deeprails 1.6.1__tar.gz → 1.8.0__tar.gz

deeprails-1.8.0/.release-please-manifest.json

@@ -0,0 +1,3 @@
+{
+  ".": "1.8.0"
+}

{deeprails-1.6.1 → deeprails-1.8.0}/CHANGELOG.md

@@ -1,5 +1,21 @@
 # Changelog
 
+## 1.8.0 (2025-10-22)
+
+Full Changelog: [v1.7.0...v1.8.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.7.0...v1.8.0)
+
+### Features
+
+* **api:** manual updates ([4b46121](https://github.com/deeprails/deeprails-sdk-python/commit/4b461213615578ca0382f044201c8343d4e9f167))
+
+## 1.7.0 (2025-10-22)
+
+Full Changelog: [v1.6.1...v1.7.0](https://github.com/deeprails/deeprails-sdk-python/compare/v1.6.1...v1.7.0)
+
+### Features
+
+* **api:** update defend naming and data structures ([ba934de](https://github.com/deeprails/deeprails-sdk-python/commit/ba934de8812d71da159a00fa8283876e682aeb7b))
+
 ## 1.6.1 (2025-10-22)
 
 Full Changelog: [v1.6.0...v1.6.1](https://github.com/deeprails/deeprails-sdk-python/compare/v1.6.0...v1.6.1)

{deeprails-1.6.1 → deeprails-1.8.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: deeprails
-Version: 1.6.1
+Version: 1.8.0
 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
@@ -67,12 +67,12 @@ client = Deeprails(
 
 defend_response = client.defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 print(defend_response.workflow_id)
 ```
@@ -99,12 +99,12 @@ client = AsyncDeeprails(
 async def main() -> None:
     defend_response = await client.defend.create_workflow(
         improvement_action="fixit",
-        metrics={
+        name="Push Alert Workflow",
+        type="custom",
+        custom_hallucination_threshold_values={
             "completeness": 0.7,
             "instruction_adherence": 0.75,
         },
-        name="Push Alert Workflow",
-        type="custom",
     )
     print(defend_response.workflow_id)
 
@@ -140,12 +140,12 @@ async def main() -> None:
     ) as client:
         defend_response = await client.defend.create_workflow(
             improvement_action="fixit",
-            metrics={
+            name="Push Alert Workflow",
+            type="custom",
+            custom_hallucination_threshold_values={
                 "completeness": 0.7,
                 "instruction_adherence": 0.75,
             },
-            name="Push Alert Workflow",
-            type="custom",
         )
         print(defend_response.workflow_id)
 
@@ -199,12 +199,12 @@ client = Deeprails()
 try:
     client.defend.create_workflow(
         improvement_action="fixit",
-        metrics={
+        name="Push Alert Workflow",
+        type="custom",
+        custom_hallucination_threshold_values={
             "completeness": 0.7,
             "instruction_adherence": 0.75,
         },
-        name="Push Alert Workflow",
-        type="custom",
     )
 except deeprails.APIConnectionError as e:
     print("The server could not be reached")
@@ -250,12 +250,12 @@ client = Deeprails(
 # Or, configure per-request:
 client.with_options(max_retries=5).defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 ```
 
@@ -281,12 +281,12 @@ client = Deeprails(
 # Override per-request:
 client.with_options(timeout=5.0).defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 ```
 
@@ -330,12 +330,12 @@ from deeprails import Deeprails
 client = Deeprails()
 response = client.defend.with_raw_response.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 print(response.headers.get('X-My-Header'))
 
@@ -356,12 +356,12 @@ To stream the response body, use `.with_streaming_response` instead, which requi
 ```python
 with client.defend.with_streaming_response.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 ) as response:
     print(response.headers.get("X-My-Header"))
 

{deeprails-1.6.1 → deeprails-1.8.0}/README.md

@@ -32,12 +32,12 @@ client = Deeprails(
 
 defend_response = client.defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 print(defend_response.workflow_id)
 ```
@@ -64,12 +64,12 @@ client = AsyncDeeprails(
 async def main() -> None:
     defend_response = await client.defend.create_workflow(
         improvement_action="fixit",
-        metrics={
+        name="Push Alert Workflow",
+        type="custom",
+        custom_hallucination_threshold_values={
             "completeness": 0.7,
             "instruction_adherence": 0.75,
         },
-        name="Push Alert Workflow",
-        type="custom",
     )
     print(defend_response.workflow_id)
 
@@ -105,12 +105,12 @@ async def main() -> None:
     ) as client:
         defend_response = await client.defend.create_workflow(
             improvement_action="fixit",
-            metrics={
+            name="Push Alert Workflow",
+            type="custom",
+            custom_hallucination_threshold_values={
                 "completeness": 0.7,
                 "instruction_adherence": 0.75,
             },
-            name="Push Alert Workflow",
-            type="custom",
         )
         print(defend_response.workflow_id)
 
@@ -164,12 +164,12 @@ client = Deeprails()
 try:
     client.defend.create_workflow(
         improvement_action="fixit",
-        metrics={
+        name="Push Alert Workflow",
+        type="custom",
+        custom_hallucination_threshold_values={
             "completeness": 0.7,
             "instruction_adherence": 0.75,
         },
-        name="Push Alert Workflow",
-        type="custom",
     )
 except deeprails.APIConnectionError as e:
     print("The server could not be reached")
@@ -215,12 +215,12 @@ client = Deeprails(
 # Or, configure per-request:
 client.with_options(max_retries=5).defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 ```
 
@@ -246,12 +246,12 @@ client = Deeprails(
 # Override per-request:
 client.with_options(timeout=5.0).defend.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 ```
 
@@ -295,12 +295,12 @@ from deeprails import Deeprails
 client = Deeprails()
 response = client.defend.with_raw_response.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 )
 print(response.headers.get('X-My-Header'))
 
@@ -321,12 +321,12 @@ To stream the response body, use `.with_streaming_response` instead, which requi
 ```python
 with client.defend.with_streaming_response.create_workflow(
     improvement_action="fixit",
-    metrics={
+    name="Push Alert Workflow",
+    type="custom",
+    custom_hallucination_threshold_values={
         "completeness": 0.7,
         "instruction_adherence": 0.75,
     },
-    name="Push Alert Workflow",
-    type="custom",
 ) as response:
     print(response.headers.get("X-My-Header"))
 

{deeprails-1.6.1 → deeprails-1.8.0}/pyproject.toml

@@ -1,6 +1,6 @@
 [project]
 name = "deeprails"
-version = "1.6.1"
+version = "1.8.0"
 description = "The official Python library for the deeprails API"
 dynamic = ["readme"]
 license = "Apache-2.0"

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/_version.py

@@ -1,4 +1,4 @@
 # File generated from our OpenAPI spec by Stainless. See CONTRIBUTING.md for details.
 
 __title__ = "deeprails"
-__version__ = "1.6.1"  # x-release-please-version
+__version__ = "1.8.0"  # x-release-please-version

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/resources/defend.py

@@ -2,7 +2,7 @@
 
 from __future__ import annotations
 
-from typing import Dict, Optional
+from typing import Dict
 from typing_extensions import Literal
 
 import httpx
@@ -48,13 +48,13 @@ class DefendResource(SyncAPIResource):
     def create_workflow(
         self,
         *,
-        improvement_action: Optional[Literal["regenerate", "fixit"]],
-        metrics: Dict[str, float],
+        improvement_action: Literal["regen", "fixit", "do_nothing"],
         name: str,
         type: Literal["automatic", "custom"],
-        automatic_tolerance: Literal["low", "medium", "high"] | Omit = omit,
+        automatic_hallucination_tolerance_levels: Dict[str, Literal["low", "medium", "high"]] | Omit = omit,
+        custom_hallucination_threshold_values: Dict[str, float] | Omit = omit,
         description: str | Omit = omit,
-        max_retries: int | Omit = omit,
+        max_improvement_attempt: int | Omit = omit,
         # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
         # The extra values given here take precedence over values defined on the client or passed to this method.
         extra_headers: Headers | None = None,
@@ -68,16 +68,10 @@ class DefendResource(SyncAPIResource):
 
         Args:
           improvement_action: The action used to improve outputs that fail one or guardrail metrics for the
-              workflow events. May be `regenerate`, `fixit`, or null which represents “do
-              nothing”. Regenerate runs the user's input prompt with minor induced variance.
-              Fixit attempts to directly address the shortcomings of the output using the
-              guardrail failure rationale. Do nothing does not attempt any improvement.
-
-          metrics: Mapping of guardrail metrics to floating point threshold values. If the workflow
-              type is automatic, only the metric names are used (`automatic_tolerance`
-              determines thresholds). Possible metrics are `correctness`, `completeness`,
-              `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
-              `comprehensive_safety`.
+              workflow events. May be `regen`, `fixit`, or `do_nothing`. ReGen runs the user's
+              input prompt with minor induced variance. FixIt attempts to directly address the
+              shortcomings of the output using the guardrail failure rationale. Do Nothing
+              does not attempt any improvement.
 
           name: Name of the workflow.
 
@@ -87,12 +81,18 @@ class DefendResource(SyncAPIResource):
               set the threshold for each metric as a floating point number between 0.0 and
               1.0.
 
-          automatic_tolerance: Hallucination tolerance for automatic workflows; may be `low`, `medium`, or
-              `high`. Ignored if `type` is `custom`.
+          automatic_hallucination_tolerance_levels: Mapping of guardrail metrics to hallucination tolerance levels (either `low`,
+              `medium`, or `high`). Possible metrics are `completeness`,
+              `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
+              `comprehensive_safety`.
+
+          custom_hallucination_threshold_values: Mapping of guardrail metrics to floating point threshold values. Possible
+              metrics are `correctness`, `completeness`, `instruction_adherence`,
+              `context_adherence`, `ground_truth_adherence`, or `comprehensive_safety`.
 
           description: Description for the workflow.
 
-          max_retries: Max. number of improvement action retries until a given event passes the
+          max_improvement_attempt: Max. number of improvement action retries until a given event passes the
               guardrails. Defaults to 10.
 
           extra_headers: Send extra headers
@@ -108,12 +108,12 @@ class DefendResource(SyncAPIResource):
             body=maybe_transform(
                 {
                     "improvement_action": improvement_action,
-                    "metrics": metrics,
                     "name": name,
                     "type": type,
-                    "automatic_tolerance": automatic_tolerance,
+                    "automatic_hallucination_tolerance_levels": automatic_hallucination_tolerance_levels,
+                    "custom_hallucination_threshold_values": custom_hallucination_threshold_values,
                     "description": description,
-                    "max_retries": max_retries,
+                    "max_improvement_attempt": max_improvement_attempt,
                 },
                 defend_create_workflow_params.DefendCreateWorkflowParams,
             ),
@@ -214,8 +214,8 @@ class DefendResource(SyncAPIResource):
 
         Args:
           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.
+              contain at least `user_prompt` or `system_prompt` field. For the
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -326,13 +326,13 @@ class AsyncDefendResource(AsyncAPIResource):
     async def create_workflow(
         self,
         *,
-        improvement_action: Optional[Literal["regenerate", "fixit"]],
-        metrics: Dict[str, float],
+        improvement_action: Literal["regen", "fixit", "do_nothing"],
         name: str,
         type: Literal["automatic", "custom"],
-        automatic_tolerance: Literal["low", "medium", "high"] | Omit = omit,
+        automatic_hallucination_tolerance_levels: Dict[str, Literal["low", "medium", "high"]] | Omit = omit,
+        custom_hallucination_threshold_values: Dict[str, float] | Omit = omit,
         description: str | Omit = omit,
-        max_retries: int | Omit = omit,
+        max_improvement_attempt: int | Omit = omit,
         # Use the following arguments if you need to pass additional parameters to the API that aren't available via kwargs.
         # The extra values given here take precedence over values defined on the client or passed to this method.
         extra_headers: Headers | None = None,
@@ -346,16 +346,10 @@ class AsyncDefendResource(AsyncAPIResource):
 
         Args:
           improvement_action: The action used to improve outputs that fail one or guardrail metrics for the
-              workflow events. May be `regenerate`, `fixit`, or null which represents “do
-              nothing”. Regenerate runs the user's input prompt with minor induced variance.
-              Fixit attempts to directly address the shortcomings of the output using the
-              guardrail failure rationale. Do nothing does not attempt any improvement.
-
-          metrics: Mapping of guardrail metrics to floating point threshold values. If the workflow
-              type is automatic, only the metric names are used (`automatic_tolerance`
-              determines thresholds). Possible metrics are `correctness`, `completeness`,
-              `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
-              `comprehensive_safety`.
+              workflow events. May be `regen`, `fixit`, or `do_nothing`. ReGen runs the user's
+              input prompt with minor induced variance. FixIt attempts to directly address the
+              shortcomings of the output using the guardrail failure rationale. Do Nothing
+              does not attempt any improvement.
 
           name: Name of the workflow.
 
@@ -365,12 +359,18 @@ class AsyncDefendResource(AsyncAPIResource):
               set the threshold for each metric as a floating point number between 0.0 and
               1.0.
 
-          automatic_tolerance: Hallucination tolerance for automatic workflows; may be `low`, `medium`, or
-              `high`. Ignored if `type` is `custom`.
+          automatic_hallucination_tolerance_levels: Mapping of guardrail metrics to hallucination tolerance levels (either `low`,
+              `medium`, or `high`). Possible metrics are `completeness`,
+              `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
+              `comprehensive_safety`.
+
+          custom_hallucination_threshold_values: Mapping of guardrail metrics to floating point threshold values. Possible
+              metrics are `correctness`, `completeness`, `instruction_adherence`,
+              `context_adherence`, `ground_truth_adherence`, or `comprehensive_safety`.
 
           description: Description for the workflow.
 
-          max_retries: Max. number of improvement action retries until a given event passes the
+          max_improvement_attempt: Max. number of improvement action retries until a given event passes the
               guardrails. Defaults to 10.
 
           extra_headers: Send extra headers
@@ -386,12 +386,12 @@ class AsyncDefendResource(AsyncAPIResource):
             body=await async_maybe_transform(
                 {
                     "improvement_action": improvement_action,
-                    "metrics": metrics,
                     "name": name,
                     "type": type,
-                    "automatic_tolerance": automatic_tolerance,
+                    "automatic_hallucination_tolerance_levels": automatic_hallucination_tolerance_levels,
+                    "custom_hallucination_threshold_values": custom_hallucination_threshold_values,
                     "description": description,
-                    "max_retries": max_retries,
+                    "max_improvement_attempt": max_improvement_attempt,
                 },
                 defend_create_workflow_params.DefendCreateWorkflowParams,
             ),
@@ -492,8 +492,8 @@ class AsyncDefendResource(AsyncAPIResource):
 
         Args:
           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.
+              contain at least `user_prompt` or `system_prompt` field. For the
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/resources/evaluate.py

@@ -76,8 +76,8 @@ class EvaluateResource(SyncAPIResource):
 
         Args:
           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.
+              contain at least `user_prompt` or `system_prompt` field. For
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -207,8 +207,8 @@ class AsyncEvaluateResource(AsyncAPIResource):
 
         Args:
           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.
+              contain at least `user_prompt` or `system_prompt` field. For
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/resources/monitor.py

@@ -220,8 +220,8 @@ class MonitorResource(SyncAPIResource):
               `ground_truth_adherence`, and/or `comprehensive_safety`.
 
           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.
+              contain at least a `user_prompt` or `system_prompt` field. For
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 
@@ -458,8 +458,8 @@ class AsyncMonitorResource(AsyncAPIResource):
               `ground_truth_adherence`, and/or `comprehensive_safety`.
 
           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.
+              contain at least a `user_prompt` or `system_prompt` field. For
+              ground_truth_adherence guardrail metric, `ground_truth` should be provided.
 
           model_output: Output generated by the LLM to be evaluated.
 

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/defend_create_workflow_params.py

@@ -2,29 +2,20 @@
 
 from __future__ import annotations
 
-from typing import Dict, Optional
+from typing import Dict
 from typing_extensions import Literal, Required, TypedDict
 
 __all__ = ["DefendCreateWorkflowParams"]
 
 
 class DefendCreateWorkflowParams(TypedDict, total=False):
-    improvement_action: Required[Optional[Literal["regenerate", "fixit"]]]
+    improvement_action: Required[Literal["regen", "fixit", "do_nothing"]]
     """
     The action used to improve outputs that fail one or guardrail metrics for the
-    workflow events. May be `regenerate`, `fixit`, or null which represents “do
-    nothing”. Regenerate runs the user's input prompt with minor induced variance.
-    Fixit attempts to directly address the shortcomings of the output using the
-    guardrail failure rationale. Do nothing does not attempt any improvement.
-    """
-
-    metrics: Required[Dict[str, float]]
-    """Mapping of guardrail metrics to floating point threshold values.
-
-    If the workflow type is automatic, only the metric names are used
-    (`automatic_tolerance` determines thresholds). Possible metrics are
-    `correctness`, `completeness`, `instruction_adherence`, `context_adherence`,
-    `ground_truth_adherence`, or `comprehensive_safety`.
+    workflow events. May be `regen`, `fixit`, or `do_nothing`. ReGen runs the user's
+    input prompt with minor induced variance. FixIt attempts to directly address the
+    shortcomings of the output using the guardrail failure rationale. Do Nothing
+    does not attempt any improvement.
     """
 
     name: Required[str]
@@ -39,16 +30,25 @@ class DefendCreateWorkflowParams(TypedDict, total=False):
     1.0.
     """
 
-    automatic_tolerance: Literal["low", "medium", "high"]
+    automatic_hallucination_tolerance_levels: Dict[str, Literal["low", "medium", "high"]]
     """
-    Hallucination tolerance for automatic workflows; may be `low`, `medium`, or
-    `high`. Ignored if `type` is `custom`.
+    Mapping of guardrail metrics to hallucination tolerance levels (either `low`,
+    `medium`, or `high`). Possible metrics are `completeness`,
+    `instruction_adherence`, `context_adherence`, `ground_truth_adherence`, or
+    `comprehensive_safety`.
+    """
+
+    custom_hallucination_threshold_values: Dict[str, float]
+    """Mapping of guardrail metrics to floating point threshold values.
+
+    Possible metrics are `correctness`, `completeness`, `instruction_adherence`,
+    `context_adherence`, `ground_truth_adherence`, or `comprehensive_safety`.
     """
 
     description: str
     """Description for the workflow."""
 
-    max_retries: int
+    max_improvement_attempt: int
     """Max.
 
     number of improvement action retries until a given event passes the guardrails.

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/defend_response.py

@@ -22,16 +22,16 @@ class DefendResponse(BaseModel):
     description: Optional[str] = None
     """Description for the workflow."""
 
-    improvement_action: Optional[Literal["regenerate", "fixit"]] = None
+    improvement_action: Optional[Literal["regen", "fixit", "do_nothing"]] = None
     """
     The action used to improve outputs that fail one or more guardrail metrics for
-    the workflow events. May be `regenerate`, `fixit`, or null which represents “do
-    nothing”. Regenerate runs the user's input prompt with minor induced variance.
-    Fixit attempts to directly address the shortcomings of the output using the
-    guardrail failure rationale. Do nothing does not attempt any improvement.
+    the workflow events. May be `regen`, `fixit`, or `do_nothing`. ReGen runs the
+    user's input prompt with minor induced variance. FixIt attempts to directly
+    address the shortcomings of the output using the guardrail failure rationale. Do
+    Nothing does not attempt any improvement.
     """
 
-    max_retries: Optional[int] = None
+    max_improvement_attempt: Optional[int] = None
     """Max.
 
     number of improvement action retries until a given event passes the guardrails.
@@ -40,10 +40,10 @@ class DefendResponse(BaseModel):
     modified_at: Optional[datetime] = None
     """The most recent time the workflow was modified in UTC."""
 
-    status: Optional[Literal["archived", "active"]] = None
+    status: Optional[Literal["inactive", "active"]] = None
     """Status of the selected workflow.
 
-    May be `archived` or `active`. Archived workflows will not accept events.
+    May be `inactive` or `active`. Inactive workflows will not accept events.
     """
 
     success_rate: Optional[float] = None

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/defend_submit_event_params.py

@@ -11,8 +11,8 @@ class DefendSubmitEventParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """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.
+    The dictionary must contain at least `user_prompt` or `system_prompt` field. For
+    the ground_truth_adherence guardrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/evaluate_create_params.py

@@ -12,8 +12,8 @@ class EvaluateCreateParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """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.
+    The dictionary must contain at least `user_prompt` or `system_prompt` field. For
+    ground_truth_adherence guardrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/evaluation.py

@@ -32,8 +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 at least one of `user_prompt` or `system_prompt`.
-    For ground_truth_aherence guadrail metric, `ground_truth` should be provided.
+    The dictionary must contain at least `user_prompt` or `system_prompt` field. For
+    ground_truth_adherence guardrail metric, `ground_truth` should be provided.
     """
 
     api_model_output: str = FieldInfo(alias="model_output")

{deeprails-1.6.1 → deeprails-1.8.0}/src/deeprails/types/monitor_submit_event_params.py

@@ -31,8 +31,8 @@ class MonitorSubmitEventParams(TypedDict, total=False):
     model_input: Required[ModelInput]
     """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.
+    The dictionary must contain at least a `user_prompt` or `system_prompt` field.
+    For ground_truth_adherence guardrail metric, `ground_truth` should be provided.
     """
 
     model_output: Required[str]

{deeprails-1.6.1 → deeprails-1.8.0}/tests/api_resources/test_defend.py

@@ -24,10 +24,10 @@ class TestDefend:
     @parametrize
     def test_method_create_workflow(self, client: Deeprails) -> None:
         defend = client.defend.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         )
         assert_matches_type(DefendResponse, defend, path=["response"])
 
@@ -35,13 +35,12 @@ class TestDefend:
     @parametrize
     def test_method_create_workflow_with_all_params(self, client: Deeprails) -> None:
         defend = client.defend.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
-            automatic_tolerance="low",
+            automatic_hallucination_tolerance_levels={"correctness": "low"},
             description="description",
-            max_retries=0,
+            max_improvement_attempt=2,
         )
         assert_matches_type(DefendResponse, defend, path=["response"])
 
@@ -49,10 +48,10 @@ class TestDefend:
     @parametrize
     def test_raw_response_create_workflow(self, client: Deeprails) -> None:
         response = client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         )
 
         assert response.is_closed is True
@@ -64,10 +63,10 @@ class TestDefend:
     @parametrize
     def test_streaming_response_create_workflow(self, client: Deeprails) -> None:
         with client.defend.with_streaming_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         ) as response:
             assert not response.is_closed
             assert response.http_request.headers.get("X-Stainless-Lang") == "python"
@@ -308,10 +307,10 @@ class TestAsyncDefend:
     @parametrize
     async def test_method_create_workflow(self, async_client: AsyncDeeprails) -> None:
         defend = await async_client.defend.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         )
         assert_matches_type(DefendResponse, defend, path=["response"])
 
@@ -319,13 +318,12 @@ class TestAsyncDefend:
     @parametrize
     async def test_method_create_workflow_with_all_params(self, async_client: AsyncDeeprails) -> None:
         defend = await async_client.defend.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
-            automatic_tolerance="low",
+            automatic_hallucination_tolerance_levels={"correctness": "low"},
             description="description",
-            max_retries=0,
+            max_improvement_attempt=2,
         )
         assert_matches_type(DefendResponse, defend, path=["response"])
 
@@ -333,10 +331,10 @@ class TestAsyncDefend:
     @parametrize
     async def test_raw_response_create_workflow(self, async_client: AsyncDeeprails) -> None:
         response = await async_client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         )
 
         assert response.is_closed is True
@@ -348,10 +346,10 @@ class TestAsyncDefend:
     @parametrize
     async def test_streaming_response_create_workflow(self, async_client: AsyncDeeprails) -> None:
         async with async_client.defend.with_streaming_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
+            improvement_action="regen",
             name="name",
             type="automatic",
+            automatic_hallucination_tolerance_levels={"completeness": "medium"},
         ) as response:
             assert not response.is_closed
             assert response.http_request.headers.get("X-Stainless-Lang") == "python"

{deeprails-1.6.1 → deeprails-1.8.0}/tests/test_client.py

@@ -718,7 +718,7 @@ class TestDeeprails:
 
         with pytest.raises(APITimeoutError):
             client.defend.with_streaming_response.create_workflow(
-                improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+                improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
             ).__enter__()
 
         assert _get_open_connections(self.client) == 0
@@ -730,7 +730,7 @@ class TestDeeprails:
 
         with pytest.raises(APIStatusError):
             client.defend.with_streaming_response.create_workflow(
-                improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+                improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
             ).__enter__()
         assert _get_open_connections(self.client) == 0
 
@@ -761,7 +761,7 @@ class TestDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
         )
 
         assert response.retries_taken == failures_before_success
@@ -787,11 +787,7 @@ class TestDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
-            name="name",
-            type="automatic",
-            extra_headers={"x-stainless-retry-count": Omit()},
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}, extra_headers={"x-stainless-retry-count": Omit()}
         )
 
         assert len(response.http_request.headers.get_list("x-stainless-retry-count")) == 0
@@ -816,11 +812,7 @@ class TestDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
-            name="name",
-            type="automatic",
-            extra_headers={"x-stainless-retry-count": "42"},
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}, extra_headers={"x-stainless-retry-count": "42"}
         )
 
         assert response.http_request.headers.get("x-stainless-retry-count") == "42"
@@ -1553,7 +1545,7 @@ class TestAsyncDeeprails:
 
         with pytest.raises(APITimeoutError):
             await async_client.defend.with_streaming_response.create_workflow(
-                improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+                improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
             ).__aenter__()
 
         assert _get_open_connections(self.client) == 0
@@ -1567,7 +1559,7 @@ class TestAsyncDeeprails:
 
         with pytest.raises(APIStatusError):
             await async_client.defend.with_streaming_response.create_workflow(
-                improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+                improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
             ).__aenter__()
         assert _get_open_connections(self.client) == 0
 
@@ -1599,7 +1591,7 @@ class TestAsyncDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = await client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate", metrics={"foo": 0}, name="name", type="automatic"
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}
         )
 
         assert response.retries_taken == failures_before_success
@@ -1626,11 +1618,7 @@ class TestAsyncDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = await client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
-            name="name",
-            type="automatic",
-            extra_headers={"x-stainless-retry-count": Omit()},
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}, extra_headers={"x-stainless-retry-count": Omit()}
         )
 
         assert len(response.http_request.headers.get_list("x-stainless-retry-count")) == 0
@@ -1656,11 +1644,7 @@ class TestAsyncDeeprails:
         respx_mock.post("/defend").mock(side_effect=retry_handler)
 
         response = await client.defend.with_raw_response.create_workflow(
-            improvement_action="regenerate",
-            metrics={"foo": 0},
-            name="name",
-            type="automatic",
-            extra_headers={"x-stainless-retry-count": "42"},
+            improvement_action="regen", name="name", type="automatic", automatic_hallucination_tolerance_levels={"correctness": "medium"}, extra_headers={"x-stainless-retry-count": "42"}
         )
 
         assert response.http_request.headers.get("x-stainless-retry-count") == "42"

deeprails-1.6.1/.release-please-manifest.json

@@ -1,3 +0,0 @@
-{
-  ".": "1.6.1"
-}