diff --git a/src/vllm_tgis_adapter/grpc/grpc_server.py b/src/vllm_tgis_adapter/grpc/grpc_server.py
index 6c2c649..26d7cc5 100644
--- a/src/vllm_tgis_adapter/grpc/grpc_server.py
+++ b/src/vllm_tgis_adapter/grpc/grpc_server.py
@@ -283,6 +283,8 @@ async def Generate(
             #     # Abort the request if the client disconnects.
             #     await self.engine.abort(f"{request_id}-{i}")
             #     return self.create_error_response("Client disconnected")
+            if res.prompt is None:
+                res.prompt = request.requests[i].text
             responses[i] = res
             service_metrics.observe_queue_time(res)
 
@@ -322,7 +324,7 @@ async def Generate(
         return BatchedGenerationResponse(responses=responses)
 
     @log_rpc_handler_errors
-    async def GenerateStream(
+    async def GenerateStream(  # noqa: PLR0915
         self,
         request: SingleGenerationRequest,
         context: ServicerContext,
@@ -378,6 +380,8 @@ async def GenerateStream(
         last_engine_response = None
         # TODO handle cancellation
         async for result in result_generator:
+            if result.prompt is None:
+                result.prompt = request.request.text
             last_engine_response = result
             if first_response is None:
                 service_metrics.observe_queue_time(result)