif request.param == "easy_prompt":
+        return weave.EasyPrompt("Hello world!")
+
+
+def test_ref_get(client, obj):
+    ref = weave.publish(obj)
+
+    obj_cls = type(obj)
+    obj2 = obj_cls.from_uri(ref.uri())
+    obj3 = ref.get()
+    assert isinstance(obj2, obj_cls)
+    assert isinstance(obj3, obj_cls)
+
+    for field_name in obj.model_fields:
+        obj_field_val = getattr(obj, field_name)
+        obj2_field_val = getattr(obj2, field_name)
+        obj3_field_val = getattr(obj3, field_name)
+
+        # This is a special case for EasyPrompt's unique init signature where `config`
+        # represents the kwargs passed into the class itself.  Since the original object
+        # has not been published, there is no ref and the key is omitted from the first
+        # `config`.  After publishing, there is a ref so the `config` dict has an
+        # additional `ref` key.  For comparison purposes, we pop the key to ensure the
+        # rest of the config dict is the same.
+        if obj_cls is EasyPrompt and field_name == "config":
+            obj2_field_val.pop("ref")
+            obj3_field_val.pop("ref")
+
+        assert obj_field_val == obj2_field_val
+        assert obj_field_val == obj3_field_val
+
+
+@pytest.mark.asyncio
+async def test_gotten_methods(client):
+    @weave.op
+    def model(a: int) -> int:
+        return a + 1
+
+    ev = weave.Evaluation(dataset=[{"a": 1}])
+    await ev.evaluate(model)
+    ref = weave.publish(ev)
+
+    ev2 = weave.Evaluation.from_uri(ref.uri())
+    await ev2.evaluate(model)
+
+    # Ensure that the Evaluation object we get back is equivalent to the one published.
+    # If they are the same, calling evaluate again should not publish new versions of any
+    # relevant objects of ops.
+    relevant_object_ids = [
+        "model",
+        "Evaluation.evaluate",
+        "Evaluation.predict_and_score",
+        "Evaluation.summarize",
+        "Dataset",
+        "example_evaluation",
+    ]
+    # TODO: Replace with client version of this query when available
+    res = client.server.objs_query(
+        ObjQueryReq(
+            project_id=client._project_id(),
+            filter=ObjectVersionFilter(object_ids=relevant_object_ids),
+        )
+    )
+    for obj in res.objs:
+        assert obj.version_index == 0
+        assert obj.is_latest == 1

tests/trace/test_weave_client_threaded.py

@@ -37,7 +37,7 @@ def test_flask_server(flask_server):
     url = flask_server
     response = requests.get(url)
     assert response.status_code == 200
-    assert response.text == "0xTDJ6hEmsx8Wg9H75y42bL2WgvW5l4IXjuhHcrMh7A"
+    assert response.text == "tiRVKBWTP7LOwjBEqe79WFS7HEibm1WG8nfe94VWZBo"
 
 
 def test_weave_client_global_accessible_in_thread(client):

weave/flow/dataset.py

@@ -2,10 +2,12 @@
 from typing import Any
 
 from pydantic import field_validator
+from typing_extensions import Self
 
 import weave
 from weave.flow.obj import Object
-from weave.trace.vals import WeaveTable
+from weave.trace.objectify import register_object
+from weave.trace.vals import WeaveObject, WeaveTable
 
 
 def short_str(obj: Any, limit: int = 25) -> str:
@@ -15,6 +17,7 @@ def short_str(obj: Any, limit: int = 25) -> str:
     return str_val
 
 
+@register_object
 class Dataset(Object):
     """
     Dataset object with easy saving and automatic versioning
@@ -42,6 +45,15 @@ class Dataset(Object):
 
     rows: weave.Table
 
+    @classmethod
+    def from_obj(cls, obj: WeaveObject) -> Self:
+        return cls(
+            name=obj.name,
+            description=obj.description,
+            ref=obj.ref,
+            rows=obj.rows,
+        )
+
     @field_validator("rows", mode="before")
     def convert_to_table(cls, rows: Any) -> weave.Table:
         if not isinstance(rows, weave.Table):

weave/flow/eval.py

@@ -7,6 +7,7 @@
 from pydantic import PrivateAttr, model_validator
 from rich import print
 from rich.console import Console
+from typing_extensions import Self
 
 import weave
 from weave.flow import util
@@ -30,6 +31,7 @@
 from weave.trace.env import get_weave_parallelism
 from weave.trace.errors import OpCallError
 from weave.trace.isinstance import weave_isinstance
+from weave.trace.objectify import register_object
 from weave.trace.op import CallDisplayNameFunc, Op, as_op, is_op
 from weave.trace.vals import WeaveObject
 from weave.trace.weave_client import Call, get_ref
@@ -57,6 +59,7 @@ class EvaluationResults(Object):
 ScorerLike = Union[Callable, Op, Scorer]
 
 
+@register_object
 class Evaluation(Object):
     """
     Sets up an evaluation which includes a set of scorers and a dataset.
@@ -114,6 +117,19 @@ def function_to_evaluate(question: str):
     # internal attr to track whether to use the new `output` or old `model_output` key for outputs
     _output_key: Literal["output", "model_output"] = PrivateAttr("output")
 
+    @classmethod
+    def from_obj(cls, obj: WeaveObject) -> Self:
+        return cls(
+            name=obj.name,
+            description=obj.description,
+            ref=obj.ref,
+            dataset=obj.dataset,
+            scorers=obj.scorers,
+            preprocess_model_input=obj.preprocess_model_input,
+            trials=obj.trials,
+            evaluation_name=obj.evaluation_name,
+        )
+
     @model_validator(mode="after")
     def _update_display_name(self) -> "Evaluation":
         if self.evaluation_name:

weave/flow/obj.py

@@ -8,7 +8,10 @@
     ValidatorFunctionWrapHandler,
     model_validator,
 )
+from typing_extensions import Self
 
+from weave.trace import api
+from weave.trace.objectify import Objectifyable
 from weave.trace.op import ObjectRef, Op
 from weave.trace.vals import WeaveObject, pydantic_getattribute
 from weave.trace.weave_client import get_ref
@@ -38,6 +41,7 @@ def setter(self: Any, value: T) -> None:
 class Object(BaseModel):
     name: Optional[str] = None
     description: Optional[str] = None
+    ref: Optional[ObjectRef] = None
 
     # Allow Op attributes
     model_config = ConfigDict(
@@ -51,6 +55,14 @@ class Object(BaseModel):
 
     __str__ = BaseModel.__repr__
 
+    @classmethod
+    def from_uri(cls, uri: str, *, objectify: bool = True) -> Self:
+        if not isinstance(cls, Objectifyable):
+            raise NotImplementedError(
+                f"`{cls.__name__}` must implement `from_obj` to support deserialization from a URI."
+            )
+        return api.ref(uri).get(objectify=objectify)
+
     # This is a "wrap" validator meaning we can run our own logic before
     # and after the standard pydantic validation.
     @model_validator(mode="wrap")

weave/flow/prompt/prompt.py

@@ -5,17 +5,20 @@
 import textwrap
 from collections import UserList
 from pathlib import Path
-from typing import IO, Any, Optional, SupportsIndex, TypedDict, Union, overload
+from typing import IO, Any, Optional, SupportsIndex, TypedDict, Union, cast, overload
 
 from pydantic import Field
 from rich.table import Table
+from typing_extensions import Self
 
 from weave.flow.obj import Object
 from weave.flow.prompt.common import ROLE_COLORS, color_role
 from weave.trace.api import publish as weave_publish
+from weave.trace.objectify import register_object
 from weave.trace.op import op
 from weave.trace.refs import ObjectRef
 from weave.trace.rich import pydantic_util
+from weave.trace.vals import WeaveObject
 
 
 class Message(TypedDict):
@@ -76,6 +79,7 @@ def format(self, **kwargs: Any) -> Any:
         raise NotImplementedError("Subclasses must implement format()")
 
 
+@register_object
 class StringPrompt(Prompt):
     content: str = ""
 
@@ -87,13 +91,15 @@ def format(self, **kwargs: Any) -> str:
         return self.content.format(**kwargs)
 
     @classmethod
-    def from_obj(cls, obj: Any) -> "StringPrompt":
+    def from_obj(cls, obj: WeaveObject) -> Self:
         prompt = cls(content=obj.content)
         prompt.name = obj.name
         prompt.description = obj.description
+        prompt.ref = cast(ObjectRef, obj.ref)
         return prompt
 
 
+@register_object
 class MessagesPrompt(Prompt):
     messages: list[dict] = Field(default_factory=list)
 
@@ -114,13 +120,15 @@ def format(self, **kwargs: Any) -> list:
         return [self.format_message(m, **kwargs) for m in self.messages]
 
     @classmethod
-    def from_obj(cls, obj: Any) -> "MessagesPrompt":
+    def from_obj(cls, obj: WeaveObject) -> Self:
         prompt = cls(messages=obj.messages)
         prompt.name = obj.name
         prompt.description = obj.description
+        prompt.ref = cast(ObjectRef, obj.ref)
         return prompt
 
 
+@register_object
 class EasyPrompt(UserList, Prompt):
     data: list = Field(default_factory=list)
     config: dict = Field(default_factory=dict)
@@ -420,21 +428,22 @@ def as_dict(self) -> dict[str, Any]:
         }
 
     @classmethod
-    def from_obj(cls, obj: Any) -> "EasyPrompt":
+    def from_obj(cls, obj: WeaveObject) -> Self:
         messages = obj.messages if hasattr(obj, "messages") else obj.data
         messages = [dict(m) for m in messages]
         config = dict(obj.config)
         requirements = dict(obj.requirements)
         return cls(
             name=obj.name,
             description=obj.description,
+            ref=obj.ref,
             messages=messages,
             config=config,
             requirements=requirements,
         )
 
-    @staticmethod
-    def load(fp: IO) -> "EasyPrompt":
+    @classmethod
+    def load(cls, fp: IO) -> Self:
         if isinstance(fp, str):  # Common mistake
             raise TypeError(
                 "Prompt.load() takes a file-like object, not a string. Did you mean Prompt.e()?"
@@ -443,8 +452,8 @@ def load(fp: IO) -> "EasyPrompt":
         prompt = EasyPrompt(**data)
         return prompt
 
-    @staticmethod
-    def load_file(filepath: Union[str, Path]) -> "Prompt":
+    @classmethod
+    def load_file(cls, filepath: Union[str, Path]) -> Self:
         expanded_path = os.path.expanduser(str(filepath))
         with open(expanded_path) as f:
             return EasyPrompt.load(f)