fix: preserve wrapper async request ids

extensions/business/edge_inference_api/base_inference_api.py

@@ -2596,6 +2596,39 @@ def maybe_mark_request_timeout(self, request_id: str, request_data: Dict[str, An
       self._decrement_active_requests()
       return True
 
+    def _extract_request_id_override(self, kwargs: Dict[str, Any]):
+      """Extract an optional caller-provided request id from endpoint kwargs.
+
+      Existing callers do not pass a request id and keep the generated-id path.
+      New paired apps may pass `request_id` so wrapper and inference tracking use
+      the same id without a separate mapping store.
+      """
+      if not isinstance(kwargs, dict):
+        return None
+      values = []
+      for key in ('request_id', 'REQUEST_ID'):
+        if key in kwargs:
+          value = kwargs.pop(key)
+          if value is not None:
+            values.append(value)
+      if not values:
+        return None
+      request_id = values[0]
+      for value in values[1:]:
+        if value != request_id:
+          raise ValueError("Conflicting request_id and REQUEST_ID values.")
+      if not isinstance(request_id, str):
+        raise ValueError("request_id must be a string.")
+      request_id = request_id.strip()
+      if not request_id:
+        raise ValueError("request_id must not be empty.")
+      if len(request_id) > 256:
+        raise ValueError("request_id must not exceed 256 characters.")
+      allowed_chars = set("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789._:-")
+      if any(ch not in allowed_chars for ch in request_id):
+        raise ValueError("request_id contains unsupported characters.")
+      return request_id
+
     def solve_postponed_request(self, request_id: str):
       """
       Resolve or requeue a postponed request by checking its current status.
@@ -2666,6 +2699,8 @@ def register_request(
         Generated request_id and the stored request data dictionary.
       """
       request_id = request_id or self.uuid()
+      if request_id in self._requests:
+        raise ValueError(f"Request ID {request_id} already exists.")
       start_time = self.time()
       request_data = {
         "request_id": request_id,
@@ -2857,6 +2892,7 @@ def predict(
     def predict_async(
         self,
         authorization: Optional[str] = None,
+        request_id: Optional[str] = None,
         **kwargs
     ):
       """
@@ -2866,6 +2902,9 @@ def predict_async(
       ----------
       authorization : str or None, optional
         Authorization token supplied by the caller.
+      request_id : str or None, optional
+        Caller-provided id to use for request tracking. If omitted, the API
+        keeps the legacy generated-id behavior.
       **kwargs
         Additional parameters forwarded to request handling.
 
@@ -2877,6 +2916,7 @@ def predict_async(
       return self._predict_entrypoint(
         authorization=authorization,
         async_request=True,
+        request_id=request_id,
         **kwargs
       )
   """END API ENDPOINTS"""
@@ -2992,6 +3032,20 @@ def _predict_entrypoint(
           return {'error': f"Unexpected error: {str(exc)}", 'status': 'error'}
         # endtry
 
+      request_id_override = None
+      if delegated_execution:
+        request_id_override = delegation_context.get('delegation_id')
+      elif async_request:
+        try:
+          request_id_override = self._extract_request_id_override(kwargs)
+        except ValueError as exc:
+          return {'error': str(exc), 'status': 'error'}
+      else:
+        # `request_id` is reserved for async tracking and must not leak into
+        # synchronous model parameters when clients reuse async payload shapes.
+        kwargs.pop('request_id', None)
+        kwargs.pop('REQUEST_ID', None)
+
       err = self.check_predict_params(**kwargs)
       if err is not None:
         return {'error': err}
@@ -3000,16 +3054,16 @@ def _predict_entrypoint(
       if 'metadata' in parameters:
         metadata = parameters.pop('metadata') or {}
       # endif 'metadata' in parameters
-      request_id_override = None
-      if delegated_execution:
-        request_id_override = delegation_context.get('delegation_id')
-      request_id, request_data = self.register_request(
-        subject=subject,
-        parameters=parameters,
-        metadata=metadata,
-        timeout=parameters.get('timeout'),
-        request_id=request_id_override,
-      )
+      try:
+        request_id, request_data = self.register_request(
+          subject=subject,
+          parameters=parameters,
+          metadata=metadata,
+          timeout=parameters.get('timeout'),
+          request_id=request_id_override,
+        )
+      except ValueError as exc:
+        return {'error': str(exc), 'status': 'error'}
       request_data['endpoint_name'] = endpoint_name
       request_data['async_request'] = async_request
 

extensions/business/edge_inference_api/cv_inference_api.py

@@ -236,6 +236,7 @@ def predict_async(
         image_data: str = '',
         metadata: Optional[Dict[str, Any]] = None,
         authorization: Optional[str] = None,
+        request_id: Optional[str] = None,
         **kwargs
     ):
       """
@@ -249,6 +250,9 @@ def predict_async(
         Optional metadata accompanying the request.
       authorization : str or None, optional
         Bearer token used for authentication.
+      request_id : str or None, optional
+        Caller-provided id to use for request tracking. If omitted, the API
+        keeps the legacy generated-id behavior.
       **kwargs
         Extra parameters forwarded to the base handler.
 
@@ -261,6 +265,7 @@ def predict_async(
         image_data=image_data,
         metadata=metadata,
         authorization=authorization,
+        request_id=request_id,
         **kwargs
       )
   """END API ENDPOINTS"""

extensions/business/edge_inference_api/llm_inference_api.py

@@ -399,6 +399,7 @@ def predict_async(
         response_format: Optional[Dict[str, Any]] = None,
         metadata: Optional[Dict[str, Any]] = None,
         authorization: Optional[str] = None,
+        request_id: Optional[str] = None,
         **kwargs
     ):
       """
@@ -422,6 +423,9 @@ def predict_async(
         Additional metadata to store with the request.
       authorization : str or None, optional
         Bearer token used for authentication.
+      request_id : str or None, optional
+        Caller-provided id to use for request tracking. If omitted, the API
+        keeps the legacy generated-id behavior.
       **kwargs
         Extra parameters forwarded to the base handler.
 
@@ -439,6 +443,7 @@ def predict_async(
         response_format=response_format,
         metadata=metadata,
         authorization=authorization,
+        request_id=request_id,
         **kwargs
       )
 

extensions/business/edge_inference_api/sd_inference_api.py

@@ -283,6 +283,7 @@ def predict_async(
         struct_data: Any = None,
         metadata: Optional[Dict[str, Any]] = None,
         authorization: Optional[str] = None,
+        request_id: Optional[str] = None,
         **kwargs
     ):
       """
@@ -296,6 +297,9 @@ def predict_async(
         Optional metadata accompanying the request.
       authorization : str or None, optional
         Bearer token used for authentication.
+      request_id : str or None, optional
+        Caller-provided id to use for request tracking. If omitted, the API
+        keeps the legacy generated-id behavior.
       **kwargs
         Extra parameters forwarded to the base handler.
 
@@ -308,6 +312,7 @@ def predict_async(
         struct_data=struct_data,
         metadata=metadata,
         authorization=authorization,
+        request_id=request_id,
         **kwargs
       )
   """END API ENDPOINTS"""

extensions/business/edge_inference_api/test_base_inference_api_balancing.py

@@ -580,6 +580,88 @@ def test_predict_entrypoint_queues_when_full_and_no_peer(self):
     request_id = result["request_id"]
     self.assertEqual(plugin._requests[request_id]["queue_state"], "queued")  # pylint: disable=protected-access
 
+  def test_predict_entrypoint_accepts_caller_request_id(self):
+    plugin = self._make_plugin()
+
+    result = plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=True,
+      request_id="client-req_1.2:3",
+      metadata={"source": "test"},
+    )
+
+    self.assertEqual(result["request_id"], "client-req_1.2:3")
+    self.assertIn("client-req_1.2:3", plugin._requests)  # pylint: disable=protected-access
+    self.assertEqual(plugin.payloads[-1]["REQUEST_ID"], "client-req_1.2:3")
+    self.assertNotIn("request_id", plugin._requests["client-req_1.2:3"]["parameters"])  # pylint: disable=protected-access
+
+  def test_predict_async_without_request_id_keeps_generated_id_behavior(self):
+    plugin = self._make_plugin()
+
+    result = plugin.predict_async(authorization=None, request_id=None)
+
+    self.assertEqual(result["request_id"], "req-1")
+    self.assertIn("req-1", plugin._requests)  # pylint: disable=protected-access
+
+  def test_sync_predict_does_not_treat_request_id_as_tracking_override(self):
+    plugin = self._make_plugin()
+
+    plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=False,
+      request_id="client-sync-id",
+      REQUEST_ID="client-sync-id-upper",
+    )
+
+    self.assertIn("req-1", plugin._requests)  # pylint: disable=protected-access
+    self.assertNotIn("client-sync-id", plugin._requests)  # pylint: disable=protected-access
+    self.assertNotIn("request_id", plugin._requests["req-1"]["parameters"])  # pylint: disable=protected-access
+    self.assertNotIn("REQUEST_ID", plugin._requests["req-1"]["parameters"])  # pylint: disable=protected-access
+    self.assertEqual(plugin.payloads[-1]["REQUEST_ID"], "req-1")
+
+  def test_predict_entrypoint_accepts_uppercase_request_id_alias(self):
+    plugin = self._make_plugin()
+
+    result = plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=True,
+      REQUEST_ID="client-req-2",
+    )
+
+    self.assertEqual(result["request_id"], "client-req-2")
+    self.assertIn("client-req-2", plugin._requests)  # pylint: disable=protected-access
+
+  def test_predict_entrypoint_rejects_duplicate_caller_request_id(self):
+    plugin = self._make_plugin()
+
+    first = plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=True,
+      request_id="client-req-dup",
+    )
+    second = plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=True,
+      request_id="client-req-dup",
+    )
+
+    self.assertEqual(first["request_id"], "client-req-dup")
+    self.assertEqual(second["status"], "error")
+    self.assertIn("already exists", second["error"])
+
+  def test_predict_entrypoint_rejects_invalid_caller_request_id(self):
+    plugin = self._make_plugin()
+
+    result = plugin._predict_entrypoint(  # pylint: disable=protected-access
+      authorization=None,
+      async_request=True,
+      request_id="../bad",
+    )
+
+    self.assertEqual(result["status"], "error")
+    self.assertIn("unsupported characters", result["error"])
+    self.assertNotIn("../bad", plugin._requests)  # pylint: disable=protected-access
+
   def test_predict_entrypoint_fails_cleanly_when_delegated_request_cannot_encode(self):
     plugin = self._make_plugin()
     plugin._active_execution_slots.add("busy")  # pylint: disable=protected-access

extensions/business/edge_inference_api/text_classifier_inference_api.py

@@ -276,6 +276,7 @@ def predict_async(
       text: str = "",
       metadata: Optional[Dict[str, Any]] = None,
       authorization: Optional[str] = None,
+      request_id: Optional[str] = None,
       **kwargs
     ):
       """
@@ -289,6 +290,9 @@ def predict_async(
         Optional metadata accompanying the request.
       authorization : str or None, optional
         Bearer token used for authentication.
+      request_id : str or None, optional
+        Caller-provided id to use for request tracking. If omitted, the API
+        keeps the legacy generated-id behavior.
       **kwargs
         Extra parameters forwarded to the base handler.
 
@@ -301,6 +305,7 @@ def predict_async(
         text=text,
         metadata=metadata,
         authorization=authorization,
+        request_id=request_id,
         **kwargs
       )
   """END API ENDPOINTS"""

ver.py

@@ -1 +1 @@
-__VER__ = '2.10.290'
+__VER__ = '2.10.300'