hud-python 0.1.4__tar.gz → 0.1.5__tar.gz

{hud_python-0.1.4 → hud_python-0.1.5}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: hud-python
-Version: 0.1.4
+Version: 0.1.5
 Summary: SDK for the HUD evaluation platform.
 Project-URL: Homepage, https://github.com/Human-Data/hud-sdk
 Project-URL: Bug Tracker, https://github.com/Human-Data/hud-sdk/issues

{hud_python-0.1.4 → hud_python-0.1.5}/docs/introduction.mdx

@@ -3,7 +3,7 @@ title: "HUD SDK | Docs"
 description: "A Python SDK for interacting with HUD environments and evaluation benchmarks for browser use and computer use models."
 ---
 
-> **Alpha Release Notice**: This SDK is currently in beta status (v0.1.4). The API is still evolving and may change in future releases as we gather feedback and improve functionality.
+> **Alpha Release Notice**: This SDK is currently in beta status (v0.1.5). The API is still evolving and may change in future releases as we gather feedback and improve functionality.
 
 [Visit our homepage hud.so →](https://hud.so)
 

{hud_python-0.1.4 → hud_python-0.1.5}/hud/__init__.py

@@ -9,7 +9,7 @@ from hud.environment import Environment, EvalSet, Observation, TaskResult
 from hud.gym import Gym
 from hud.run import Run
 
-__version__ = "0.1.4"
+__version__ = "0.1.5"
 
 __all__ = [
     "Environment",

{hud_python-0.1.4 → hud_python-0.1.5}/hud/environment.py

@@ -210,7 +210,7 @@ class Environment:
             return self.adapter.adapt_list(action)
         return [self.adapter.adapt(action)]
 
-    async def evaluate(self) -> RewardResponse:
+    async def evaluate(self, final_response: str | None = None) -> RewardResponse:
         """
         Get final evaluation score.
 
@@ -220,6 +220,7 @@ class Environment:
         data = await make_request(
             method="POST",
             url=f"{settings.base_url}/evaluation/{self.id}",
+            json={"final_response": final_response},
             api_key=settings.api_key,
         )
         return RewardResponse(**data)

{hud_python-0.1.4 → hud_python-0.1.5}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "hud-python"
-version = "0.1.4"
+version = "0.1.5"
 description = "SDK for the HUD evaluation platform."
 readme = "README.md"
 requires-python = ">=3.9, <3.14"

{hud_python-0.1.4 → hud_python-0.1.5}/tests/test_import.py

@@ -4,4 +4,4 @@ from __future__ import annotations
 def test_import():
     """Test that the package can be imported."""
     import hud
-    assert hud.__version__ == "0.1.4"
+    assert hud.__version__ == "0.1.5"