sondera-harness 0.6.3__tar.gz → 0.7.1__tar.gz

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sondera-harness
-Version: 0.6.3
+Version: 0.7.1
 Summary: Sondera Harness SDK for Python - Agent governance and policy enforcement
 Author-email: Sondera AI <sdk@sondera.ai>
 License-Expression: MIT
@@ -27,7 +27,6 @@ Requires-Dist: cedar-python>=0.1.1
 Requires-Dist: click>=8.0.0
 Requires-Dist: click-default-group>=1.2.4
 Requires-Dist: grpcio>=1.76.0
-Requires-Dist: grpcio-tools>=1.76.0
 Requires-Dist: httpx>=0.27.0
 Requires-Dist: pydantic>=2.12.0
 Requires-Dist: pydantic-settings>=2.12.0
@@ -47,11 +46,6 @@ Requires-Dist: strands-agents>=1.21.0; extra == "all"
 Dynamic: license-file
 
 <div align="center">
-  <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-dark.svg">
-    <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg">
-    <img alt="Sondera" src="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg" height="60">
-  </picture>
 
   <h1>Sondera Harness</h1>
 
@@ -66,7 +60,7 @@ Dynamic: license-file
     ·
     <a href="https://github.com/sondera-ai/sondera-harness-python/tree/main/examples">Examples</a>
     ·
-    <a href="https://discord.gg/8zMbcnDnZs">Discord</a>
+    <a href="https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw">Slack</a>
   </p>
 
   <p>
@@ -94,12 +88,16 @@ This policy stops your agent from running `rm -rf`, every time.
 
 ## Quickstart
 
+> **Try it now:** [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/sondera-ai/sondera-harness-python/blob/main/docs/src/notebooks/quickstart.ipynb) - no install required.
+
 ### 1. Install
 
 ```bash
 uv add "sondera-harness[langgraph]"   # or: pip install "sondera-harness[langgraph]"
 ```
 
+Works with [LangChain/LangGraph](https://docs.sondera.ai/integrations/langgraph/), [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom agents](https://docs.sondera.ai/integrations/custom/).
+
 ### 2. Add to Your Agent (LangGraph)
 
 ```python
@@ -132,8 +130,6 @@ agent = create_agent(
 )
 ```
 
-Also supports [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom integrations](https://docs.sondera.ai/integrations/custom/).
-
 > [!NOTE]
 > This example uses Sondera Platform ([free account](https://sondera.ai)), which also enables the TUI below. For local-only development, see [CedarPolicyHarness](https://docs.sondera.ai/integrations/custom/).
 
@@ -163,7 +159,7 @@ uv run sondera   # or: sondera (if installed via pip)
 
 ## Community
 
-- [Discord](https://discord.gg/8zMbcnDnZs) for questions and feedback
+- [Slack](https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw) for questions and feedback
 - [GitHub Issues](https://github.com/sondera-ai/sondera-harness-python/issues) for bugs
 - [Contributing](CONTRIBUTING.md) for development setup
 

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/README.md

@@ -1,9 +1,4 @@
 <div align="center">
-  <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-dark.svg">
-    <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg">
-    <img alt="Sondera" src="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg" height="60">
-  </picture>
 
   <h1>Sondera Harness</h1>
 
@@ -18,7 +13,7 @@
     ·
     <a href="https://github.com/sondera-ai/sondera-harness-python/tree/main/examples">Examples</a>
     ·
-    <a href="https://discord.gg/8zMbcnDnZs">Discord</a>
+    <a href="https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw">Slack</a>
   </p>
 
   <p>
@@ -46,12 +41,16 @@ This policy stops your agent from running `rm -rf`, every time.
 
 ## Quickstart
 
+> **Try it now:** [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/sondera-ai/sondera-harness-python/blob/main/docs/src/notebooks/quickstart.ipynb) - no install required.
+
 ### 1. Install
 
 ```bash
 uv add "sondera-harness[langgraph]"   # or: pip install "sondera-harness[langgraph]"
 ```
 
+Works with [LangChain/LangGraph](https://docs.sondera.ai/integrations/langgraph/), [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom agents](https://docs.sondera.ai/integrations/custom/).
+
 ### 2. Add to Your Agent (LangGraph)
 
 ```python
@@ -84,8 +83,6 @@ agent = create_agent(
 )
 ```
 
-Also supports [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom integrations](https://docs.sondera.ai/integrations/custom/).
-
 > [!NOTE]
 > This example uses Sondera Platform ([free account](https://sondera.ai)), which also enables the TUI below. For local-only development, see [CedarPolicyHarness](https://docs.sondera.ai/integrations/custom/).
 
@@ -115,7 +112,7 @@ uv run sondera   # or: sondera (if installed via pip)
 
 ## Community
 
-- [Discord](https://discord.gg/8zMbcnDnZs) for questions and feedback
+- [Slack](https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw) for questions and feedback
 - [GitHub Issues](https://github.com/sondera-ai/sondera-harness-python/issues) for bugs
 - [Contributing](CONTRIBUTING.md) for development setup
 

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "sondera-harness"
-version = "0.6.3"
+version = "0.7.1"
 description = "Sondera Harness SDK for Python - Agent governance and policy enforcement"
 readme = "README.md"
 requires-python = ">=3.12,<3.15"
@@ -30,7 +30,6 @@ dependencies = [
     "click>=8.0.0",
     "click-default-group>=1.2.4",
     "grpcio>=1.76.0",
-    "grpcio-tools>=1.76.0",
     "httpx>=0.27.0",
     "pydantic>=2.12.0",
     "pydantic-settings>=2.12.0",
@@ -71,6 +70,7 @@ dev = [
     "textual-dev>=1.8.0",
     "ruff>=0.8.0",
     "pyright",
+    "grpcio-tools>=1.76.0",
 ]
 examples = [
     "sondera-examples-adk",

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/__init__.py

@@ -68,8 +68,8 @@ from sondera.types import (
     Content,
     Decision,
     Parameter,
-    PolicyAnnotation,
     PolicyEngineMode,
+    PolicyMetadata,
     PromptContent,
     Role,
     SourceCode,
@@ -112,7 +112,7 @@ __all__ = [
     "AdjudicatedStep",
     "AdjudicatedTrajectory",
     "AdjudicationRecord",
-    "PolicyAnnotation",
+    "PolicyMetadata",
     "Decision",
     # Exceptions
     "SonderaError",

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/adk/plugin.py

@@ -23,6 +23,7 @@ from google.genai import types as genai_types
 from sondera.adk.analyze import format
 from sondera.harness import Harness
 from sondera.types import (
+    Decision,
     PromptContent,
     Role,
     Stage,
@@ -135,7 +136,7 @@ class SonderaHarnessPlugin(BasePlugin):
             f"[SonderaHarness] User message adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             return genai_types.Content(
                 parts=[genai_types.Part(text=adjudication.reason)]
             )
@@ -212,7 +213,7 @@ class SonderaHarnessPlugin(BasePlugin):
             f"[SonderaHarness] Before model adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             return LlmResponse(
                 content=genai_types.Content(
                     parts=[genai_types.Part(text=adjudication.reason)]
@@ -254,7 +255,7 @@ class SonderaHarnessPlugin(BasePlugin):
             f"[SonderaHarness] After model adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             return LlmResponse(
                 content=genai_types.Content(
                     parts=[genai_types.Part(text=adjudication.reason)]
@@ -296,7 +297,7 @@ class SonderaHarnessPlugin(BasePlugin):
             f"[SonderaHarness] Before tool adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             return {"error": f"Tool blocked: {adjudication.reason}"}
         return None
 
@@ -332,7 +333,7 @@ class SonderaHarnessPlugin(BasePlugin):
             f"[SonderaHarness] After tool adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             return {"error": f"Tool result blocked: {adjudication.reason}"}
         return None
 

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/harness/cedar/harness.py

@@ -15,6 +15,7 @@ from cedar import (
     EntityUid,
     PolicySet,
     Request,
+    Response,
     Schema,
 )
 from sondera.harness.abc import Harness as AbstractHarness
@@ -23,7 +24,7 @@ from sondera.types import (
     Agent,
     Content,
     Decision,
-    PolicyAnnotation,
+    PolicyMetadata,
     PromptContent,
     Role,
     Stage,
@@ -95,7 +96,6 @@ class CedarPolicyHarness(AbstractHarness):
         if policy_set is None:
             raise ValueError("policy_set is required")
 
-        self._cedar_schema = schema
         # Exclude None values when serializing to JSON for Cedar compatibility
         self._schema = Schema.from_json(schema.model_dump_json(exclude_none=True))
 
@@ -105,10 +105,18 @@ class CedarPolicyHarness(AbstractHarness):
         else:
             self._policy_set = policy_set
 
+        seen_ids: set[str] = set()
         for policy in self._policy_set.policies():
             annotations = policy.annotations()
+            if "id" not in annotations:
+                raise ValueError(
+                    f"Policy '{policy.id()}' is missing required @id annotation."
+                )
+            policy_id = annotations["id"]
+            if policy_id in seen_ids:
+                self._logger.warning(f"Duplicate policy @id: '{policy_id}'")
+            seen_ids.add(policy_id)
             if "escalate" in annotations and str(policy.effect()) != "Forbid":
-                policy_id = annotations.get("id", policy.id())
                 raise ValueError(
                     f"Policy '{policy_id}' has @escalate but is not a forbid policy. "
                     "@escalate is only valid on forbid policies."
@@ -270,50 +278,65 @@ class CedarPolicyHarness(AbstractHarness):
                     decision=Decision.ALLOW,
                     reason="Non-tool content allowed by default",
                 )
-        assert request is not None, "Unexpected none request"
         response = self._authorizer.is_authorized(request, self._policy_set)
-        if str(response.decision) == "Allow":
-            return Adjudication(
-                decision=Decision.ALLOW,
-                reason=f"Allowed by policies: {response.reason}",
-            )
+        return self._convert_cedar_response_to_adjudication(response)
 
-        annotations: list[PolicyAnnotation] = []
-        hard_deny_ids = []
+    def _convert_cedar_response_to_adjudication(
+        self, response: Response
+    ) -> Adjudication:
+        escalate_policies = []
+        non_escalate_policies = []
         for internal_id in response.reason:
             policy = self._policy_set.policy(internal_id)
             if policy is None:
                 raise RuntimeError(f"Policy '{internal_id}' not found in policy set")
-            policy_annotations = policy.annotations()
-            if "escalate" not in policy_annotations:
-                hard_deny_ids.append(internal_id)
+            cedar_policy_annotations = policy.annotations()
+            non_default_annotations = {
+                k: v
+                for k, v in cedar_policy_annotations.items()
+                if k not in ("id", "reason", "escalate")
+            }
+            is_escalate = "escalate" in cedar_policy_annotations
+            policy_metadata = PolicyMetadata(
+                id=cedar_policy_annotations["id"],
+                description=cedar_policy_annotations.get("reason", ""),
+                escalate=is_escalate,
+                escalate_arg=cedar_policy_annotations.get("escalate", ""),
+                custom=non_default_annotations,
+            )
+            if is_escalate:
+                escalate_policies.append(policy_metadata)
             else:
-                custom = {
-                    k: v
-                    for k, v in policy_annotations.items()
-                    if k not in ("id", "reason", "escalate")
-                }
-                annotations.append(
-                    PolicyAnnotation(
-                        id=policy_annotations.get("id", internal_id),
-                        description=policy_annotations.get("reason", ""),
-                        escalate=True,
-                        escalate_arg=policy_annotations["escalate"],
-                        custom=custom,
-                    )
-                )
+                non_escalate_policies.append(policy_metadata)
 
-        if not hard_deny_ids and annotations:
+        if str(response.decision) == "Allow":
+            # The @escalate annotation is only valid for `forbid` Cedar policies, so we can
+            # just return the non-escalate policies here:
             return Adjudication(
-                decision=Decision.ESCALATE,
-                reason=f"Escalated by policies: {response.reason}",
-                annotations=annotations,
+                decision=Decision.ALLOW,
+                reason="Allowed by all policies",
+                policies=non_escalate_policies,
             )
-        else:
+        # At this point we know the Cedar decision was DENY, so we just need to figure out if the
+        # final decision should be a hard DENY or else ESCALATE.
+        if non_escalate_policies:
             return Adjudication(
                 decision=Decision.DENY,
-                reason=f"Denied by policies: {hard_deny_ids}",
+                reason="Denied by policies",
+                policies=non_escalate_policies,
             )
+        if escalate_policies:
+            return Adjudication(
+                decision=Decision.ESCALATE,
+                reason="Escalated by policies",
+                policies=escalate_policies,
+            )
+        # Default deny because no policies matched (neither permit nor forbid/escalate)
+        return Adjudication(
+            decision=Decision.DENY,
+            reason="No matching permit policy",
+            policies=[],
+        )
 
     def _message_request(
         self,

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/harness/sondera/_grpc.py

@@ -16,8 +16,8 @@ from sondera.types import (
     Decision,
     GuardrailContext,
     Parameter,
-    PolicyAnnotation,
     PolicyEngineMode,
+    PolicyMetadata,
     PromptContent,
     Role,
     SourceCode,
@@ -127,9 +127,9 @@ def _convert_pb_adjudication_to_sdk(
         primitives_pb2.DECISION_ESCALATE: Decision.ESCALATE,
     }
 
-    # Convert annotations
-    annotations = [
-        PolicyAnnotation(
+    # Convert annotations to PolicyMetadata
+    policies = [
+        PolicyMetadata(
             id=ann.id if ann.HasField("id") else "",
             description=ann.description if ann.HasField("description") else "",
             custom=dict(ann.custom),
@@ -140,8 +140,7 @@ def _convert_pb_adjudication_to_sdk(
     return Adjudication(
         decision=decision_map[adjudication.decision],
         reason=adjudication.reason,
-        policy_ids=list(adjudication.policy_ids),
-        annotations=annotations,
+        policies=policies,
     )
 
 

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/harness/sondera/harness.py

@@ -562,30 +562,6 @@ class SonderaRemoteHarness(AbstractHarness):
             )
             raise
 
-    async def _list_trajectory_steps(
-        self, trajectory_id: str
-    ) -> list[primitives_pb2.AdjudicatedStep]:
-        """List trajectory steps internally."""
-        request = harness_pb2.GetTrajectoryRequest(
-            trajectory_id=trajectory_id,
-        )
-        await self._ensure_connected()
-        assert self._stub is not None, "Client not connected"
-
-        # Inject organization_id and auth metadata
-        metadata = self._get_metadata()
-
-        try:
-            response = await self._stub.GetTrajectory(request, metadata=metadata)
-            return list(response.steps)
-        except grpc.aio.AioRpcError as e:
-            if e.code() == grpc.StatusCode.NOT_FOUND:
-                return []
-            logging.error(
-                f"Failed to list trajectory steps for {trajectory_id}: {e.code()} - {e.details()}"
-            )
-            raise
-
     async def _list_agents(
         self,
         provider_id: str | None = None,

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/langgraph/middleware.py

@@ -28,6 +28,7 @@ except ImportError:
 
 from sondera.harness import Harness
 from sondera.types import (
+    Decision,
     PromptContent,
     Role,
     Stage,
@@ -173,7 +174,7 @@ class SonderaHarnessMiddleware(AgentMiddleware[State]):
             f"[SonderaHarness] Before Agent Adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if adjudication.is_denied:
+        if adjudication.decision == Decision.DENY:
             self._log.warning(
                 f"[SonderaHarness] Policy violation detected (strategy={self._strategy.value}): "
                 f"{adjudication.reason}"
@@ -226,7 +227,7 @@ class SonderaHarnessMiddleware(AgentMiddleware[State]):
                 PromptContent(text=_message_to_text(request.messages[-1])),
             )
 
-            if pre_adjudication.is_denied:
+            if pre_adjudication.decision == Decision.DENY:
                 _LOGGER.warning(
                     f"[SonderaHarness] Pre-model policy violation (strategy={self._strategy.value}): "
                     f"{pre_adjudication.reason}"
@@ -259,7 +260,7 @@ class SonderaHarnessMiddleware(AgentMiddleware[State]):
                 self._log.info(
                     f"[SonderaHarness] Post-model Adjudication for trajectory {self._harness.trajectory_id}"
                 )
-                if post_adjudication.is_denied:
+                if post_adjudication.decision == Decision.DENY:
                     self._log.warning(
                         f"[SonderaHarness] Post-model policy violation (strategy={self._strategy.value}): "
                         f"{post_adjudication.reason}"
@@ -324,7 +325,7 @@ class SonderaHarnessMiddleware(AgentMiddleware[State]):
             f"[SonderaHarness] Before Tool Adjudication for trajectory {self._harness.trajectory_id}"
         )
 
-        if pre_adjudication.is_denied:
+        if pre_adjudication.decision == Decision.DENY:
             self._log.warning(
                 f"[SonderaHarness] Pre-tool policy violation for {tool_name} "
                 f"(strategy={self._strategy.value}): {pre_adjudication.reason}"
@@ -367,7 +368,7 @@ class SonderaHarnessMiddleware(AgentMiddleware[State]):
                 f"[SonderaHarness] After Tool Adjudication for trajectory {self._harness.trajectory_id}"
             )
 
-            if post_adjudication.is_denied:
+            if post_adjudication.decision == Decision.DENY:
                 self._log.warning(
                     f"[SonderaHarness] Post-tool policy violation for {tool_name} "
                     f"(strategy={self._strategy.value}): {post_adjudication.reason}"

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/strands/harness.py

@@ -16,6 +16,7 @@ from strands.hooks.events import (
 from sondera.harness import Harness
 from sondera.strands.analyze import format_strands_agent
 from sondera.types import (
+    Decision,
     PromptContent,
     Role,
     Stage,
@@ -163,7 +164,7 @@ class SonderaHarnessHook(HookProvider):
                 f"[SonderaHarness] Before model adjudication for trajectory {self._harness.trajectory_id}"
             )
 
-            if adjudication.is_denied:
+            if adjudication.decision == Decision.DENY:
                 self._log.warning(
                     f"[SonderaHarness] Model call blocked: {adjudication.reason}"
                 )
@@ -192,7 +193,7 @@ class SonderaHarnessHook(HookProvider):
                 f"[SonderaHarness] After model adjudication for trajectory {self._harness.trajectory_id}"
             )
 
-            if adjudication.is_denied:
+            if adjudication.decision == Decision.DENY:
                 self._log.warning(
                     f"[SonderaHarness] Model response blocked: {adjudication.reason}"
                 )
@@ -231,7 +232,7 @@ class SonderaHarnessHook(HookProvider):
                 f"[SonderaHarness] Before tool adjudication for trajectory {self._harness.trajectory_id}"
             )
 
-            if adjudication.is_denied:
+            if adjudication.decision == Decision.DENY:
                 # Cancel the tool call using Strands' cancel_tool mechanism
                 event.cancel_tool = f"Tool blocked by policy: {adjudication.reason}"
                 self._log.warning(
@@ -262,7 +263,7 @@ class SonderaHarnessHook(HookProvider):
                 f"[SonderaHarness] After tool adjudication for trajectory {self._harness.trajectory_id}"
             )
 
-            if adjudication.is_denied:
+            if adjudication.decision == Decision.DENY:
                 # Modify the result to indicate policy violation
                 event.result = {
                     "content": [

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/tui/widgets/violation_panel.py

@@ -53,20 +53,22 @@ class ViolationPanel(Widget):
         with Container(classes="reason-container"):
             yield Markdown(record.adjudication.reason, classes="reason-text")
 
-        # Annotations section
-        annotations = record.adjudication.annotations
-        if annotations:
-            yield Static("[bold]Policy Annotations[/bold]", classes="section-header")
+        # Policies section
+        policies = record.adjudication.policies
+        if policies:
+            yield Static("[bold]Policies[/bold]", classes="section-header")
             with VerticalScroll(classes="annotations-container"):
-                for ann in annotations:
+                for policy in policies:
                     with Container(classes="annotation-card"):
-                        yield Static(f"[bold]{ann.id}[/bold]", classes="annotation-id")
-                        if ann.description:
+                        yield Static(
+                            f"[bold]{policy.id}[/bold]", classes="annotation-id"
+                        )
+                        if policy.description:
                             yield Static(
-                                ann.description, classes="annotation-description"
+                                policy.description, classes="annotation-description"
                             )
-                        if ann.custom:
+                        if policy.custom:
                             with Grid(classes="annotation-custom-grid"):
-                                for key, value in ann.custom.items():
+                                for key, value in policy.custom.items():
                                     yield Static(f"{key}:", classes="label")
                                     yield Static(value, classes="value")

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/tui/widgets/violations_list.py

@@ -44,15 +44,13 @@ class ViolationsList(Widget):
             reason = record.adjudication.reason
             reason_display = reason[:40] + "..." if len(reason) > 43 else reason
 
-            # Format annotations as comma-separated policy IDs
-            annotations = record.adjudication.annotations
-            if annotations:
-                ann_ids = [ann.id for ann in annotations]
-                annotations_display = ", ".join(ann_ids)
-                if len(annotations_display) > 30:
-                    annotations_display = annotations_display[:27] + "..."
+            # Format policies as comma-separated policy IDs
+            if record.adjudication.policies:
+                policies_display = ", ".join(p.id for p in record.adjudication.policies)
+                if len(policies_display) > 30:
+                    policies_display = policies_display[:27] + "..."
             else:
-                annotations_display = "-"
+                policies_display = "-"
 
             table.add_row(
                 decision_display,
@@ -66,7 +64,7 @@ class ViolationsList(Widget):
                 if len(record.step_id) > 11
                 else record.step_id,
                 reason_display,
-                annotations_display,
+                policies_display,
             )
 
     def get_selected_adjudication(self) -> AdjudicationRecord | None:

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera/types.py

@@ -231,19 +231,19 @@ class GuardrailContext(Model):
     """Map of check name to check result."""
 
 
-class PolicyAnnotation(Model):
-    """Annotation from a policy evaluation."""
+class PolicyMetadata(Model):
+    """Metadata about a policy that contributed to an adjudication decision."""
 
     id: str
-    """Unique identifier of the policy that produced this annotation."""
+    """Unique identifier of the policy."""
     description: str
-    """Human-readable description of why this annotation was added."""
+    """Human-readable description from the policy's @reason annotation."""
     escalate: bool = False
     """Whether this policy requires escalation to a human or other oracle to decide the final verdict."""
     escalate_arg: str = ""
     """The argument passed to @escalate, if any."""
     custom: dict[str, str] = Field(default_factory=dict)
-    """Custom key-value metadata from the policy."""
+    """Custom key-value metadata from the policy's annotations."""
 
 
 class Adjudication(Model):
@@ -253,25 +253,9 @@ class Adjudication(Model):
     """Whether the input is allowed."""
     reason: str
     """Reason for the adjudication decision."""
-    policy_ids: list[str] = Field(default_factory=list)
-    """IDs of policies that contributed to this decision."""
-    annotations: list[PolicyAnnotation] = Field(default_factory=list)
-    """Annotations from policy evaluations."""
-
-    @property
-    def is_denied(self) -> bool:
-        """Check if is denied."""
-        return self.decision == Decision.DENY
-
-    @property
-    def is_allowed(self) -> bool:
-        """Check if allowed."""
-        return self.decision == Decision.ALLOW
-
-    @property
-    def is_escalated(self) -> bool:
-        """Check if result requires escalation."""
-        return self.decision == Decision.ESCALATE
+    policies: list[PolicyMetadata] = Field(default_factory=list)
+    """Policies that determined this decision. Each entry contains the policy's
+    id, description (from @reason annotation), escalation info, and custom metadata."""
 
 
 class AdjudicatedStep(Model):
@@ -286,27 +270,6 @@ class AdjudicatedStep(Model):
     guardrails: GuardrailContext | None = None
     """Guardrail check results for this step."""
 
-    @property
-    def is_denied(self) -> bool:
-        """Check if result is denied."""
-        return (
-            self.adjudication.decision == Decision.DENY
-            and self.mode == PolicyEngineMode.GOVERN
-        )
-
-    @property
-    def is_allowed(self) -> bool:
-        """Check if result is allowed."""
-        return self.adjudication.decision == Decision.ALLOW
-
-    @property
-    def is_escalated(self) -> bool:
-        """Check if result requires escalation."""
-        return (
-            self.adjudication.decision == Decision.ESCALATE
-            and self.mode == PolicyEngineMode.GOVERN
-        )
-
     @property
     def message(self) -> str:
         """Get the adjudication reason in a friendly format."""

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera_harness.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: sondera-harness
-Version: 0.6.3
+Version: 0.7.1
 Summary: Sondera Harness SDK for Python - Agent governance and policy enforcement
 Author-email: Sondera AI <sdk@sondera.ai>
 License-Expression: MIT
@@ -27,7 +27,6 @@ Requires-Dist: cedar-python>=0.1.1
 Requires-Dist: click>=8.0.0
 Requires-Dist: click-default-group>=1.2.4
 Requires-Dist: grpcio>=1.76.0
-Requires-Dist: grpcio-tools>=1.76.0
 Requires-Dist: httpx>=0.27.0
 Requires-Dist: pydantic>=2.12.0
 Requires-Dist: pydantic-settings>=2.12.0
@@ -47,11 +46,6 @@ Requires-Dist: strands-agents>=1.21.0; extra == "all"
 Dynamic: license-file
 
 <div align="center">
-  <picture>
-    <source media="(prefers-color-scheme: dark)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-dark.svg">
-    <source media="(prefers-color-scheme: light)" srcset="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg">
-    <img alt="Sondera" src="https://raw.githubusercontent.com/sondera-ai/harness-sdk-python/main/assets/sondera-logo-light.svg" height="60">
-  </picture>
 
   <h1>Sondera Harness</h1>
 
@@ -66,7 +60,7 @@ Dynamic: license-file
     ·
     <a href="https://github.com/sondera-ai/sondera-harness-python/tree/main/examples">Examples</a>
     ·
-    <a href="https://discord.gg/8zMbcnDnZs">Discord</a>
+    <a href="https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw">Slack</a>
   </p>
 
   <p>
@@ -94,12 +88,16 @@ This policy stops your agent from running `rm -rf`, every time.
 
 ## Quickstart
 
+> **Try it now:** [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/sondera-ai/sondera-harness-python/blob/main/docs/src/notebooks/quickstart.ipynb) - no install required.
+
 ### 1. Install
 
 ```bash
 uv add "sondera-harness[langgraph]"   # or: pip install "sondera-harness[langgraph]"
 ```
 
+Works with [LangChain/LangGraph](https://docs.sondera.ai/integrations/langgraph/), [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom agents](https://docs.sondera.ai/integrations/custom/).
+
 ### 2. Add to Your Agent (LangGraph)
 
 ```python
@@ -132,8 +130,6 @@ agent = create_agent(
 )
 ```
 
-Also supports [Google ADK](https://docs.sondera.ai/integrations/adk/), [Strands](https://docs.sondera.ai/integrations/strands/), and [custom integrations](https://docs.sondera.ai/integrations/custom/).
-
 > [!NOTE]
 > This example uses Sondera Platform ([free account](https://sondera.ai)), which also enables the TUI below. For local-only development, see [CedarPolicyHarness](https://docs.sondera.ai/integrations/custom/).
 
@@ -163,7 +159,7 @@ uv run sondera   # or: sondera (if installed via pip)
 
 ## Community
 
-- [Discord](https://discord.gg/8zMbcnDnZs) for questions and feedback
+- [Slack](https://join.slack.com/t/sonderacommunity/shared_invite/zt-3onw10qhj-5UNQ7EMuAbPk0nTwh_sNcw) for questions and feedback
 - [GitHub Issues](https://github.com/sondera-ai/sondera-harness-python/issues) for bugs
 - [Contributing](CONTRIBUTING.md) for development setup
 

{sondera_harness-0.6.3 → sondera_harness-0.7.1}/src/sondera_harness.egg-info/requires.txt

@@ -2,7 +2,6 @@ cedar-python>=0.1.1
 click>=8.0.0
 click-default-group>=1.2.4
 grpcio>=1.76.0
-grpcio-tools>=1.76.0
 httpx>=0.27.0
 pydantic>=2.12.0
 pydantic-settings>=2.12.0