docs(sdk): explain remote stuck detection thresholds

sdk/guides/agent-stuck-detector.mdx

@@ -33,6 +33,45 @@ command 5 times in a row. The stuck detector analyzes the event history and iden
 The example demonstrates that stuck detection is enabled by default (`stuck_detection=True`), and you can check the
 stuck status at any point using `conversation.stuck_detector.is_stuck()`.
 
+## Configure thresholds for remote conversations
+
+Pass `stuck_detection_thresholds` when creating a Python `Conversation`, including
+when using a `RemoteWorkspace`:
+
+```python
+conversation = Conversation(
+    agent=agent,
+    workspace=remote_workspace,
+    stuck_detection_thresholds={
+        "action_observation": 6,
+        "alternating_pattern": 8,
+    },
+)
+```
+
+The Agent Server accepts the same `stuck_detection_thresholds` object in
+`POST /api/conversations`. Each value must be a positive integer. Omitted fields
+use the Python defaults: `action_observation=4`, `action_error=3`, `monologue=3`,
+and `alternating_pattern=6`. Omitting the object uses all defaults;
+`stuck_detection=False` disables detection even when thresholds are supplied.
+
+The server stores this configuration in conversation metadata and restores it
+when reloading or forking the conversation. Thresholds are configured at creation;
+there is no endpoint to update them on an existing conversation.
+
+For the TypeScript remote client, use camelCase options:
+
+```typescript
+const conversation = await manager.createConversation(agent, {
+  stuckDetectionThresholds: { actionObservation: 6, alternatingPattern: 8 },
+});
+```
+
+The client maps these keys to the server's snake_case fields. This requires an
+Agent Server version containing the remote-threshold propagation fix; older
+servers silently discard the field. This configuration does not change detector
+matching rules or the bounded history scan.
+
 ## Pattern Detection
 
 The stuck detector compares events based on their semantic content rather than object identity. For example: