When using the GPT model, the Agent node cannot streaming reply #21927

Closed
opened 2026-02-21 20:14:53 -05:00 by yindo · 1 comment
Owner

Originally created by @zio525 on GitHub (Jan 27, 2026).

Self Checks

  • I have read the Contributing Guide and Language Policy.
  • 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, otherwise it will be closed.
  • 【中文用户 & Non English User】请使用英语提交,否则会被关闭 :)
  • Please do not modify this template :) and fill in all the required fields.

Dify version

1.11.2

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

The apikey of the GPT model used is verified by the organization

The test chatflow:

Image

for a same question:

GPT-5.2:

Image Image

It took about 5 seconds in total, and the reply started in about 5 seconds

DeepSeek:

Image Image

It took a total of 21 seconds, and the reply will start after about 1 second

✔️ Expected Behavior

Agent node should streaming reply when I use the GPT.

Actual Behavior

the agent node doesn't streaming reply

Originally created by @zio525 on GitHub (Jan 27, 2026). ### Self Checks - [x] I have read the [Contributing Guide](https://github.com/langgenius/dify/blob/main/CONTRIBUTING.md) and [Language Policy](https://github.com/langgenius/dify/issues/1542). - [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, otherwise it will be closed. - [x] 【中文用户 & Non English User】请使用英语提交,否则会被关闭 :) - [x] Please do not modify this template :) and fill in all the required fields. ### Dify version 1.11.2 ### Cloud or Self Hosted Self Hosted (Docker) ### Steps to reproduce The apikey of the GPT model used is verified by the organization ### The test chatflow: <img width="1890" height="627" alt="Image" src="https://github.com/user-attachments/assets/7cd0df9b-a01b-4aa7-9acd-0d449b18962b" /> for a same question: ### GPT-5.2: <img width="1111" height="1199" alt="Image" src="https://github.com/user-attachments/assets/b043ab97-dbcd-4ac5-b6e8-f750a5643d23" /> <img width="2305" height="961" alt="Image" src="https://github.com/user-attachments/assets/4903b166-3c3b-4b66-b92d-b82b8253f303" /> It took about 5 seconds in total, and the reply started in about 5 seconds ### DeepSeek: <img width="1111" height="1375" alt="Image" src="https://github.com/user-attachments/assets/fcd0e4c5-b574-4667-9cdb-0a09b08869c1" /> <img width="2300" height="1319" alt="Image" src="https://github.com/user-attachments/assets/0e564786-9a31-4e35-8a39-abf312fc1192" /> It took a total of 21 seconds, and the reply will start after about 1 second ### ✔️ Expected Behavior Agent node should streaming reply when I use the GPT. ### ❌ Actual Behavior the agent node doesn't streaming reply
yindo added the 🐞 bug label 2026-02-21 20:14:53 -05:00
yindo closed this issue 2026-02-21 20:14:53 -05:00
Author
Owner

@dosubot[bot] commented on GitHub (Jan 27, 2026):

This is a known limitation in Dify: Agent nodes do not always stream replies, even when using GPT models, because streaming support depends on both the model/plugin declaring the stream-tool-call feature and the workflow design. There’s no config flag to enable streaming at the Agent node level—streaming is hard-wired but only works if the model/plugin emits incremental messages and the workflow allows it. Common causes for non-streaming include missing stream-tool-call in the plugin YAML, using ReAct mode instead of Function Calling, merges/aggregations after the agent node, or reverse proxy buffering (like nginx) blocking Server-Sent Events. Disabling exception handling in the workflow config can sometimes help as a workaround. For more details and troubleshooting steps, see this issue and this discussion.

To reply, just mention @dosu.


How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other

Chat with Dosu Open in Cursor Join Discord Share on X

@dosubot[bot] commented on GitHub (Jan 27, 2026): <!-- Answer --> This is a known limitation in Dify: Agent nodes do not always stream replies, even when using GPT models, because streaming support depends on both the model/plugin declaring the `stream-tool-call` feature and the workflow design. There’s no config flag to enable streaming at the Agent node level—streaming is hard-wired but only works if the model/plugin emits incremental messages and the workflow allows it. Common causes for non-streaming include missing `stream-tool-call` in the plugin YAML, using ReAct mode instead of Function Calling, merges/aggregations after the agent node, or reverse proxy buffering (like nginx) blocking Server-Sent Events. Disabling exception handling in the workflow config can sometimes help as a workaround. For more details and troubleshooting steps, see [this issue](https://github.com/langgenius/dify/issues/31259) and [this discussion](https://github.com/langgenius/dify/issues/28258). <!-- Dosu Comment Footer --> *To reply, just mention [@dosu](https://go.dosu.dev/dosubot).* --- <sup>How did I do? [Good](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=great_response) | [Irrelevant](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=irrelevant_answer) | [Incorrect](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=incorrect_sources) | [Verbose](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=too_verbose) | [Hallucination](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=hallucination) | [Report 🐛](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=bug_report) | [Other](https://app.dosu.dev/response-feedback/77e40d05-8f1f-43f0-a700-8a88c01ca726?feedback_type=other)</sup> [![Chat with Dosu](https://dosu.dev/dosu-chat-badge.svg)](https://app.dosu.dev/6088a4f4-073a-43f3-b3d4-4f06fb0c05a0/ask?utm_source=github)&nbsp;[![Open in Cursor](https://dosu.dev/dosu-in-cursor.svg)](https://cursor.com/link/prompt?text=This%20is%20a%20known%20limitation%20in%20Dify%3A%20Agent%20nodes%20do%20not%20always%20stream%20replies%2C%20even%20when%20using%20GPT%20models%2C%20because%20streaming%20support%20depends%20on%20both%20the%20model/plugin%20declaring%20the%20%60stream-tool-call%60%20feature%20and%20the%20workflow%20design.%20There%E2%80%99s%20no%20config%20flag%20to%20enable%20streaming%20at%20the%20Agent%20node%20level%E2%80%94streaming%20is%20hard-wired%20but%20only%20works%20if%20the%20model/plugin%20emits%20incremental%20messages%20and%20the%20workflow%20allows%20it.%20Common%20causes%20for%20non-streaming%20include%20missing%20%60stream-tool-call%60%20in%20the%20plugin%20YAML%2C%20using%20ReAct%20mode%20instead%20of%20Function%20Calling%2C%20merges/aggregations%20after%20the%20agent%20node%2C%20or%20reverse%20proxy%20buffering%20%28like%20nginx%29%20blocking%20Server-Sent%20Events.%20Disabling%20exception%20handling%20in%20the%20workflow%20config%20can%20sometimes%20help%20as%20a%20workaround.%20For%20more%20details%20and%20troubleshooting%20steps%2C%20see%20%5Bthis%20issue%5D%28https%3A//github.com/langgenius/dify/issues/31259%29%20and%20%5Bthis%20discussion%5D%28https%3A//github.com/langgenius/dify/issues/28258%29.)&nbsp;[![Join Discord](https://img.shields.io/badge/join-5865F2?logo=discord&logoColor=white&label=)](https://go.dosu.dev/discord-bot)&nbsp;[![Share on X](https://img.shields.io/badge/X-share-black)](https://twitter.com/intent/tweet?text=%40dosu_ai%20helped%20me%20solve%20this%20issue!&url=https%3A//github.com/langgenius/dify/issues/31609)
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: langgenius/dify#21927