diff --git a/agent/component/agent_with_tools.py b/agent/component/agent_with_tools.py
index d59d8eb80..859064046 100644
--- a/agent/component/agent_with_tools.py
+++ b/agent/component/agent_with_tools.py
@@ -277,10 +277,13 @@ class Agent(LLM, ToolBase):
                 return
             if delta.find("**ERROR**") >= 0:
                 if self.get_exception_default_value():
-                    self.set_output("content", self.get_exception_default_value())
-                    yield self.get_exception_default_value()
+                    fallback = self.get_exception_default_value()
+                    self.set_output("content", fallback)
+                    yield fallback
                 else:
                     self.set_output("_ERROR", delta)
+                    self.set_output("content", delta)
+                    yield delta
                 return
             if not need2cite or cited:
                 yield delta
diff --git a/api/apps/llm_app.py b/api/apps/llm_app.py
index 1b520ec29..eaf56628f 100644
--- a/api/apps/llm_app.py
+++ b/api/apps/llm_app.py
@@ -29,6 +29,23 @@ from rag.utils.base64_image import test_image
 from rag.llm import EmbeddingModel, ChatModel, RerankModel, CvModel, TTSModel, OcrModel, Seq2txtModel
 
 
+def _resolve_my_llm_is_tools(o_dict: dict) -> bool:
+    decode_api_key_config = getattr(TenantLLMService, "_decode_api_key_config", None)
+    if callable(decode_api_key_config):
+        _, is_tools, _ = decode_api_key_config(o_dict.get("api_key", ""))
+        if is_tools is not None:
+            return bool(is_tools)
+
+    try:
+        base_name, fid = TenantLLMService.split_model_name_and_factory(o_dict["llm_name"])
+        llm_cfg = LLMService.query(llm_name=base_name, fid=fid) if fid else LLMService.query(llm_name=base_name)
+        if not llm_cfg and fid:
+            llm_cfg = LLMService.query(llm_name=base_name)
+        return bool(llm_cfg[0].is_tools) if llm_cfg else False
+    except Exception:
+        return False
+
+
 @manager.route("/factories", methods=["GET"])  # noqa: F821
 @login_required
 def factories():
@@ -229,6 +246,19 @@ async def add_llm():
     elif factory == "OpenDataLoader":
         api_key = apikey_json(["api_key", "provider_order"])
 
+    existing_llm = None
+    existing_api_key = None
+    if req.get("api_key") is None:
+        existing_llms = TenantLLMService.query(tenant_id=current_user.id, llm_factory=factory, llm_name=llm_name)
+        if existing_llms:
+            existing_llm = existing_llms[0]
+            existing_api_key, _, existing_api_key_payload = TenantLLMService._decode_api_key_config(existing_llm.api_key)
+            if existing_api_key_payload is not None:
+                existing_api_key = existing_api_key_payload
+
+    if req.get("api_key") is None:
+        api_key = existing_api_key if existing_api_key is not None else "x"
+
     llm = {
         "tenant_id": current_user.id,
         "llm_factory": factory,
@@ -353,6 +383,9 @@ async def add_llm():
     if msg:
         return get_data_error_result(message=msg)
 
+    if "is_tools" in req:
+        llm["api_key"] = TenantLLMService._encode_api_key_config(llm["api_key"], bool(req["is_tools"]))
+
     if not TenantLLMService.filter_update([TenantLLM.tenant_id == current_user.id, TenantLLM.llm_factory == factory, TenantLLM.llm_name == llm["llm_name"]], llm):
         TenantLLMService.save(**llm)
 
@@ -421,6 +454,7 @@ def my_llms():
                         "api_base": o_dict["api_base"] or "",
                         "max_tokens": o_dict["max_tokens"] or 8192,
                         "status": o_dict["status"] or "1",
+                        "is_tools": _resolve_my_llm_is_tools(o_dict),
                     }
                 )
         else:
diff --git a/api/db/joint_services/tenant_model_service.py b/api/db/joint_services/tenant_model_service.py
index 8e745d8e0..9f9487286 100644
--- a/api/db/joint_services/tenant_model_service.py
+++ b/api/db/joint_services/tenant_model_service.py
@@ -26,8 +26,14 @@ def get_model_config_by_id(tenant_model_id: int) -> dict:
     if not found:
         raise LookupError(f"Tenant Model with id {tenant_model_id} not found")
     config_dict = model_config.to_dict()
+    api_key, is_tools, api_key_payload = TenantLLMService._decode_api_key_config(config_dict.get("api_key", ""))
+    config_dict["api_key"] = api_key
+    if api_key_payload is not None:
+        config_dict["api_key_payload"] = api_key_payload
+    if is_tools is not None:
+        config_dict["is_tools"] = is_tools
     llm = LLMService.query(llm_name=config_dict["llm_name"])
-    if llm:
+    if "is_tools" not in config_dict and llm:
         config_dict["is_tools"] = llm[0].is_tools
     return config_dict
 
@@ -73,6 +79,12 @@ def get_model_config_by_type_and_name(tenant_id: str, model_type: str, model_nam
     else:
         # model_name without @factory
         config_dict = model_config.to_dict()
+    api_key, is_tools, api_key_payload = TenantLLMService._decode_api_key_config(config_dict.get("api_key", ""))
+    config_dict["api_key"] = api_key
+    if api_key_payload is not None:
+        config_dict["api_key_payload"] = api_key_payload
+    if is_tools is not None:
+        config_dict["is_tools"] = is_tools
     config_model_type = config_dict.get("model_type")
     config_model_type = config_model_type.value if hasattr(config_model_type, "value") else config_model_type
     if config_model_type != model_type_val and not (
@@ -83,7 +95,7 @@ def get_model_config_by_type_and_name(tenant_id: str, model_type: str, model_nam
             f"Tenant Model with name {model_name} has type {config_model_type}, expected {model_type_val}"
         )
     llm = LLMService.query(llm_name=config_dict["llm_name"])
-    if llm:
+    if "is_tools" not in config_dict and llm:
         config_dict["is_tools"] = llm[0].is_tools
     return config_dict
 
diff --git a/api/db/services/tenant_llm_service.py b/api/db/services/tenant_llm_service.py
index fe99aee49..5bf0c17d5 100644
--- a/api/db/services/tenant_llm_service.py
+++ b/api/db/services/tenant_llm_service.py
@@ -34,6 +34,42 @@ class LLMFactoriesService(CommonService):
 class TenantLLMService(CommonService):
     model = TenantLLM
 
+    @staticmethod
+    def _decode_api_key_config(raw_api_key: str) -> tuple[str, bool | None, str | None]:
+        if not raw_api_key:
+            return raw_api_key, None, None
+
+        try:
+            parsed = json.loads(raw_api_key)
+        except Exception:
+            return raw_api_key, None, None
+
+        if not isinstance(parsed, dict):
+            return raw_api_key, None, None
+
+        is_tools = bool(parsed["is_tools"]) if "is_tools" in parsed else None
+        if set(parsed.keys()) <= {"api_key", "is_tools"}:
+            return parsed.get("api_key", ""), is_tools, None
+
+        return parsed.get("api_key", raw_api_key), is_tools, raw_api_key
+
+    @staticmethod
+    def _encode_api_key_config(raw_api_key: str, is_tools: bool | None) -> str:
+        if is_tools is None:
+            return raw_api_key
+
+        try:
+            parsed = json.loads(raw_api_key or "{}")
+        except Exception:
+            parsed = None
+
+        if isinstance(parsed, dict):
+            payload = dict(parsed)
+            payload["is_tools"] = bool(is_tools)
+            return json.dumps(payload)
+
+        return json.dumps({"api_key": raw_api_key or "", "is_tools": bool(is_tools)})
+
     @classmethod
     @DB.connection_context()
     def get_api_key(cls, tenant_id, model_name, model_type=None):
@@ -123,6 +159,12 @@ class TenantLLMService(CommonService):
             model_config = cls.get_api_key(tenant_id, mdlnm, llm_type)
         if model_config:
             model_config = model_config.to_dict()
+            api_key, is_tools, api_key_payload = cls._decode_api_key_config(model_config.get("api_key", ""))
+            model_config["api_key"] = api_key
+            if api_key_payload is not None:
+                model_config["api_key_payload"] = api_key_payload
+            if is_tools is not None:
+                model_config["is_tools"] = is_tools
         elif llm_type == LLMType.EMBEDDING and fid == "Builtin" and "tei-" in os.getenv("COMPOSE_PROFILES", "") and mdlnm == os.getenv("TEI_MODEL", ""):
             embedding_cfg = settings.EMBEDDING_CFG
             model_config = {"llm_factory": "Builtin", "api_key": embedding_cfg["api_key"], "llm_name": mdlnm, "api_base": embedding_cfg["base_url"]}
@@ -132,7 +174,7 @@ class TenantLLMService(CommonService):
         llm = LLMService.query(llm_name=mdlnm) if not fid else LLMService.query(llm_name=mdlnm, fid=fid)
         if not llm and fid:  # for some cases seems fid mismatch
             llm = LLMService.query(llm_name=mdlnm)
-        if llm:
+        if "is_tools" not in model_config and llm:
             model_config["is_tools"] = llm[0].is_tools
         return model_config
 
@@ -142,35 +184,36 @@ class TenantLLMService(CommonService):
         if not model_config:
             raise LookupError("Model config is required")
         kwargs.update({"provider": model_config["llm_factory"]})
+        api_key = model_config.get("api_key_payload", model_config["api_key"])
         if model_config["model_type"] == LLMType.EMBEDDING.value:
             if model_config["llm_factory"] not in EmbeddingModel:
                 return None
-            return EmbeddingModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
+            return EmbeddingModel[model_config["llm_factory"]](api_key, model_config["llm_name"], base_url=model_config["api_base"])
 
         elif model_config["model_type"] == LLMType.RERANK:
             if model_config["llm_factory"] not in RerankModel:
                 return None
-            return RerankModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"])
+            return RerankModel[model_config["llm_factory"]](api_key, model_config["llm_name"], base_url=model_config["api_base"])
 
         elif model_config["model_type"] == LLMType.IMAGE2TEXT.value:
             if model_config["llm_factory"] not in CvModel:
                 return None
-            return CvModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], lang, base_url=model_config["api_base"], **kwargs)
+            return CvModel[model_config["llm_factory"]](api_key, model_config["llm_name"], lang, base_url=model_config["api_base"], **kwargs)
 
         elif model_config["model_type"] == LLMType.CHAT.value:
             if model_config["llm_factory"] not in ChatModel:
                 return None
-            return ChatModel[model_config["llm_factory"]](model_config["api_key"], model_config["llm_name"], base_url=model_config["api_base"], **kwargs)
+            return ChatModel[model_config["llm_factory"]](api_key, model_config["llm_name"], base_url=model_config["api_base"], **kwargs)
 
         elif model_config["model_type"] == LLMType.SPEECH2TEXT:
             if model_config["llm_factory"] not in Seq2txtModel:
                 return None
-            return Seq2txtModel[model_config["llm_factory"]](key=model_config["api_key"], model_name=model_config["llm_name"], lang=lang, base_url=model_config["api_base"])
+            return Seq2txtModel[model_config["llm_factory"]](key=api_key, model_name=model_config["llm_name"], lang=lang, base_url=model_config["api_base"])
         elif model_config["model_type"] == LLMType.TTS:
             if model_config["llm_factory"] not in TTSModel:
                 return None
             return TTSModel[model_config["llm_factory"]](
-                model_config["api_key"],
+                api_key,
                 model_config["llm_name"],
                 base_url=model_config["api_base"],
             )
@@ -179,7 +222,7 @@ class TenantLLMService(CommonService):
             if model_config["llm_factory"] not in OcrModel:
                 return None
             return OcrModel[model_config["llm_factory"]](
-                key=model_config["api_key"],
+                key=api_key,
                 model_name=model_config["llm_name"],
                 base_url=model_config.get("api_base", ""),
                 **kwargs,
diff --git a/web/src/interfaces/request/llm.ts b/web/src/interfaces/request/llm.ts
index 687d13aca..f8690784e 100644
--- a/web/src/interfaces/request/llm.ts
+++ b/web/src/interfaces/request/llm.ts
@@ -5,6 +5,7 @@ export interface IAddLlmRequestBody {
   api_base?: string; // chat|embedding|speech2text|image2text
   api_key?: string | Record<string, any>;
   max_tokens: number;
+  is_tools?: boolean;
 }
 
 export interface IDeleteLlmRequestBody {
diff --git a/web/src/locales/en.ts b/web/src/locales/en.ts
index ee4da4d14..9078dc749 100644
--- a/web/src/locales/en.ts
+++ b/web/src/locales/en.ts
@@ -1126,6 +1126,9 @@ This auto-tagging feature enhances retrieval by adding another layer of domain-s
       Verify: 'Verify',
       keyValid: 'Your API key is valid.',
       keyInvalid: 'Your API key is invalid.',
+      enableToolCall: 'Enable tool call',
+      enableToolCallTip:
+        'Allow this model to call tools when the selected model type supports tool calling.',
       deleteModel: 'Delete model',
       bedrockCredentialsHint:
         'Tip: Leave Access Key / Secret Key blank to use AWS IAM authentication.',
diff --git a/web/src/locales/zh.ts b/web/src/locales/zh.ts
index 8bbcc6a7e..97ebb5d7c 100644
--- a/web/src/locales/zh.ts
+++ b/web/src/locales/zh.ts
@@ -1036,6 +1036,8 @@ General：实体和关系提取提示来自 GitHub - microsoft/graphrag：基于
       Verify: '验证',
       keyValid: '你的 API 密钥有效。',
       keyInvalid: '你的 API 密钥无效。',
+      enableToolCall: '启用工具调用',
+      enableToolCallTip: '当所选模型类型支持工具调用时，允许该模型调用工具。',
       deleteModel: '删除模型',
       modelEmptyTip: '暂无可用模型,<br>请先在右侧面板添加模型。',
       sourceEmptyTip: '暂未添加任何数据源，请从下方选择一个进行连接。',
diff --git a/web/src/pages/user-setting/setting-model/hooks.tsx b/web/src/pages/user-setting/setting-model/hooks.tsx
index 47cfaa37c..1ddf3ac77 100644
--- a/web/src/pages/user-setting/setting-model/hooks.tsx
+++ b/web/src/pages/user-setting/setting-model/hooks.tsx
@@ -228,6 +228,7 @@ export const useSubmitOllama = () => {
         api_base: detailedData.api_base || '',
         max_tokens: detailedData.max_tokens || 8192,
         api_key: '',
+        is_tools: detailedData.is_tools || false,
       };
       setInitialValues(initialVals);
     } else {
diff --git a/web/src/pages/user-setting/setting-model/modal/ollama-modal/index.tsx b/web/src/pages/user-setting/setting-model/modal/ollama-modal/index.tsx
index a1c00e5aa..1b59ecb42 100644
--- a/web/src/pages/user-setting/setting-model/modal/ollama-modal/index.tsx
+++ b/web/src/pages/user-setting/setting-model/modal/ollama-modal/index.tsx
@@ -115,6 +115,7 @@ const OllamaModal = ({
     const getOptions = (factory: string) => {
       return optionsMap[factory as LLMFactory] || optionsMap.Default;
     };
+    const defaultToolCallEnabled = initialValues?.is_tools ?? false;
 
     const baseFields: FormFieldConfig[] = [
       {
@@ -177,6 +178,20 @@ const OllamaModal = ({
       },
     ];
 
+    baseFields.push({
+      name: 'is_tools',
+      label: t('enableToolCall'),
+      type: FormFieldType.Switch,
+      required: false,
+      dependencies: ['model_type'],
+      shouldRender: (formValues: any) => {
+        const modelType = formValues?.model_type;
+        return modelType === 'chat' || modelType === 'image2text';
+      },
+      tooltip: t('enableToolCallTip'),
+      defaultValue: defaultToolCallEnabled,
+    });
+
     // Add provider_order field only for OpenRouter
     if (llmFactory === 'OpenRouter') {
       baseFields.push({
@@ -214,14 +229,18 @@ const OllamaModal = ({
         api_key: '',
         vision: initialValues.model_type === 'image2text',
         provider_order: initialValues.provider_order || '',
+        is_tools: initialValues.is_tools || false,
       };
     }
     return {
       model_type:
-        llmFactory in optionsMap
-          ? optionsMap[llmFactory as LLMFactory]?.at(0)?.value
-          : 'embedding',
+        llmFactory === LLMFactory.Ollama || llmFactory === LLMFactory.VLLM
+          ? 'chat'
+          : llmFactory in optionsMap
+            ? optionsMap[llmFactory as LLMFactory]?.at(0)?.value
+            : 'embedding',
       vision: false,
+      is_tools: false,
     };
   }, [editMode, initialValues, llmFactory]);
 
@@ -232,6 +251,7 @@ const OllamaModal = ({
       values.model_type === 'chat' && values.vision
         ? 'image2text'
         : values.model_type;
+    const supportsToolCall = modelType === 'chat' || modelType === 'image2text';
 
     const data: IAddLlmRequestBody & { provider_order?: string } = {
       llm_factory: llmFactory,
@@ -241,6 +261,9 @@ const OllamaModal = ({
       api_key: values.api_key as string,
       max_tokens: values.max_tokens as number,
     };
+    if (supportsToolCall) {
+      data.is_tools = Boolean(values.is_tools);
+    }
 
     // Add provider_order only if it exists (for OpenRouter)
     if (values.provider_order) {