diff --git a/.github/workflows/codespell.yaml b/.github/workflows/codespell.yaml
index 5773b9afb..21487e64f 100644
--- a/.github/workflows/codespell.yaml
+++ b/.github/workflows/codespell.yaml
@@ -15,5 +15,5 @@ jobs:
         with:
           ignore_words_file: .github/.codespellignore
           check_filenames: true
-          skip: go.*,**/*.drawio,./deploy/charts/*,./config/crd/*,./deploy/llms/*,./deploy/pgvector
+          skip: go.*,**/*.drawio,./deploy/charts/*,./config/crd/*,./deploy/llms/*,./deploy/pgvector,./pkg/documentloaders/testdata
 
diff --git a/apiserver/graph/generated/generated.go b/apiserver/graph/generated/generated.go
index 6c591459c..5cb5fa9df 100644
--- a/apiserver/graph/generated/generated.go
+++ b/apiserver/graph/generated/generated.go
@@ -5922,7 +5922,7 @@ input ListCommonInput {
 
     """
     每页数量，
-    规则: 默认10
+    规则: -1,返回全部
     """
     pageSize: Int
 }
diff --git a/apiserver/graph/generated/models_gen.go b/apiserver/graph/generated/models_gen.go
index 6a5caa577..cc74da00e 100644
--- a/apiserver/graph/generated/models_gen.go
+++ b/apiserver/graph/generated/models_gen.go
@@ -973,7 +973,7 @@ type ListCommonInput struct {
 	// 规则: 从1开始，默认是1
 	Page *int `json:"page,omitempty"`
 	// 每页数量，
-	// 规则: 默认10
+	// 规则: -1,返回全部
 	PageSize *int `json:"pageSize,omitempty"`
 }
 
diff --git a/config/samples/app_llmchain_abstract.yaml b/config/samples/app_llmchain_abstract.yaml
index 089a2c9a4..84df0e4f1 100644
--- a/config/samples/app_llmchain_abstract.yaml
+++ b/config/samples/app_llmchain_abstract.yaml
@@ -2,7 +2,7 @@ apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1
 kind: Application
 metadata:
   name: base-chat-document-assistant
-  namespace: kubeagi-system
+  namespace: arcadia
 spec:
   displayName: "AI文档对话助手"
   description: "最简单的AI文档对话助手"
@@ -37,7 +37,7 @@ spec:
       ref:
         apiGroup: arcadia.kubeagi.k8s.com.cn
         kind: LLM
-        name: qwen-0-5b-external
+        name: app-shared-llm-service
       nextNodeName: ["chain-node"]
     - name: chain-node
       displayName: "llm chain"
@@ -58,7 +58,7 @@ apiVersion: prompt.arcadia.kubeagi.k8s.com.cn/v1alpha1
 kind: Prompt
 metadata:
   name: base-chat-document-assistant
-  namespace: kubeagi-system
+  namespace: arcadia
   annotations:
     arcadia.kubeagi.k8s.com.cn/input-rules: '[{"kind":"Input","length":1}]'
     arcadia.kubeagi.k8s.com.cn/output-rules: '[{"length":1}]'
@@ -73,7 +73,7 @@ apiVersion: arcadia.kubeagi.k8s.com.cn/v1alpha1
 kind: DocumentLoader
 metadata:
   name: base-chat-document-assistant
-  namespace: kubeagi-system
+  namespace: arcadia
 spec:
   displayName: "llm chain"
   description: "llm chain"
@@ -84,16 +84,13 @@ apiVersion: chain.arcadia.kubeagi.k8s.com.cn/v1alpha1
 kind: LLMChain
 metadata:
   name: base-chat-document-assistant
-  namespace: kubeagi-system
+  namespace: arcadia
   annotations:
     arcadia.kubeagi.k8s.com.cn/input-rules: '[{"kind":"LLM","group":"arcadia.kubeagi.k8s.com.cn","length":1},{"kind":"prompt","group":"prompt.arcadia.kubeagi.k8s.com.cn","length":1}]'
     arcadia.kubeagi.k8s.com.cn/output-rules: '[{"kind":"Output","length":1}]'
 spec:
   displayName: "llm chain"
   description: "llm chain"
-  maxNumberOfConccurent: 1
-  maxTokens: 20480
-  maxLength: 20480
-  model: "6ac7baa2-71e7-4ffc-bd49-9356e743ecbb"
   memory:
-    maxTokenLimit: 20480
+    conversionWindowSize: 2
+  model: glm-4
diff --git a/pkg/appruntime/chain/mpchain.go b/pkg/appruntime/chain/mpchain.go
index 1e0474cf4..4203972ef 100644
--- a/pkg/appruntime/chain/mpchain.go
+++ b/pkg/appruntime/chain/mpchain.go
@@ -127,7 +127,7 @@ func (l *MapReduceChain) Run(ctx context.Context, _ client.Client, args map[stri
 	if err != nil {
 		return args, fmt.Errorf("failed to run MapReduceChain due to %s", err.Error())
 	}
-	args["_answer"] = fmt.Sprintf("Here is the document summary: %s \n", out)
+	args[base.AgentOutputInArg] = fmt.Sprintf("Here is the document summary: %s \n", out)
 	return args, nil
 }
 
diff --git a/pkg/documentloaders/testdata/llava.pdf b/pkg/documentloaders/testdata/llava.pdf
new file mode 100644
index 000000000..88da76cec
Binary files /dev/null and b/pkg/documentloaders/testdata/llava.pdf differ
diff --git a/tests/example-test.sh b/tests/example-test.sh
index 093ef2ea3..b3c36a072 100755
--- a/tests/example-test.sh
+++ b/tests/example-test.sh
@@ -262,6 +262,92 @@ function getRespInAppChat() {
 	fi
 }
 
+function fileUploadSummarise() {
+	appname=$1
+	namespace=$2
+	filename=$3
+	attempt=0
+	while true; do
+		info "sleep 3 seconds"
+		sleep 3
+		resp=$(curl --max-time $TimeoutSeconds -s --show-error -XPOST --form file=@$filename --form app_name=$appname --form app_namespace=$namespace -H "Content-Type: multipart/form-data" http://127.0.0.1:8081/chat/conversations/file)
+		doc_data=$(echo $resp | jq -r '.document')
+		if [ -z "$doc_data" ]; then
+			echo $resp
+			EnableAPIServerPortForward
+			if [[ $resp == *"googleapi: Error"* ]]; then
+				echo "google api error, will retry after 60s"
+				sleep 60
+			fi
+			attempt=$((attempt + 1))
+			if [ $attempt -gt $RETRY_COUNT ]; then
+				echo "❌: Failed. Retry count exceeded."
+				exit 1
+			fi
+			echo "🔄: Failed. Attempt $attempt/$RETRY_COUNT"
+			continue
+		fi
+		echo "👤: ${filename}"
+		echo "🤖: ${doc_data}"
+		break
+	done
+	file_id=$(echo $resp | jq -r '.document.object')
+	resp_conversation_id=$(echo $resp | jq -r '.conversation_id')
+	attempt=0
+	while true; do
+		info "sleep 3 seconds to sumerize doc"
+		sleep 3
+		data=$(jq -n --arg fileid "$file_id" --arg appname "$appname" --arg query "总结一下" --arg namespace "$namespace" --arg conversationID "$resp_conversation_id" '{"query":$query,"response_mode":"blocking","conversation_id":$conversationID,"app_name":$appname, "app_namespace":$namespace, "files": [$fileid]}')
+		resp=$(curl --max-time $TimeoutSeconds -s --show-error -XPOST http://127.0.0.1:8081/chat --data "$data")
+		ai_data=$(echo $resp | jq -r '.message')
+		references=$(echo $resp | jq -r '.references')
+		if [ -z "$ai_data" ] || [ "$ai_data" = "null" ]; then
+			echo $resp
+			EnableAPIServerPortForward
+			if [[ $resp == *"googleapi: Error"* ]]; then
+				echo "google api error, will retry after 60s"
+				sleep 60
+			fi
+			attempt=$((attempt + 1))
+			if [ $attempt -gt $RETRY_COUNT ]; then
+				echo "❌: Failed. Retry count exceeded."
+				exit 1
+			fi
+			echo "🔄: Failed. Attempt $attempt/$RETRY_COUNT"
+			continue
+		fi
+		echo "👤: 总结一下"
+		echo "🤖: ${ai_data}"
+		echo "🔗: ${references}"
+		break
+	done
+	resp_conversation_id=$(echo $resp | jq -r '.conversation_id')
+
+	if [ $testStream == "true" ]; then
+		attempt=0
+		while true; do
+			info "sleep 5 seconds"
+			sleep 5
+			info "just test stream mode"
+			data=$(jq -n --arg fileid "$file_id" --arg appname "$appname" --arg query "总结一下" --arg namespace "$namespace" --arg conversationID "$resp_conversation_id" '{"query":$query,"response_mode":"blocking","conversation_id":$conversationID,"app_name":$appname, "app_namespace":$namespace, "files": [$fileid]}')
+			curl --max-time $TimeoutSeconds -s --show-error -XPOST http://127.0.0.1:8081/chat --data "$data"
+			if [[ $? -ne 0 ]]; then
+				attempt=$((attempt + 1))
+				if [ $attempt -gt $RETRY_COUNT ]; then
+					echo "❌: Failed. Retry count exceeded."
+					exit 1
+				fi
+				echo "🔄: Failed. Attempt $attempt/$RETRY_COUNT"
+				EnableAPIServerPortForward
+				echo "and wait 60s for google api error"
+				sleep 60
+				continue
+			fi
+			break
+		done
+	fi
+}
+
 info "1. create kind cluster"
 make kind
 df -h
@@ -645,6 +731,12 @@ while true; do
 	break
 done
 
+info "8.4.6 chat with document"
+kubectl apply -f config/samples/app_llmchain_abstract.yaml
+waitCRDStatusReady "Application" "arcadia" "base-chat-document-assistant"
+fileUploadSummarise "base-chat-document-assistant" "arcadia" "./pkg/documentloaders/testdata/llava.pdf"
+getRespInAppChat "base-chat-document-assistant" "arcadia" "what is LLaVA？" ${resp_conversation_id} "false"
+
 # There is uncertainty in the AI replies, most of the time, it will pass the test, a small percentage of the time, the AI will call names in each reply, causing the test to fail, therefore, temporarily disable the following tests
 #getRespInAppChat "base-chat-with-bot" "arcadia" "What is your model?" ${resp_conversation_id} "false"
 #getRespInAppChat "base-chat-with-bot" "arcadia" "Does your model based on gpt-3.5?" ${resp_conversation_id} "false"