langgenius/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-07-22 20:45:26 -04:00
Code Issues 775 Packages Projects Releases 10 Wiki Activity

The agent's invocation of the ReAct function for MCP is unstable. #13742

New Issue
Closed
opened 2026-02-21 19:13:38 -05:00 by yindo · 5 comments
yindo commented 2026-02-21 19:13:38 -05:00
Owner
Copy Link
Copy Source

Originally created by @haofengW on GitHub (May 9, 2025).

Self Checks

  • This is only for bug report, if you would like to ask a question, please head to Discussions.
  • I have searched for existing issues search for existing issues, including closed ones.
  • I confirm that I am using English to submit this report (我已阅读并同意 Language Policy).
  • [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:)
  • Please do not modify this template :) and fill in all the required fields.

Dify version

1.3.0

Cloud or Self Hosted

Self Hosted (Source)

Steps to reproduce

  1. Create a workflow and introduce an Agent into the workflow.
  2. The Agent's strategy is set to ReAct.
  3. Select the model.
  4. Set the prompt words and the number of iterations to 2.
    As shown in the figure below.

Image

Image

✔️ Expected Behavior

Successfully invoke MCP

❌ Actual Behavior

In ROUND2, the MCP will be called, but sometimes there is ROUND2, and sometimes there isn't.

Originally created by @haofengW on GitHub (May 9, 2025). ### Self Checks - [x] This is only for bug report, if you would like to ask a question, please head to [Discussions](https://github.com/langgenius/dify/discussions/categories/general). - [x] I have searched for existing issues [search for existing issues](https://github.com/langgenius/dify/issues), including closed ones. - [x] I confirm that I am using English to submit this report (我已阅读并同意 [Language Policy](https://github.com/langgenius/dify/issues/1542)). - [x] [FOR CHINESE USERS] 请务必使用英文提交 Issue，否则会被关闭。谢谢！:) - [x] Please do not modify this template :) and fill in all the required fields. ### Dify version 1.3.0 ### Cloud or Self Hosted Self Hosted (Source) ### Steps to reproduce 1. Create a workflow and introduce an Agent into the workflow. 2. The Agent's strategy is set to ReAct. 3. Select the model. 4. Set the prompt words and the number of iterations to 2. As shown in the figure below. ![Image](https://github.com/user-attachments/assets/af105cca-15aa-426e-8744-8e4f7fb403ce) ![Image](https://github.com/user-attachments/assets/0d8279c8-9a77-4dde-87f9-85f41b3e2ce0) ### ✔️ Expected Behavior Successfully invoke MCP ### ❌ Actual Behavior In ROUND2, the MCP will be called, but sometimes there is ROUND2, and sometimes there isn't.
yindo added the 🐞 bug🤖 feat:agent labels 2026-02-21 19:13:38 -05:00
yindo closed this issue 2026-02-21 19:13:38 -05:00
yindo commented 2026-02-21 19:13:39 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (May 9, 2025):

I found a similar unsolved discussion that might be relevant to your issue: Sometimes the agent says it invoked the MCP tool but actually hasn't, yet still returns a result. This discussion was last updated on May 7, 2025 [1].

Regarding your issue, it seems that the instability might be related to the number of iterations set for the ReAct strategy. The qwen3:32b model has known limitations when using the ReAct strategy, particularly with tool-based operations when the maximum iterations are set to less than 3. It requires a configuration of max iterations equal to 3 to successfully execute such operations [2]. You might want to try adjusting the number of iterations to see if it resolves the instability.

To continue talking to Dosu, mention @dosu.

Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Bug Report | Other

@dosubot[bot] commented on GitHub (May 9, 2025): <!-- Answer --> I found a similar unsolved discussion that might be relevant to your issue: [Sometimes the agent says it invoked the MCP tool but actually hasn't, yet still returns a result](https://github.com/langgenius/dify/discussions/18922). This discussion was last updated on May 7, 2025 <sup>[[1]](https://github.com/langgenius/dify/discussions/18922)</sup>. Regarding your issue, it seems that the instability might be related to the number of iterations set for the ReAct strategy. The qwen3:32b model has known limitations when using the ReAct strategy, particularly with tool-based operations when the maximum iterations are set to less than 3. It requires a configuration of max iterations equal to 3 to successfully execute such operations <sup>[[2]](https://github.com/langgenius/dify/issues/16739)</sup>. You might want to try adjusting the number of iterations to see if it resolves the instability. <!-- Dosu Comment Footer --> *To continue talking to [Dosu](https://dosu.dev), mention @dosu.* --- <sup>Help Dosu learn! Give it feedback: [Great Response](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=great_response) | [Irrelevant Answer](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=irrelevant_answer) | [Incorrect Sources](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=incorrect_sources) | [Too Verbose](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=too_verbose) | [Hallucination](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=hallucination) | [Bug Report](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=bug_report) | [Other](https://app.dosu.dev/response-feedback/e545bcd2-21be-425e-b2c1-8af5dbd41ea6?feedback_type=other)</sup>
yindo commented 2026-02-21 19:13:39 -05:00
Author
Owner
Copy Link
Copy Source

@haofengW commented on GitHub (May 9, 2025):

When I changed it to 4 times, even if the tool invocation was successful, the iteration would still continue.
As shown in the figure below

Image

@haofengW commented on GitHub (May 9, 2025): When I changed it to 4 times, even if the tool invocation was successful, the iteration would still continue. As shown in the figure below ![Image](https://github.com/user-attachments/assets/61f60d50-2652-43e8-bfe7-e28861b5d268)
yindo commented 2026-02-21 19:13:39 -05:00
Author
Owner
Copy Link
Copy Source

@Jeff-CH-Y commented on GitHub (May 15, 2025):

AgentNode非常需要优化，现在非常难用。MCP的各种策略也很难用。但是一旦好用dify就牛逼了，伸手党许愿

@Jeff-CH-Y commented on GitHub (May 15, 2025): AgentNode非常需要优化，现在非常难用。MCP的各种策略也很难用。但是一旦好用dify就牛逼了，伸手党许愿
yindo commented 2026-02-21 19:13:39 -05:00
Author
Owner
Copy Link
Copy Source

@crazywoola commented on GitHub (Jul 6, 2025):

Stability under ReACT mode depends on the model itself

@crazywoola commented on GitHub (Jul 6, 2025): Stability under ReACT mode depends on the model itself
yindo commented 2026-02-21 19:13:39 -05:00
Author
Owner
Copy Link
Copy Source

@haofengW commented on GitHub (Aug 2, 2025):

Stability under ReACT mode depends on the model itself

I don't think so. With the same prompt words and the same model, in the directly - created Agent, the tool was correctly invoked and the result was returned immediately (as shown in Figure 1 below). However, in ChatFlow, after two rounds of thinking, it seems that the context was truncated in the second round (as shown in Figures 2 and 3 below). So it has nothing to do with the model itself. Moreover, the time taken to call the model in ChatFlow is far longer than that in the Agent.
The complete thinking process is as follows
{
"text": "\n好的，我现在需要处理用户的查询。首先，用户发来的是",
"files": [],
"json": [
{
"id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e",
"parent_id": null,
"error": null,
"status": "success",
"data": {
"action_input": {
"arguments": "{"transformer_id": "TR001"}",
"tool_name": "get_basic_info"
},
"action_name": "mcp_sse_call_tool",
"observation": "Tool execution result: [{'type': 'text', 'text': '{\n "status": "success",\n "message": "\\u6210\\u529f\\u83b7\\u53d6\\u53d8\\u538b\\u5668\\u57fa\\u7840\\u4fe1\\u606f",\n "transformer_id": "TR001",\n "data": {\n "items": [\n {\n "transformer_id": "TR001",\n "created_at": "2025-05-09T07:47:51.895051",\n "rated_capacity": "647MVA",\n "rated_voltage": "331kV",\n "rated_current": "331A",\n "impedance_voltage": "14%",\n "no_load_loss": "44kW",\n "load_loss": "303kW",\n "manufacture_date": "2016-02-14",\n "serial_number": "0a516424-8768-4aee-8cc9-366105e9c264",\n "winding_type": "YNd11",\n "core_material": "\\u975e\\u6676\\u5408\\u91d1",\n "cooling_type": "OFAF",\n "insulation_class": "A",\n "voltage_regulation_type": "\\u6709\\u8f7d\\u8c03\\u538b",\n "tap_changer_type": "\\u7535\\u5b50\\u5f0f",\n "tank_type": "\\u81ea\\u51b7\\u6cb9\\u7bb1",\n "updated_at": "2025-05-09T07:47:51.895051"\n }\n ]\n },\n "created_at": "2025-08-02 12:14:24"\n}'}]",
"thought": "\n好的，我现在需要处理用户的查询：“查询TR001的基础信息”。首先，我要确定用户的需求属于哪种情况。根据核心工作流程，情况1是当输入中不包含“observation: Tool execution result”时，需要调用工具。用户这里只是简单查询，没有之前的工具调用结果，所以属于情况1。\n\n接下来，我需要从接口清单中找到对应的服务。用户提到“基础信息”，根据参数映射表，“基础信息”对应的是get_basic_info接口。确认一下，参数需要transformer_id，这里用户提供了TR001，所以参数是transformer_id: "TR001"。其他参数如start_time和end_time是可选的，用户没有提到时间范围，所以不需要填写。\n\n然后，我需要构造正确的JSON动作。工具名称是mcp_sse_call_tool，参数需要tool_name和arguments。tool_name应该是get_basic_info，arguments需要是JSON字符串，包含transformer_id。注意参数格式要正确，用双引号，避免语法错误。\n\n检查一下参数是否正确：transformer_id是TR001，没有时间参数。所以arguments应该是{"transformer_id": "TR001"}，转换成JSON字符串。然后调用mcp_sse_call_tool工具。\n\n确认没有其他工具需要调用，比如mcp_sse_list_tools是列出工具，但这里不需要。所以正确的动作是调用mcp_sse_call_tool，参数正确。现在生成对应的JSON blob。\n"
},
"label": "ROUND 1",
"metadata": {
"currency": "USD",
"elapsed_time": 28.70030255499296,
"finished_at": 199162.256750017,
"started_at": 199133.556447628,
"total_price": "0",
"total_tokens": 2255
},
"node_id": "1745158769067"
},
{
"id": "d4ce72c6-b4e4-4e9e-8420-dd4bb63edd4e",
"parent_id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e",
"error": null,
"status": "success",
"data": {
"action": "mcp_sse_call_tool",
"action_input": {
"arguments": "{"transformer_id": "TR001"}",
"tool_name": "get_basic_info"
},
"thought": "\n好的，我现在需要处理用户的查询：“查询TR001的基础信息”。首先，我要确定用户的需求属于哪种情况。根据核心工作流程，情况1是当输入中不包含“observation: Tool execution result”时，需要调用工具。用户这里只是简单查询，没有之前的工具调用结果，所以属于情况1。\n\n接下来，我需要从接口清单中找到对应的服务。用户提到“基础信息”，根据参数映射表，“基础信息”对应的是get_basic_info接口。确认一下，参数需要transformer_id，这里用户提供了TR001，所以参数是transformer_id: "TR001"。其他参数如start_time和end_time是可选的，用户没有提到时间范围，所以不需要填写。\n\n然后，我需要构造正确的JSON动作。工具名称是mcp_sse_call_tool，参数需要tool_name和arguments。tool_name应该是get_basic_info，arguments需要是JSON字符串，包含transformer_id。注意参数格式要正确，用双引号，避免语法错误。\n\n检查一下参数是否正确：transformer_id是TR001，没有时间参数。所以arguments应该是{"transformer_id": "TR001"}，转换成JSON字符串。然后调用mcp_sse_call_tool工具。\n\n确认没有其他工具需要调用，比如mcp_sse_list_tools是列出工具，但这里不需要。所以正确的动作是调用mcp_sse_call_tool，参数正确。现在生成对应的JSON blob。\n"
},
"label": "qwen3:30b-a3b-thinking-2507-q8_0 Thought",
"metadata": {
"currency": "USD",
"elapsed_time": 28.578585013980046,
"finished_at": 199162.137023225,
"provider": "langgenius/ollama/ollama",
"started_at": 199133.558453503,
"total_price": "0",
"total_tokens": 2255,
"icon": "bfff83a66922c09cb5a5aa68829742b8b4f4e818579db42f53c7b8a30912cd8b.svg",
"icon_dark": null
},
"node_id": "1745158769067"
},
{
"id": "42c18eb0-7ca6-4ae5-b9e8-be0aa43cc371",
"parent_id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e",
"error": null,
"status": "success",
"data": {
"output": "Tool execution result: [{'type': 'text', 'text': '{\n "status": "success",\n "message": "\\u6210\\u529f\\u83b7\\u53d6\\u53d8\\u538b\\u5668\\u57fa\\u7840\\u4fe1\\u606f",\n "transformer_id": "TR001",\n "data": {\n "items": [\n {\n "transformer_id": "TR001",\n "created_at": "2025-05-09T07:47:51.895051",\n "rated_capacity": "647MVA",\n "rated_voltage": "331kV",\n "rated_current": "331A",\n "impedance_voltage": "14%",\n "no_load_loss": "44kW",\n "load_loss": "303kW",\n "manufacture_date": "2016-02-14",\n "serial_number": "0a516424-8768-4aee-8cc9-366105e9c264",\n "winding_type": "YNd11",\n "core_material": "\\u975e\\u6676\\u5408\\u91d1",\n "cooling_type": "OFAF",\n "insulation_class": "A",\n "voltage_regulation_type": "\\u6709\\u8f7d\\u8c03\\u538b",\n "tap_changer_type": "\\u7535\\u5b50\\u5f0f",\n "tank_type": "\\u81ea\\u51b7\\u6cb9\\u7bb1",\n "updated_at": "2025-05-09T07:47:51.895051"\n }\n ]\n },\n "created_at": "2025-08-02 12:14:24"\n}'}]",
"tool_call_args": {
"arguments": "{"transformer_id": "TR001"}",
"tool_name": "get_basic_info"
},
"tool_name": "mcp_sse_call_tool"
},
"label": "CALL mcp_sse_call_tool",
"metadata": {
"elapsed_time": 0.11897187499562278,
"finished_at": 199162.256590433,
"provider": "junjiem/mcp_sse/mcp_sse",
"started_at": 199162.137619017,
"icon": "/console/api/workspaces/current/plugin/icon?tenant_id=5c32de7c-12bc-4cf0-a269-1f1aa4f25609&filename=221f13757d18412c5db453374c10f539c671216dcc29456ed4695e86b9c756e8.png",
"icon_dark": ""
},
"node_id": "1745158769067"
},
{
"id": "bdca36ee-eeb9-4db1-9db2-89ec069b0fd7",
"parent_id": null,
"error": null,
"status": "success",
"data": {
"action_input": "",
"action_name": "",
"observation": "",
"thought": "\n好的，我现在需要处理用户的查询。首先，用户发来的是"
},
"label": "ROUND 2",
"metadata": {
"currency": "USD",
"elapsed_time": 11.441905921004945,
"finished_at": 199173.698725146,
"started_at": 199162.256819433,
"total_price": "0",
"total_tokens": 1896
},
"node_id": "1745158769067"
},
{
"id": "fe96f84a-d8ce-4823-aeec-a4a85862ba59",
"parent_id": "bdca36ee-eeb9-4db1-9db2-89ec069b0fd7",
"error": null,
"status": "success",
"data": {
"action": "\n好的，我现在需要处理用户的查询。首先，用户发来的是",
"thought": "\n好的，我现在需要处理用户的查询。首先，用户发来的是"
},
"label": "qwen3:30b-a3b-thinking-2507-q8_0 Thought",
"metadata": {
"currency": "USD",
"elapsed_time": 11.434410963003756,
"finished_at": 199173.697465104,
"provider": "langgenius/ollama/ollama",
"started_at": 199162.263066475,
"total_price": "0",
"total_tokens": 1896,
"icon": "bfff83a66922c09cb5a5aa68829742b8b4f4e818579db42f53c7b8a30912cd8b.svg",
"icon_dark": null
},
"node_id": "1745158769067"
},
{}
]
}

Image Image Image
@haofengW commented on GitHub (Aug 2, 2025): > Stability under ReACT mode depends on the model itself I don't think so. With the same prompt words and the same model, in the directly - created Agent, the tool was correctly invoked and the result was returned immediately (as shown in Figure 1 below). However, in ChatFlow, after two rounds of thinking, it seems that the context was truncated in the second round (as shown in Figures 2 and 3 below). So it has nothing to do with the model itself. Moreover, the time taken to call the model in ChatFlow is far longer than that in the Agent. The complete thinking process is as follows { "text": "<think>\n好的，我现在需要处理用户的查询。首先，用户发来的是", "files": [], "json": [ { "id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e", "parent_id": null, "error": null, "status": "success", "data": { "action_input": { "arguments": "{\"transformer_id\": \"TR001\"}", "tool_name": "get_basic_info" }, "action_name": "mcp_sse_call_tool", "observation": "Tool execution result: [{'type': 'text', 'text': '{\\n \"status\": \"success\",\\n \"message\": \"\\\\u6210\\\\u529f\\\\u83b7\\\\u53d6\\\\u53d8\\\\u538b\\\\u5668\\\\u57fa\\\\u7840\\\\u4fe1\\\\u606f\",\\n \"transformer_id\": \"TR001\",\\n \"data\": {\\n \"items\": [\\n {\\n \"transformer_id\": \"TR001\",\\n \"created_at\": \"2025-05-09T07:47:51.895051\",\\n \"rated_capacity\": \"647MVA\",\\n \"rated_voltage\": \"331kV\",\\n \"rated_current\": \"331A\",\\n \"impedance_voltage\": \"14%\",\\n \"no_load_loss\": \"44kW\",\\n \"load_loss\": \"303kW\",\\n \"manufacture_date\": \"2016-02-14\",\\n \"serial_number\": \"0a516424-8768-4aee-8cc9-366105e9c264\",\\n \"winding_type\": \"YNd11\",\\n \"core_material\": \"\\\\u975e\\\\u6676\\\\u5408\\\\u91d1\",\\n \"cooling_type\": \"OFAF\",\\n \"insulation_class\": \"A\",\\n \"voltage_regulation_type\": \"\\\\u6709\\\\u8f7d\\\\u8c03\\\\u538b\",\\n \"tap_changer_type\": \"\\\\u7535\\\\u5b50\\\\u5f0f\",\\n \"tank_type\": \"\\\\u81ea\\\\u51b7\\\\u6cb9\\\\u7bb1\",\\n \"updated_at\": \"2025-05-09T07:47:51.895051\"\\n }\\n ]\\n },\\n \"created_at\": \"2025-08-02 12:14:24\"\\n}'}]", "thought": "<think>\n好的，我现在需要处理用户的查询：“查询TR001的基础信息”。首先，我要确定用户的需求属于哪种情况。根据核心工作流程，情况1是当输入中不包含“observation: Tool execution result”时，需要调用工具。用户这里只是简单查询，没有之前的工具调用结果，所以属于情况1。\n\n接下来，我需要从接口清单中找到对应的服务。用户提到“基础信息”，根据参数映射表，“基础信息”对应的是get_basic_info接口。确认一下，参数需要transformer_id，这里用户提供了TR001，所以参数是transformer_id: \"TR001\"。其他参数如start_time和end_time是可选的，用户没有提到时间范围，所以不需要填写。\n\n然后，我需要构造正确的JSON动作。工具名称是mcp_sse_call_tool，参数需要tool_name和arguments。tool_name应该是get_basic_info，arguments需要是JSON字符串，包含transformer_id。注意参数格式要正确，用双引号，避免语法错误。\n\n检查一下参数是否正确：transformer_id是TR001，没有时间参数。所以arguments应该是{\"transformer_id\": \"TR001\"}，转换成JSON字符串。然后调用mcp_sse_call_tool工具。\n\n确认没有其他工具需要调用，比如mcp_sse_list_tools是列出工具，但这里不需要。所以正确的动作是调用mcp_sse_call_tool，参数正确。现在生成对应的JSON blob。\n</think>" }, "label": "ROUND 1", "metadata": { "currency": "USD", "elapsed_time": 28.70030255499296, "finished_at": 199162.256750017, "started_at": 199133.556447628, "total_price": "0", "total_tokens": 2255 }, "node_id": "1745158769067" }, { "id": "d4ce72c6-b4e4-4e9e-8420-dd4bb63edd4e", "parent_id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e", "error": null, "status": "success", "data": { "action": "mcp_sse_call_tool", "action_input": { "arguments": "{\"transformer_id\": \"TR001\"}", "tool_name": "get_basic_info" }, "thought": "<think>\n好的，我现在需要处理用户的查询：“查询TR001的基础信息”。首先，我要确定用户的需求属于哪种情况。根据核心工作流程，情况1是当输入中不包含“observation: Tool execution result”时，需要调用工具。用户这里只是简单查询，没有之前的工具调用结果，所以属于情况1。\n\n接下来，我需要从接口清单中找到对应的服务。用户提到“基础信息”，根据参数映射表，“基础信息”对应的是get_basic_info接口。确认一下，参数需要transformer_id，这里用户提供了TR001，所以参数是transformer_id: \"TR001\"。其他参数如start_time和end_time是可选的，用户没有提到时间范围，所以不需要填写。\n\n然后，我需要构造正确的JSON动作。工具名称是mcp_sse_call_tool，参数需要tool_name和arguments。tool_name应该是get_basic_info，arguments需要是JSON字符串，包含transformer_id。注意参数格式要正确，用双引号，避免语法错误。\n\n检查一下参数是否正确：transformer_id是TR001，没有时间参数。所以arguments应该是{\"transformer_id\": \"TR001\"}，转换成JSON字符串。然后调用mcp_sse_call_tool工具。\n\n确认没有其他工具需要调用，比如mcp_sse_list_tools是列出工具，但这里不需要。所以正确的动作是调用mcp_sse_call_tool，参数正确。现在生成对应的JSON blob。\n</think>" }, "label": "qwen3:30b-a3b-thinking-2507-q8_0 Thought", "metadata": { "currency": "USD", "elapsed_time": 28.578585013980046, "finished_at": 199162.137023225, "provider": "langgenius/ollama/ollama", "started_at": 199133.558453503, "total_price": "0", "total_tokens": 2255, "icon": "bfff83a66922c09cb5a5aa68829742b8b4f4e818579db42f53c7b8a30912cd8b.svg", "icon_dark": null }, "node_id": "1745158769067" }, { "id": "42c18eb0-7ca6-4ae5-b9e8-be0aa43cc371", "parent_id": "e4bfbd23-d42e-4b5d-af42-661c181ebe8e", "error": null, "status": "success", "data": { "output": "Tool execution result: [{'type': 'text', 'text': '{\\n \"status\": \"success\",\\n \"message\": \"\\\\u6210\\\\u529f\\\\u83b7\\\\u53d6\\\\u53d8\\\\u538b\\\\u5668\\\\u57fa\\\\u7840\\\\u4fe1\\\\u606f\",\\n \"transformer_id\": \"TR001\",\\n \"data\": {\\n \"items\": [\\n {\\n \"transformer_id\": \"TR001\",\\n \"created_at\": \"2025-05-09T07:47:51.895051\",\\n \"rated_capacity\": \"647MVA\",\\n \"rated_voltage\": \"331kV\",\\n \"rated_current\": \"331A\",\\n \"impedance_voltage\": \"14%\",\\n \"no_load_loss\": \"44kW\",\\n \"load_loss\": \"303kW\",\\n \"manufacture_date\": \"2016-02-14\",\\n \"serial_number\": \"0a516424-8768-4aee-8cc9-366105e9c264\",\\n \"winding_type\": \"YNd11\",\\n \"core_material\": \"\\\\u975e\\\\u6676\\\\u5408\\\\u91d1\",\\n \"cooling_type\": \"OFAF\",\\n \"insulation_class\": \"A\",\\n \"voltage_regulation_type\": \"\\\\u6709\\\\u8f7d\\\\u8c03\\\\u538b\",\\n \"tap_changer_type\": \"\\\\u7535\\\\u5b50\\\\u5f0f\",\\n \"tank_type\": \"\\\\u81ea\\\\u51b7\\\\u6cb9\\\\u7bb1\",\\n \"updated_at\": \"2025-05-09T07:47:51.895051\"\\n }\\n ]\\n },\\n \"created_at\": \"2025-08-02 12:14:24\"\\n}'}]", "tool_call_args": { "arguments": "{\"transformer_id\": \"TR001\"}", "tool_name": "get_basic_info" }, "tool_name": "mcp_sse_call_tool" }, "label": "CALL mcp_sse_call_tool", "metadata": { "elapsed_time": 0.11897187499562278, "finished_at": 199162.256590433, "provider": "junjiem/mcp_sse/mcp_sse", "started_at": 199162.137619017, "icon": "/console/api/workspaces/current/plugin/icon?tenant_id=5c32de7c-12bc-4cf0-a269-1f1aa4f25609&filename=221f13757d18412c5db453374c10f539c671216dcc29456ed4695e86b9c756e8.png", "icon_dark": "" }, "node_id": "1745158769067" }, { "id": "bdca36ee-eeb9-4db1-9db2-89ec069b0fd7", "parent_id": null, "error": null, "status": "success", "data": { "action_input": "", "action_name": "", "observation": "", "thought": "<think>\n好的，我现在需要处理用户的查询。首先，用户发来的是" }, "label": "ROUND 2", "metadata": { "currency": "USD", "elapsed_time": 11.441905921004945, "finished_at": 199173.698725146, "started_at": 199162.256819433, "total_price": "0", "total_tokens": 1896 }, "node_id": "1745158769067" }, { "id": "fe96f84a-d8ce-4823-aeec-a4a85862ba59", "parent_id": "bdca36ee-eeb9-4db1-9db2-89ec069b0fd7", "error": null, "status": "success", "data": { "action": "<think>\n好的，我现在需要处理用户的查询。首先，用户发来的是", "thought": "<think>\n好的，我现在需要处理用户的查询。首先，用户发来的是" }, "label": "qwen3:30b-a3b-thinking-2507-q8_0 Thought", "metadata": { "currency": "USD", "elapsed_time": 11.434410963003756, "finished_at": 199173.697465104, "provider": "langgenius/ollama/ollama", "started_at": 199162.263066475, "total_price": "0", "total_tokens": 1896, "icon": "bfff83a66922c09cb5a5aa68829742b8b4f4e818579db42f53c7b8a30912cd8b.svg", "icon_dark": null }, "node_id": "1745158769067" }, {} ] } <img width="1495" height="806" alt="Image" src="https://github.com/user-attachments/assets/2755f7f0-0df8-477c-9d4f-06f236d9665e" /> <img width="875" height="650" alt="Image" src="https://github.com/user-attachments/assets/13024c5c-185e-4985-95d0-0df1f2e839af" /> <img width="1507" height="830" alt="Image" src="https://github.com/user-attachments/assets/38a95607-4723-4c3b-9fa7-33456c1215a6" />
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
Knowledge Pipeline
Proposal
WIP
cloud
dependencies
feat:webapp
good first issue
outdated
performance
pull-request
Mirrored from GitHub Pull Request
 python:uv
refactor
size:L
stale
status: accepting prs
status: blocked
status: in discussion
stt & tts
vulnerablity
web
wont-fix
🌊 feat:workflow
🌚 invalid
🐞 bug
👻 feat:rag
💪 enhancement
📚 documentation
🔨 feat:tools
🙋‍♂️ question
🙏 help wanted
🤔 cant-reproduce
🤖 feat:agent
No Label 🐞 bug 🤖 feat:agent
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
yindo
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#13742
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?