langgenius/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2026-07-21 20:15:31 -04:00
Code Issues 775 Packages Projects Releases 10 Wiki Activity

When the parameter extraction node is used in combination with the qwen3 series model, parameters cannot be extracted #15356

New Issue
Closed
opened 2026-02-21 19:21:21 -05:00 by yindo · 4 comments
yindo commented 2026-02-21 19:21:21 -05:00
Owner
Copy Link
Copy Source

Originally created by @GitHub-20200408 on GitHub (Jul 14, 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

1.5.1 and 1.4.3

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

I want to extract parameters from the query
Image

✔️ Expected Behavior

The qwen2.5 series is all fine
Image

❌ Actual Behavior

None of the qwen3 series models are allowed
Image

Originally created by @GitHub-20200408 on GitHub (Jul 14, 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 1.5.1 and 1.4.3 ### Cloud or Self Hosted Self Hosted (Docker) ### Steps to reproduce I want to extract parameters from the query <img width="682" height="258" alt="Image" src="https://github.com/user-attachments/assets/483d6617-d2c2-4351-97f7-ff25aefa1b3f" /> ### ✔️ Expected Behavior The qwen2.5 series is all fine <img width="679" height="519" alt="Image" src="https://github.com/user-attachments/assets/6c29b46f-e033-4e34-be48-702a3e60fdab" /> ### ❌ Actual Behavior None of the qwen3 series models are allowed <img width="761" height="495" alt="Image" src="https://github.com/user-attachments/assets/9f410bb3-b635-467c-a509-3085ef94efed" />
yindo added the 🐞 bug label 2026-02-21 19:21:21 -05:00
yindo closed this issue 2026-02-21 19:21:21 -05:00
yindo commented 2026-02-21 19:21:21 -05:00
Author
Owner
Copy Link
Copy Source

@dosubot[bot] commented on GitHub (Jul 14, 2025):

Hi @GitHub-20200408! I'm Dosu and I’m helping the dify team.

The parameter extraction issue with Qwen3 models is likely caused by their output format, which sometimes includes leftover <think></think> tags or extra reasoning text that breaks JSON parsing. Dify's extraction logic is model-agnostic and can handle minor JSON issues (like those from Qwen2.5), but it can't recover if the output isn't parseable as JSON.

To fix this, set the environment variable LLM_NODE_THINKING_TAGS_ENABLED=false in your deployment. This will strip out the entire <think>...</think> block from Qwen3 outputs, ensuring cleaner, parser-safe results for extraction. This approach is confirmed to help with Qwen3's unreliable output formatting and should resolve the extraction failures you are seeing [source].

Also, make sure Qwen3 is configured to use function/tool calling or structured output modes, and that the response_format and schema are set correctly in your model configuration. If issues persist, try prompt engineering to explicitly instruct Qwen3 to return only a JSON object, without any extra text or tags.

If you still have trouble, test Qwen3's raw output for a simple extraction case and compare it to Qwen2.5 to spot any format differences. This will help pinpoint if further prompt or config adjustments are needed.

To reply, just mention @dosu.

How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other  Join Discord Share on X

@dosubot[bot] commented on GitHub (Jul 14, 2025): <!-- Greeting --> Hi @GitHub-20200408! I'm [Dosu](https://go.dosu.dev/dosubot) and I’m helping the dify team. <!-- Answer --> The parameter extraction issue with Qwen3 models is likely caused by their output format, which sometimes includes leftover `<think></think>` tags or extra reasoning text that breaks JSON parsing. Dify's extraction logic is model-agnostic and can handle minor JSON issues (like those from Qwen2.5), but it can't recover if the output isn't parseable as JSON. To fix this, set the environment variable `LLM_NODE_THINKING_TAGS_ENABLED=false` in your deployment. This will strip out the entire `<think>...</think>` block from Qwen3 outputs, ensuring cleaner, parser-safe results for extraction. This approach is confirmed to help with Qwen3's unreliable output formatting and should resolve the extraction failures you are seeing [[source]](https://github.com/langgenius/dify/pull/21897). Also, make sure Qwen3 is configured to use function/tool calling or structured output modes, and that the `response_format` and schema are set correctly in your model configuration. If issues persist, try prompt engineering to explicitly instruct Qwen3 to return only a JSON object, without any extra text or tags. If you still have trouble, test Qwen3's raw output for a simple extraction case and compare it to Qwen2.5 to spot any format differences. This will help pinpoint if further prompt or config adjustments are needed. <!-- 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/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=great_response) | [Irrelevant](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=irrelevant_answer) | [Incorrect](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=incorrect_sources) | [Verbose](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=too_verbose) | [Hallucination](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=hallucination) | [Report 🐛](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=bug_report) | [Other](https://app.dosu.dev/response-feedback/fd525e83-8fb8-4666-a5ff-ce0565c42e28?feedback_type=other)</sup>&nbsp;&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/22377)
yindo commented 2026-02-21 19:21:21 -05:00
Author
Owner
Copy Link
Copy Source

@GitHub-20200408 commented on GitHub (Jul 14, 2025):

I set up to shut down my thoughts, but it's still the same as before
Image

Image
@GitHub-20200408 commented on GitHub (Jul 14, 2025): I set up to shut down my thoughts, but it's still the same as before <img width="434" height="61" alt="Image" src="https://github.com/user-attachments/assets/5b51fbaf-d68d-4305-a447-546cbdf0f1b6" /> <img width="438" height="257" alt="Image" src="https://github.com/user-attachments/assets/31478857-5fd5-4ea7-84f8-6755057719fa" />
yindo commented 2026-02-21 19:21:21 -05:00
Author
Owner
Copy Link
Copy Source

@crazywoola commented on GitHub (Jul 14, 2025):

This is only for bug report, if you would like to ask a question, please head to Discussions.

@crazywoola commented on GitHub (Jul 14, 2025): 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).
yindo commented 2026-02-21 19:21:21 -05:00
Author
Owner
Copy Link
Copy Source

@junjiem commented on GitHub (Jul 28, 2025):

This is a bug. The current parameter extractor not support stream, and qwen3 models only supports streaming output.

@junjiem commented on GitHub (Jul 28, 2025): This is a bug. The current parameter extractor not support stream, and qwen3 models only supports streaming output.
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
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#15356
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?