[AgentServer] Refine error handling

sdk/agentserver/azure-ai-agentserver-agentframework/CHANGELOG.md

@@ -11,7 +11,6 @@
 
 - Fixed error response handling in stream and non-stream modes
 
-
 ## 1.0.0b6 (2025-11-26)
 
 ### Features Added

sdk/agentserver/azure-ai-agentserver-agentframework/azure/ai/agentserver/agentframework/agent_framework.py

@@ -167,20 +167,23 @@ async def _resolve_agent_for_request(self, context: AgentRunContext):
         return agent, tool_client_wrapper
 
     def init_tracing(self):
-        exporter = os.environ.get(AdapterConstants.OTEL_EXPORTER_ENDPOINT)
-        app_insights_conn_str = os.environ.get(APPINSIGHT_CONNSTR_ENV_NAME)
-        project_endpoint = os.environ.get(AdapterConstants.AZURE_AI_PROJECT_ENDPOINT)
-
-        if exporter or app_insights_conn_str:
-            from agent_framework.observability import setup_observability
-
-            setup_observability(
-                enable_sensitive_data=True,
-                otlp_endpoint=exporter,
-                applicationinsights_connection_string=app_insights_conn_str,
-            )
-        elif project_endpoint:
-            self.setup_tracing_with_azure_ai_client(project_endpoint)
+        try:
+            exporter = os.environ.get(AdapterConstants.OTEL_EXPORTER_ENDPOINT)
+            app_insights_conn_str = os.environ.get(APPINSIGHT_CONNSTR_ENV_NAME)
+            project_endpoint = os.environ.get(AdapterConstants.AZURE_AI_PROJECT_ENDPOINT)
+
+            if exporter or app_insights_conn_str:
+                from agent_framework.observability import setup_observability
+
+                setup_observability(
+                    enable_sensitive_data=True,
+                    otlp_endpoint=exporter,
+                    applicationinsights_connection_string=app_insights_conn_str,
+                )
+            elif project_endpoint:
+                self.setup_tracing_with_azure_ai_client(project_endpoint)
+        except Exception as e:
+            logger.warning(f"Failed to initialize tracing: {e}", exc_info=True)
         self.tracer = trace.get_tracer(__name__)
 
     def setup_tracing_with_azure_ai_client(self, project_endpoint: str):

sdk/agentserver/azure-ai-agentserver-agentframework/pyproject.toml

@@ -20,7 +20,7 @@ classifiers = [
 keywords = ["azure", "azure sdk"]
 
 dependencies = [
-    "azure-ai-agentserver-core>=1.0.0b5",
+    "azure-ai-agentserver-core>=1.0.0b7",
     "agent-framework-azure-ai>=1.0.0b251112",
     "agent-framework-core>=1.0.0b251112",
     "opentelemetry-exporter-otlp-proto-grpc>=1.36.0",
@@ -65,5 +65,9 @@ breaking = false   # incompatible python version
 pyright = false
 verifytypes = false   # incompatible python version for -core
 verify_keywords = false
-mindependency = false  # depends on -core package
+#mindependency = false  # depends on -core package
+#latestdependency = false
+#whl = false
+#depends = false
+#pylint = false
 whl_no_aio = false

sdk/agentserver/azure-ai-agentserver-core/CHANGELOG.md

@@ -11,7 +11,6 @@
 
 - Fixed error response handling in stream and non-stream modes
 
-
 ## 1.0.0b6 (2025-11-26)
 
 ### Features Added

sdk/agentserver/azure-ai-agentserver-core/azure/ai/agentserver/core/server/base.py

@@ -14,6 +14,7 @@
 from opentelemetry import context as otel_context, trace
 from opentelemetry.trace.propagation.tracecontext import TraceContextTextMapPropagator
 from starlette.applications import Starlette
+from starlette.concurrency import iterate_in_threadpool
 from starlette.middleware.base import BaseHTTPMiddleware
 from starlette.middleware.cors import CORSMiddleware
 from starlette.requests import Request
@@ -137,61 +138,46 @@ async def runs_endpoint(request):
                     context_carrier = {}
                     TraceContextTextMapPropagator().inject(context_carrier)
 
+                    ex = None
                     resp = await self.agent_run(context)
-
-                    if inspect.isgenerator(resp):
-                        def gen():
-                            ctx = TraceContextTextMapPropagator().extract(carrier=context_carrier)
-                            token = otel_context.attach(ctx)
-                            seq = 0
-                            try:
-                                for event in resp:
-                                    seq += 1
-                                    yield _event_to_sse_chunk(event)
-                            except Exception as e:  # noqa: BLE001
-                                logger.error("Error in non-async generator: %s", e, exc_info=True)
-                                err = project_models.ResponseErrorEvent(
-                                    sequence_number=seq + 1,
-                                    code=project_models.ResponseErrorCode.SERVER_ERROR,
-                                    message=_format_error(e),
-                                    param="")
-                                yield _event_to_sse_chunk(err)
-                            finally:
-                                logger.info("End of processing CreateResponse request.")
-                                otel_context.detach(token)
-
-                        return StreamingResponse(gen(), media_type="text/event-stream")
-                    if inspect.isasyncgen(resp):
-                        async def gen_async():
-                            ctx = TraceContextTextMapPropagator().extract(carrier=context_carrier)
-                            token = otel_context.attach(ctx)
-                            seq = 0
-                            try:
-                                async for event in resp:
-                                    seq += 1
-                                    yield _event_to_sse_chunk(event)
-                            except Exception as e:  # noqa: BLE001
-                                logger.error("Error in async generator: %s", e, exc_info=True)
-                                err = project_models.ResponseErrorEvent(
-                                    sequence_number=seq + 1,
-                                    code=project_models.ResponseErrorCode.SERVER_ERROR,
-                                    message=_format_error(e),
-                                    param="")
-                                yield _event_to_sse_chunk(err)
-                            finally:
-                                logger.info("End of processing CreateResponse request.")
-                                otel_context.detach(token)
-
-                        return StreamingResponse(gen_async(), media_type="text/event-stream")
-                    logger.info("End of processing CreateResponse request.")
-                    return JSONResponse(resp.as_dict())
                 except Exception as e:
                     # TODO: extract status code from exception
                     logger.error(f"Error processing CreateResponse request: {e}", exc_info=True)
-                    err = project_models.ResponseError(
+                    ex = e
+
+                if not context.stream:
+                    logger.info("End of processing CreateResponse request.")
+                    result = resp if not ex else project_models.ResponseError(
                         code=project_models.ResponseErrorCode.SERVER_ERROR,
-                        message=_format_error(e))
-                    return JSONResponse(err.as_dict())
+                        message=_format_error(ex))
+                    return JSONResponse(result.as_dict())
+
+                async def gen_async(ex):
+                    ctx = TraceContextTextMapPropagator().extract(carrier=context_carrier)
+                    token = otel_context.attach(ctx)
+                    seq = 0
+                    try:
+                        if ex:
+                            return
+                        it = iterate_in_threadpool(resp) if inspect.isgenerator(resp) else resp
+                        async for event in it:
+                            seq += 1
+                            yield _event_to_sse_chunk(event)
+                        logger.info("End of processing CreateResponse request.")
+                    except Exception as e:  # noqa: BLE001
+                        logger.error("Error in async generator: %s", e, exc_info=True)
+                        ex = e
+                    finally:
+                        if ex:
+                            err = project_models.ResponseErrorEvent(
+                                sequence_number=seq + 1,
+                                code=project_models.ResponseErrorCode.SERVER_ERROR,
+                                message=_format_error(ex),
+                                param="")
+                            yield _event_to_sse_chunk(err)
+                        otel_context.detach(token)
+
+                return StreamingResponse(gen_async(ex), media_type="text/event-stream")
 
         async def liveness_endpoint(request):
             result = await self.agent_liveness(request)

sdk/agentserver/azure-ai-agentserver-langgraph/CHANGELOG.md

@@ -11,7 +11,6 @@
 
 - Fixed error response handling in stream and non-stream modes
 
-
 ## 1.0.0b6 (2025-11-26)
 
 ### Feature Added

sdk/agentserver/azure-ai-agentserver-langgraph/azure/ai/agentserver/langgraph/langgraph.py

@@ -268,7 +268,7 @@ async def agent_run_non_stream(self, input_data: dict, context: AgentRunContext,
             output = self.state_converter.state_to_response(result, context)
             return output
         except Exception as e:
-            logger.error(f"Error during agent run: {e}")
+            logger.error(f"Error during agent run: {e}", exc_info=True)
             raise e
 
     async def agent_run_astream(
@@ -301,7 +301,7 @@ async def agent_run_astream(
             async for result in self.state_converter.state_to_response_stream(stream, context):
                 yield result
         except Exception as e:
-            logger.error(f"Error during streaming agent run: {e}")
+            logger.error(f"Error during streaming agent run: {e}", exc_info=True)
             raise e
         finally:
             # Close tool_client if provided

sdk/agentserver/azure-ai-agentserver-langgraph/pyproject.toml

@@ -19,7 +19,7 @@ classifiers = [
 keywords = ["azure", "azure sdk"]
 
 dependencies = [
-    "azure-ai-agentserver-core>=1.0.0b5",
+    "azure-ai-agentserver-core>=1.0.0b7",
     "langchain>0.3.20",
     "langchain-openai>0.3.10",
     "langchain-azure-ai[opentelemetry]>=0.1.8",
@@ -64,5 +64,9 @@ breaking = false   # incompatible python version
 pyright = false
 verifytypes = false   # incompatible python version for -core
 verify_keywords = false
-mindependency = false  # depends on -core package
+#mindependency = false  # depends on -core package
+#latestdependency = false
+#whl = false
+#depends = false
+#pylint = false
 whl_no_aio = false