fix: async_client parity — stub memory_step/code, align context_trust default

- AsyncTraceSession.memory_step() now raises NotImplementedError
- AsyncTraceSession.code() now raises NotImplementedError
- AsyncGroundingClient.rag() context_trust_enabled defaults to False
- scoring_mode Literal narrowed to "rag" only
- Remove stale sessions.md links from README

Author: Latence AI

README.md

@@ -15,8 +15,6 @@
   ·
   <a href="docs/quickstart.md">Quickstart</a>
   ·
-  <a href="docs/sessions.md">Sessions</a>
-  ·
   <a href="docs/integrations.md">Integrations</a>
   ·
   <a href="examples/">Examples</a>
@@ -276,7 +274,6 @@ from latence import LatenceTraceClient, AsyncLatenceTraceClient
 ## More
 
 - [Quickstart](docs/quickstart.md)
-- [Sessions](docs/sessions.md)
 - [Integrations](docs/integrations.md)
 - [Phase 5 Integration Checkpoint](docs/phase5_integration_checkpoint.md)
 - [Errors](docs/errors.md)

src/latence/async_client.py

@@ -83,7 +83,7 @@ def __init__(self, owner: AsyncLatence) -> None:
     async def rag(
         self,
         *,
-        context_trust_enabled: bool = True,
+        context_trust_enabled: bool = False,
         **kwargs: Any,
     ) -> GroundednessResponse:
         extra = dict(kwargs.pop("extra", {}) or {})
@@ -151,18 +151,7 @@ def event(self, event_type: str, content: str, **metadata: Any) -> Mapping[str,
         return event
 
     async def memory_step(self, **payload: Any) -> MemoryUpdateResponse:
-        key = str(payload.pop("idempotency_key", "") or new_idempotency_key("memory"))
-        result = await self._owner.memory.step(
-            prior_memory_state=self.memory_state,
-            idempotency_key=key,
-            **payload,
-        )
-        next_state = result.next_memory_state
-        if isinstance(next_state, Mapping):
-            self.memory_state = dict(next_state)
-        self.idempotency_keys.append(key)
-        self.save()
-        return result
+        raise NotImplementedError("Memory API has been removed.")
 
     async def rag(
         self,
@@ -186,27 +175,8 @@ async def rag(
             **kwargs,
         )
 
-    async def code(
-        self,
-        *,
-        context_trust_enabled: bool | None = None,
-        **kwargs: Any,
-    ) -> GroundednessResponse:
-        extra = dict(kwargs.pop("extra", {}) or {})
-        if self.session_id:
-            extra.setdefault("session_id", self.session_id)
-        if self.memory_state:
-            extra.setdefault("memory_state", self.memory_state)
-        extra.setdefault("metadata", dict(self.metadata))
-        return await self._owner.grounding.code(
-            extra=extra,
-            context_trust_enabled=(
-                self.context_trust_enabled
-                if context_trust_enabled is None
-                else context_trust_enabled
-            ),
-            **kwargs,
-        )
+    async def code(self, **kwargs: Any) -> GroundednessResponse:
+        raise NotImplementedError("Code scoring has been removed. Use .rag() instead.")
 
     async def rollup(self, **options: Any) -> Mapping[str, Any]:
         return await self._owner.rollup(normalize_events(self.events), **options)

src/latence/models.py

@@ -266,7 +266,7 @@ class GroundednessRequest(BaseModel):
     evidence_limit: int | None = None
     heatmap_format: str | None = None
     auto_decide: bool | None = None
-    scoring_mode: Literal["rag", "code"] | None = None
+    scoring_mode: Literal["rag"] | None = None
 
 
 class RuntimeEvidenceUnit(BaseModel):