deeprails 1.4.1__tar.gz → 1.5.0__tar.gz

deeprails-1.5.0/.release-please-manifest.json

@@ -0,0 +1,3 @@
+{
+  ".": "1.5.0"
+}

{deeprails-1.4.1 → deeprails-1.5.0}/CHANGELOG.md

@@ -1,5 +1,13 @@
 # Changelog
 
+## 1.5.0 (2025-10-16)
+
+Full Changelog: [v1.4.1...v1.5.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.4.1...v1.5.0)
+
+### Features
+
+* **api:** adding support for system_prompt and ground_truth ([be4cd74](https://github.com/deeprails/deeprails-sdk-python/commit/be4cd74a118dc1e4b1ef90f25f311a0862807176))
+
 ## 1.4.1 (2025-10-11)
 
 Full Changelog: [v1.4.0...v1.4.1](https://github.com/deeprails/deeprails-sdk-python/compare/v1.4.0...v1.4.1)

{deeprails-1.4.1 → deeprails-1.5.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: deeprails
-Version: 1.4.1
+Version: 1.5.0
 Summary: The official Python library for the deeprails API
 Project-URL: Homepage, https://docs.deeprails.com/
 Project-URL: Repository, https://github.com/deeprails/deeprails-sdk-python

{deeprails-1.4.1 → deeprails-1.5.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "deeprails"
-version = "1.4.1"
+version = "1.5.0"
 description = "The official Python library for the deeprails API"
 dynamic = ["readme"]
 license = "Apache-2.0"

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/_version.py

@@ -1,4 +1,4 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 __title__ = "deeprails"
-__version__ = "1.4.1"  # x-release-please-version
+__version__ = "1.5.0"  # x-release-please-version

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/resources/defend.py

@@ -213,9 +213,9 @@ class DefendResource(SyncAPIResource):
         evaluation
 
         Args:
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -491,9 +491,9 @@ class AsyncDefendResource(AsyncAPIResource):
         evaluation
 
         Args:
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/resources/evaluate.py

@@ -75,9 +75,9 @@ class EvaluateResource(SyncAPIResource):
         guardrail metrics
 
         Args:
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -206,9 +206,9 @@ class AsyncEvaluateResource(AsyncAPIResource):
         guardrail metrics
 
         Args:
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/resources/monitor.py

@@ -219,9 +219,9 @@ class MonitorResource(SyncAPIResource):
               `completeness`, `instruction_adherence`, `context_adherence`,
               `ground_truth_adherence`, and/or `comprehensive_safety`.
 
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -457,9 +457,9 @@ class AsyncMonitorResource(AsyncAPIResource):
               `completeness`, `instruction_adherence`, `context_adherence`,
               `ground_truth_adherence`, and/or `comprehensive_safety`.
 
-          model_input: A dictionary of inputs sent to the LLM to generate output. This must contain a
-              `user_prompt` field and an optional `context` field. Additional properties are
-              allowed.
+          model_input: A dictionary of inputs sent to the LLM to generate output. The dictionary must
+              contain at least one of `user_prompt` or `system_prompt`. For
+              ground_truth_aherence guadrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/types/defend_submit_event_params.py

@@ -2,8 +2,7 @@
 
 from __future__ import annotations
 
-from typing import Dict, Union
-from typing_extensions import Literal, Required, TypeAlias, TypedDict
+from typing_extensions import Literal, Required, TypedDict
 
 __all__ = ["DefendSubmitEventParams", "ModelInput"]
 
@@ -12,8 +11,8 @@ class DefendSubmitEventParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """A dictionary of inputs sent to the LLM to generate output.
 
-    This must contain a `user_prompt` field and an optional `context` field.
-    Additional properties are allowed.
+    The dictionary must contain at least one of `user_prompt` or `system_prompt`.
+    For ground_truth_aherence guadrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]
@@ -35,10 +34,12 @@ class DefendSubmitEventParams(TypedDict, total=False):
     """An optional, user-defined tag for the event."""
 
 
-class ModelInputTyped(TypedDict, total=False):
-    user_prompt: Required[str]
+class ModelInput(TypedDict, total=False):
+    ground_truth: str
+    """The ground truth for evaluating Ground Truth Adherence guardrail."""
 
-    context: str
+    system_prompt: str
+    """The system prompt used to generate the output."""
 
-
-ModelInput: TypeAlias = Union[ModelInputTyped, Dict[str, object]]
+    user_prompt: str
+    """The user prompt used to generate the output."""

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/types/evaluate_create_params.py

@@ -2,8 +2,8 @@
 
 from __future__ import annotations
 
-from typing import Dict, List, Union
-from typing_extensions import Literal, Required, TypeAlias, TypedDict
+from typing import List
+from typing_extensions import Literal, Required, TypedDict
 
 __all__ = ["EvaluateCreateParams", "ModelInput"]
 
@@ -12,8 +12,8 @@ class EvaluateCreateParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """A dictionary of inputs sent to the LLM to generate output.
 
-    This must contain a `user_prompt` field and an optional `context` field.
-    Additional properties are allowed.
+    The dictionary must contain at least one of `user_prompt` or `system_prompt`.
+    For ground_truth_aherence guadrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]
@@ -51,10 +51,12 @@ class EvaluateCreateParams(TypedDict, total=False):
     """An optional, user-defined tag for the evaluation."""
 
 
-class ModelInputTyped(TypedDict, total=False):
-    user_prompt: Required[str]
+class ModelInput(TypedDict, total=False):
+    ground_truth: str
+    """The ground truth for evaluating Ground Truth Adherence guardrail."""
 
-    context: str
+    system_prompt: str
+    """The system prompt used to generate the output."""
 
-
-ModelInput: TypeAlias = Union[ModelInputTyped, Dict[str, object]]
+    user_prompt: str
+    """The user prompt used to generate the output."""

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/types/evaluation.py

@@ -1,6 +1,6 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
-from typing import TYPE_CHECKING, Dict, List, Optional
+from typing import Dict, List, Optional
 from datetime import datetime
 from typing_extensions import Literal
 
@@ -12,23 +12,14 @@ __all__ = ["Evaluation", "ModelInput"]
 
 
 class ModelInput(BaseModel):
-    user_prompt: str
-    """The user prompt used to generate the output."""
-
-    context: Optional[str] = None
-    """Optional context supplied to the LLM when generating the output."""
+    ground_truth: Optional[str] = None
+    """The ground truth for evaluating Ground Truth Adherence guardrail."""
 
-    if TYPE_CHECKING:
-        # Some versions of Pydantic <2.8.0 have a bug and don’t allow assigning a
-        # value to this field, so for compatibility we avoid doing it at runtime.
-        __pydantic_extra__: Dict[str, object] = FieldInfo(init=False)  # pyright: ignore[reportIncompatibleVariableOverride]
+    system_prompt: Optional[str] = None
+    """The system prompt used to generate the output."""
 
-        # Stub to indicate that arbitrary properties are accepted.
-        # To access properties that are not valid identifiers you can use `getattr`, e.g.
-        # `getattr(obj, '$type')`
-        def __getattr__(self, attr: str) -> object: ...
-    else:
-        __pydantic_extra__: Dict[str, object]
+    user_prompt: Optional[str] = None
+    """The user prompt used to generate the output."""
 
 
 class Evaluation(BaseModel):
@@ -41,8 +32,8 @@ class Evaluation(BaseModel):
     api_model_input: ModelInput = FieldInfo(alias="model_input")
     """A dictionary of inputs sent to the LLM to generate output.
 
-    The dictionary must contain a `user_prompt` field and an optional `context`
-    field. Additional properties are allowed.
+    The dictionary must contain at least one of `user_prompt` or `system_prompt`.
+    For ground_truth_aherence guadrail metric, `ground_truth` should be provided.
     """
 
     api_model_output: str = FieldInfo(alias="model_output")

{deeprails-1.4.1 → deeprails-1.5.0}/src/deeprails/types/monitor_submit_event_params.py

@@ -2,8 +2,8 @@
 
 from __future__ import annotations
 
-from typing import Dict, List, Union
-from typing_extensions import Literal, Required, TypeAlias, TypedDict
+from typing import List
+from typing_extensions import Literal, Required, TypedDict
 
 __all__ = ["MonitorSubmitEventParams", "ModelInput"]
 
@@ -31,8 +31,8 @@ class MonitorSubmitEventParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """A dictionary of inputs sent to the LLM to generate output.
 
-    This must contain a `user_prompt` field and an optional `context` field.
-    Additional properties are allowed.
+    The dictionary must contain at least one of `user_prompt` or `system_prompt`.
+    For ground_truth_aherence guadrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]
@@ -54,10 +54,12 @@ class MonitorSubmitEventParams(TypedDict, total=False):
     """
 
 
-class ModelInputTyped(TypedDict, total=False):
-    user_prompt: Required[str]
+class ModelInput(TypedDict, total=False):
+    ground_truth: str
+    """The ground truth for evaluating Ground Truth Adherence guardrail."""
 
-    context: str
+    system_prompt: str
+    """The system prompt used to generate the output."""
 
-
-ModelInput: TypeAlias = Union[ModelInputTyped, Dict[str, object]]
+    user_prompt: str
+    """The user prompt used to generate the output."""

{deeprails-1.4.1 → deeprails-1.5.0}/tests/api_resources/test_defend.py

@@ -176,7 +176,7 @@ class TestDefend:
     def test_method_submit_event(self, client: Deeprails) -> None:
         defend = client.defend.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -189,8 +189,9 @@ class TestDefend:
         defend = client.defend.submit_event(
             workflow_id="workflow_id",
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             model_used="model_used",
@@ -204,7 +205,7 @@ class TestDefend:
     def test_raw_response_submit_event(self, client: Deeprails) -> None:
         response = client.defend.with_raw_response.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -220,7 +221,7 @@ class TestDefend:
     def test_streaming_response_submit_event(self, client: Deeprails) -> None:
         with client.defend.with_streaming_response.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -239,7 +240,7 @@ class TestDefend:
         with pytest.raises(ValueError, match=r"Expected a non-empty value for `workflow_id` but received ''"):
             client.defend.with_raw_response.submit_event(
                 workflow_id="",
-                model_input={"user_prompt": "user_prompt"},
+                model_input={},
                 model_output="model_output",
                 model_used="model_used",
                 run_mode="precision_plus",
@@ -459,7 +460,7 @@ class TestAsyncDefend:
     async def test_method_submit_event(self, async_client: AsyncDeeprails) -> None:
         defend = await async_client.defend.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -472,8 +473,9 @@ class TestAsyncDefend:
         defend = await async_client.defend.submit_event(
             workflow_id="workflow_id",
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             model_used="model_used",
@@ -487,7 +489,7 @@ class TestAsyncDefend:
     async def test_raw_response_submit_event(self, async_client: AsyncDeeprails) -> None:
         response = await async_client.defend.with_raw_response.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -503,7 +505,7 @@ class TestAsyncDefend:
     async def test_streaming_response_submit_event(self, async_client: AsyncDeeprails) -> None:
         async with async_client.defend.with_streaming_response.submit_event(
             workflow_id="workflow_id",
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             model_used="model_used",
             run_mode="precision_plus",
@@ -522,7 +524,7 @@ class TestAsyncDefend:
         with pytest.raises(ValueError, match=r"Expected a non-empty value for `workflow_id` but received ''"):
             await async_client.defend.with_raw_response.submit_event(
                 workflow_id="",
-                model_input={"user_prompt": "user_prompt"},
+                model_input={},
                 model_output="model_output",
                 model_used="model_used",
                 run_mode="precision_plus",

{deeprails-1.4.1 → deeprails-1.5.0}/tests/api_resources/test_evaluate.py

@@ -21,7 +21,7 @@ class TestEvaluate:
     @parametrize
     def test_method_create(self, client: Deeprails) -> None:
         evaluate = client.evaluate.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         )
@@ -32,8 +32,9 @@ class TestEvaluate:
     def test_method_create_with_all_params(self, client: Deeprails) -> None:
         evaluate = client.evaluate.create(
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             run_mode="precision_plus",
@@ -47,7 +48,7 @@ class TestEvaluate:
     @parametrize
     def test_raw_response_create(self, client: Deeprails) -> None:
         response = client.evaluate.with_raw_response.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         )
@@ -61,7 +62,7 @@ class TestEvaluate:
     @parametrize
     def test_streaming_response_create(self, client: Deeprails) -> None:
         with client.evaluate.with_streaming_response.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         ) as response:
@@ -125,7 +126,7 @@ class TestAsyncEvaluate:
     @parametrize
     async def test_method_create(self, async_client: AsyncDeeprails) -> None:
         evaluate = await async_client.evaluate.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         )
@@ -136,8 +137,9 @@ class TestAsyncEvaluate:
     async def test_method_create_with_all_params(self, async_client: AsyncDeeprails) -> None:
         evaluate = await async_client.evaluate.create(
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             run_mode="precision_plus",
@@ -151,7 +153,7 @@ class TestAsyncEvaluate:
     @parametrize
     async def test_raw_response_create(self, async_client: AsyncDeeprails) -> None:
         response = await async_client.evaluate.with_raw_response.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         )
@@ -165,7 +167,7 @@ class TestAsyncEvaluate:
     @parametrize
     async def test_streaming_response_create(self, async_client: AsyncDeeprails) -> None:
         async with async_client.evaluate.with_streaming_response.create(
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
             run_mode="precision_plus",
         ) as response:

{deeprails-1.4.1 → deeprails-1.5.0}/tests/api_resources/test_monitor.py

@@ -174,7 +174,7 @@ class TestMonitor:
         monitor = client.monitor.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         )
         assert_matches_type(MonitorSubmitEventResponse, monitor, path=["response"])
@@ -186,8 +186,9 @@ class TestMonitor:
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             model_used="model_used",
@@ -202,7 +203,7 @@ class TestMonitor:
         response = client.monitor.with_raw_response.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         )
 
@@ -217,7 +218,7 @@ class TestMonitor:
         with client.monitor.with_streaming_response.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         ) as response:
             assert not response.is_closed
@@ -235,7 +236,7 @@ class TestMonitor:
             client.monitor.with_raw_response.submit_event(
                 monitor_id="",
                 guardrail_metrics=["correctness"],
-                model_input={"user_prompt": "user_prompt"},
+                model_input={},
                 model_output="model_output",
             )
 
@@ -398,7 +399,7 @@ class TestAsyncMonitor:
         monitor = await async_client.monitor.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         )
         assert_matches_type(MonitorSubmitEventResponse, monitor, path=["response"])
@@ -410,8 +411,9 @@ class TestAsyncMonitor:
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
             model_input={
+                "ground_truth": "ground_truth",
+                "system_prompt": "system_prompt",
                 "user_prompt": "user_prompt",
-                "context": "context",
             },
             model_output="model_output",
             model_used="model_used",
@@ -426,7 +428,7 @@ class TestAsyncMonitor:
         response = await async_client.monitor.with_raw_response.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         )
 
@@ -441,7 +443,7 @@ class TestAsyncMonitor:
         async with async_client.monitor.with_streaming_response.submit_event(
             monitor_id="monitor_id",
             guardrail_metrics=["correctness"],
-            model_input={"user_prompt": "user_prompt"},
+            model_input={},
             model_output="model_output",
         ) as response:
             assert not response.is_closed
@@ -459,6 +461,6 @@ class TestAsyncMonitor:
             await async_client.monitor.with_raw_response.submit_event(
                 monitor_id="",
                 guardrail_metrics=["correctness"],
-                model_input={"user_prompt": "user_prompt"},
+                model_input={},
                 model_output="model_output",
             )

deeprails-1.4.1/.release-please-manifest.json

@@ -1,3 +0,0 @@
-{
-  ".": "1.4.1"
-}