OpenAI-API-compatible Model throws "Non-JSON encountered." error #45

New Issue

Closed

opened 2026-02-16 10:17:51 -05:00 by yindo · 16 comments

yindo commented 2026-02-16 10:17:51 -05:00

Owner

Copy Link

Copy Source

Originally created by @TinyFatty on GitHub (Mar 6, 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 Dify issues & Dify Official Plugins, 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

1.0.0

Plugin version

Built-in LLM block（I don't see any version here）

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

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

1.0.0

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

I used OpenAI-API-compatible Models (offered by China Telecom) in workflow LLM blocks and it throws "Non-JSON encountered." error (3 blocks share the same configuration, Deepseek official api and Silconflow deepseek api return the correct answer).

Here is the configuration for all

Here is the error message

OUTPUT data:

{
  "text": "",
  "usage": {
    "prompt_tokens": 0,
    "prompt_unit_price": "0",
    "prompt_price_unit": "0",
    "prompt_price": "0",
    "completion_tokens": 0,
    "completion_unit_price": "0",
    "completion_price_unit": "0",
    "completion_price": "0",
    "total_tokens": 0,
    "total_price": "0",
    "currency": "USD",
    "latency": 0.9801772950013401
  },
  "finish_reason": "Non-JSON encountered."
}

It takes only 1 second and no token is used,so I think the error occurs before the model starts reasoning.
Then I tested this model with python scripts

import openai

client= openai.OpenAI(base_url="***",
                           api_key="***")


response = client.chat.completions.create(
    model="7ba7726dad4c4ea4ab7f39c7741aea68",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {"role": "user", "content": "Who are you?"}
    ],
    temperature=0.7
)

It shows the corrent answer
{"code":0,"id":"chatcmpl-00-89c57718adba43dd5650511e4ecece0f-ba26f79321d736fa-01","choices":[{"index":0,"message":{"content":"<think>\nGreetings! I'm DeepSeek-R1, an artificial intelligence assistant created by DeepSeek. I'm at your service and would be delighted to assist you with any inquiries or tasks you may have.\n</think>\n\nGreetings! I'm DeepSeek-R1, an artificial intelligence assistant created by DeepSeek. I'm at your service and would be delighted to assist you with any inquiries or tasks you may have.","role":"assistant"},"finish_reason":"stop"}],"created":1741254044,"model":"DeepSeek-R1","object":"chat.completion","usage":{"completion_tokens":83,"prompt_tokens":16,"total_tokens":99}}
Is there any special configuration I missed for OpenAI-API-compatible Model?

✔️ Error log

OUTPUT data:

{
  "text": "",
  "usage": {
    "prompt_tokens": 0,
    "prompt_unit_price": "0",
    "prompt_price_unit": "0",
    "prompt_price": "0",
    "completion_tokens": 0,
    "completion_unit_price": "0",
    "completion_price_unit": "0",
    "completion_price": "0",
    "total_tokens": 0,
    "total_price": "0",
    "currency": "USD",
    "latency": 0.9801772950013401
  },
  "finish_reason": "Non-JSON encountered."
}

Originally created by @TinyFatty on GitHub (Mar 6, 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 [Dify issues](https://github.com/langgenius/dify/issues) & [Dify Official Plugins](https://github.com/langgenius/dify-official-plugins/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 1.0.0 ### Plugin version Built-in LLM block（I don't see any version here） ### Cloud or Self Hosted Self Hosted (Docker) ### Steps to reproduce ### 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 1.0.0 ### Cloud or Self Hosted Self Hosted (Docker) ### Steps to reproduce I used OpenAI-API-compatible Models (offered by China Telecom) in workflow LLM blocks and it throws "Non-JSON encountered." error (3 blocks share the same configuration, Deepseek official api and Silconflow deepseek api return the correct answer). Here is the configuration for all  Here is the error message  OUTPUT data: ``` { "text": "", "usage": { "prompt_tokens": 0, "prompt_unit_price": "0", "prompt_price_unit": "0", "prompt_price": "0", "completion_tokens": 0, "completion_unit_price": "0", "completion_price_unit": "0", "completion_price": "0", "total_tokens": 0, "total_price": "0", "currency": "USD", "latency": 0.9801772950013401 }, "finish_reason": "Non-JSON encountered." } ``` It takes only 1 second and no token is used,so I think the error occurs before the model starts reasoning. Then I tested this model with python scripts ``` import openai client= openai.OpenAI(base_url="***", api_key="***") response = client.chat.completions.create( model="7ba7726dad4c4ea4ab7f39c7741aea68", messages=[ {"role": "system", "content": "You are a helpful assistant."}, {"role": "user", "content": "Who are you?"} ], temperature=0.7 ) ``` It shows the corrent answer `{"code":0,"id":"chatcmpl-00-89c57718adba43dd5650511e4ecece0f-ba26f79321d736fa-01","choices":[{"index":0,"message":{"content":"<think>\nGreetings! I'm DeepSeek-R1, an artificial intelligence assistant created by DeepSeek. I'm at your service and would be delighted to assist you with any inquiries or tasks you may have.\n</think>\n\nGreetings! I'm DeepSeek-R1, an artificial intelligence assistant created by DeepSeek. I'm at your service and would be delighted to assist you with any inquiries or tasks you may have.","role":"assistant"},"finish_reason":"stop"}],"created":1741254044,"model":"DeepSeek-R1","object":"chat.completion","usage":{"completion_tokens":83,"prompt_tokens":16,"total_tokens":99}}` Is there any special configuration I missed for OpenAI-API-compatible Model? ### ✔️ Error log OUTPUT data: ``` { "text": "", "usage": { "prompt_tokens": 0, "prompt_unit_price": "0", "prompt_price_unit": "0", "prompt_price": "0", "completion_tokens": 0, "completion_unit_price": "0", "completion_price_unit": "0", "completion_price": "0", "total_tokens": 0, "total_price": "0", "currency": "USD", "latency": 0.9801772950013401 }, "finish_reason": "Non-JSON encountered." } ```

yindo added the bug label 2026-02-16 10:17:51 -05:00

yindo closed this issue 2026-02-16 10:17:51 -05:00

yindo commented 2026-02-16 10:17:51 -05:00

Author

Owner

Copy Link

Copy Source

@TinyFatty commented on GitHub (Mar 6, 2025):

Additional Information。
I rolled back to ver 0.15.3. The same model worked fine.
So I think maybe ver 1.0.0 changed something and didn't mention it.

@TinyFatty commented on GitHub (Mar 6, 2025): Additional Information。 I rolled back to ver 0.15.3. The same model worked fine. So I think maybe ver 1.0.0 changed something and didn't mention it.

yindo commented 2026-02-16 10:17:51 -05:00

Author

Owner

Copy Link

Copy Source

@ccll commented on GitHub (Mar 6, 2025):

Same here with 1.0.0, and works as @TinyFatty says in 0.15.3

The model API works with curl and the returned streaming JSONL is correct.

@ccll commented on GitHub (Mar 6, 2025): Same here with 1.0.0, and works as @TinyFatty says in 0.15.3 The model API works with `curl` and the returned streaming JSONL is correct.

yindo commented 2026-02-16 10:17:51 -05:00

Author

Owner

Copy Link

Copy Source

@oiisongyu commented on GitHub (Mar 13, 2025):

Same here.
In the Dify version 1.0.0 and above, there is a new parameter named "API endpoint中的模型名称" in the openai-api-compatible configuration page. However, in version 0.15.3, this parameter doesn't exist, and the Q&A functionality works properly.

@oiisongyu commented on GitHub (Mar 13, 2025): Same here. In the Dify version 1.0.0 and above, there is a new parameter named "API endpoint中的模型名称" in the openai-api-compatible configuration page. However, in version 0.15.3, this parameter doesn't exist, and the Q&A functionality works properly.

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@zouxiaodong commented on GitHub (Mar 14, 2025):

same here

@zouxiaodong commented on GitHub (Mar 14, 2025): same here

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@westdata commented on GitHub (Mar 15, 2025):

same here

@westdata commented on GitHub (Mar 15, 2025): same here

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@zouxiaodong commented on GitHub (Mar 16, 2025):

same error after upgrading to version 1.0.1, model using ctyun xirang deepseek-r1 :https://www.ctyun.cn/act/xirang/deepseek

@zouxiaodong commented on GitHub (Mar 16, 2025): same error after upgrading to version 1.0.1, model using ctyun xirang deepseek-r1 :https://www.ctyun.cn/act/xirang/deepseek

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@Lamber-maybe commented on GitHub (Mar 17, 2025):

same here

@Lamber-maybe commented on GitHub (Mar 17, 2025): same here

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@zouxiaodong commented on GitHub (Mar 17, 2025):

The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.

@zouxiaodong commented on GitHub (Mar 17, 2025):  The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@lastchiliarch commented on GitHub (Mar 20, 2025):

same problem. any progress?

@lastchiliarch commented on GitHub (Mar 20, 2025): same problem. any progress?

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@XiaoCC commented on GitHub (Mar 24, 2025):

The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.

提交一个pr修复吧

@XiaoCC commented on GitHub (Mar 24, 2025): >  > > The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang. 提交一个pr修复吧

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@Aniwine commented on GitHub (Mar 25, 2025):

The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.问题的根本原因是秋云西郎返回的数据在 “data：” 后面没有空格。但是，源代码当前假定 “data：” 后面会有一个空格。为了解决这个问题，我们可以修改源码，去掉对 “data：” 后面空格的依赖，保证与秋云西郎提供的数据格式兼容。

Hi, thanks for your solution, but where is the file?

@Aniwine commented on GitHub (Mar 25, 2025): >  > > The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.问题的根本原因是秋云西郎返回的数据在 “data：” 后面没有空格。但是，源代码当前假定 “data：” 后面会有一个空格。为了解决这个问题，我们可以修改源码，去掉对 “data：” 后面空格的依赖，保证与秋云西郎提供的数据格式兼容。 Hi, thanks for your solution, but where is the file?

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@Aniwine commented on GitHub (Mar 26, 2025):

The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.问题的根本原因是秋云西郎返回的数据在 “data：” 后面没有空格。但是，源代码当前假定 “data：” 后面会有一个空格。为了解决这个问题，我们可以修改源码，去掉对 “data：” 后面空格的依赖，保证与秋云西郎提供的数据格式兼容。

Hi, thanks for your solution, but where is the file?

I have already solved it by this method, in plugin-daemon's container,find the openai_compatible's llm.py of the package called dify-plugin,which located in langgenius/openai_api_compatible-0.0.11@410445eba2fa0f693d26dea2c3b9ffe51ad0777e021146ff877af6098412efc7/.venv/lib/python3.12/site-packages/dify_plugin/interfaces/model/openai_compatible/llm.py.Then modify it as above, it works for me!!

@Aniwine commented on GitHub (Mar 26, 2025): > >  > > The root cause of the issue is that the data returned by Ctyun Xirang does not include a space after "data:". However, the source code currently assumes that there will be a space after "data:". To fix this, we can modify the source code to remove the dependency on the space after "data:", ensuring compatibility with the data format provided by Ctyun Xirang.问题的根本原因是秋云西郎返回的数据在 “data：” 后面没有空格。但是，源代码当前假定 “data：” 后面会有一个空格。为了解决这个问题，我们可以修改源码，去掉对 “data：” 后面空格的依赖，保证与秋云西郎提供的数据格式兼容。 > > Hi, thanks for your solution, but where is the file? I have already solved it by this method, in plugin-daemon's container,find the openai_compatible's llm.py of the package called dify-plugin,which located in langgenius/openai_api_compatible-0.0.11@410445eba2fa0f693d26dea2c3b9ffe51ad0777e021146ff877af6098412efc7/.venv/lib/python3.12/site-packages/dify_plugin/interfaces/model/openai_compatible/llm.py.Then modify it as above, it works for me!!

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@Osama-Faza commented on GitHub (Jun 15, 2025):

I'm facing this issue on the cloud Dify website, version 1.4.2.
Is there anything I can do to fix the problem from my end, or do I have to wait for a pull request to be made and merged?

@Osama-Faza commented on GitHub (Jun 15, 2025): I'm facing this issue on the cloud Dify website, version 1.4.2. Is there anything I can do to fix the problem from my end, or do I have to wait for a pull request to be made and merged?

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

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

Same here.
Please solve it

@dwdcth commented on GitHub (Jul 14, 2025): Same here. Please solve it

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@shun-qi commented on GitHub (Jul 22, 2025):

same here
please !!!!!

@shun-qi commented on GitHub (Jul 22, 2025): same here please !!!!!

yindo commented 2026-02-16 10:17:52 -05:00

Author

Owner

Copy Link

Copy Source

@dosubot[bot] commented on GitHub (Aug 21, 2025):

Hi, @TinyFatty. I'm Dosu, and I'm helping the dify-official-plugins team manage their backlog and am marking this issue as stale.

Issue Summary:

• You reported a "Non-JSON encountered." error with China Telecom OpenAI-API-compatible models on Dify v1.0.0, while v0.15.3 worked fine.
• I identified the root cause as the source code assuming a space after "data:" in the returned data, which was not always present.
• I found a workaround by modifying the llm.py file in the dify-plugin package to remove the space dependency.
• Users on cloud versions like 1.4.2 also experienced this issue and are awaiting an official fix or PR.

Next Steps:

• Please confirm if this issue is still relevant with the latest version of dify-official-plugins by commenting here.
• If no further activity occurs, I will automatically close this issue in 5 days.

Thank you for your understanding and contribution!

@dosubot[bot] commented on GitHub (Aug 21, 2025): Hi, @TinyFatty. I'm [Dosu](https://dosu.dev), and I'm helping the dify-official-plugins team manage their backlog and am marking this issue as stale. **Issue Summary:** - You reported a "Non-JSON encountered." error with China Telecom OpenAI-API-compatible models on Dify v1.0.0, while v0.15.3 worked fine. - I identified the root cause as the source code assuming a space after "data:" in the returned data, which was not always present. - I found a workaround by modifying the llm.py file in the dify-plugin package to remove the space dependency. - Users on cloud versions like 1.4.2 also experienced this issue and are awaiting an official fix or PR. **Next Steps:** - Please confirm if this issue is still relevant with the latest version of dify-official-plugins by commenting here. - If no further activity occurs, I will automatically close this issue in 5 days. Thank you for your understanding and contribution!

yindo referenced this issue 2026-02-16 10:21:41 -05:00

[PR #45] [MERGED] Docs:Add Novita-AI Readme #1020

Sign in to join this conversation.

No Branch/Tag Specified

Branches Tags

main

feat/3479-gemini-flash-models

dependabot/uv/tools/chart/uv-cdf8d38825

feat/moonshot-kimi-k3-k27-code

feat/gemini-system-instruction-semantics

fix/gemini-text-system-parts

dosu/doc-update-3436

feat/openai-gpt-5-6-support

fix/outlook-trigger-readme-permissions

fix/microsoft-oauth-datasource-tenant

fix/byteplus-api-base-url-label

feat/ark-video-generation-custom-models

fix/openai-stream-reasoning-thinking

fix/skip-live-tests-missing-keys

codex/enhance-discord-integrations

codex/harden-trigger-oauth-permissions

update-lemonade-default-port

codex/fix-pr-review-ci

feat/ark-multimodal-polling

copilot/update-mineru-3x-plugin-file-types

feat/action-security-fix

feat/claude-fable5

codex/fix-ollama-model-category

codex/upgrade-ollama-latest

feat/minimax-m3

codex/refine-plugin-readmes

fix/fireworks-model-label-zh-hans

chore/version-bump-plugins-batch-06

chore/version-bump-plugins-batch-05

chore/version-bump-plugins-batch-04

chore/version-bump-plugins-batch-03

chore/version-bump-plugins-batch-02

fix/plugin-sdk-i18n-fields

chore/version-bump-plugins-batch-01

chore/dependencies-high-priority

chore/dependencies-medium-priority

chore/dependencies-low-priority-clean

chore/dependencies-low-priority

fix/tongyi-align-deepseek-model-metadata

codex/update-openrouter-models

codex/azure-ai-studio-cjk-version-bump

feat/ernie-image-tool

claude/cranky-thompson-b6a281

fix/remove-unavailable-volcengine-models

fix/legacy-plugin-daemon-compatibility

copilot/add-deepseek-v4-to-openrouter

codex/merge-tongyi-tts-prs

codex/depot-runners

chore/migrate-selected-plugins-to-uv-lock

fix/repair-failed-plugin-deploys

ci/fix-plugin-change-detection

chore/migrate-aws-s3-storage-to-uv-lock

chore/migrate-somark-dependencies-to-uv-lock

ci/uv-lock-dependency-validation

fix/2911-openai-api-compatible-tool-call-content

chore/remove-deprecated-plugins

copilot/add-more-models-openrouter-provider

copilot/fix-error-connecting-fireworks-ai

fix/openai-compatible-max-completion-tokens-validation

fix/gemini-lockfile

chore/add-React-fallback-desc

feat/add-gemini-embedding-2-preview

copilot/add-claude-sonnet-opus-4-6

codex/update-xai-models-2721

fix/remove-sentence-transformer-for-hubspot

feat/add-gpt-5.4

copilot/fix-ci-errors-and-bump-version

copilot/fix-security-errors-combined

copilot/update-gemini-model-version

fix/anthropic-parallel-tool-call

codex/fix-azure-ai-studio-azure-core-1-38

codex/fix-azure-blob-azure-core-1-38

codex/fix-azure-openai-azure-core-1-38

gitlab-credentials-c

fix/openrouter-validation-model

fix/gemini-null-checks

fix/deepseek-tool-message-ignore

fix/gemini-dependencies

No results found.

Labels

Clear labels

bug

bug

bug

dependencies

documentation

enhancement

good first issue

invalid

pull-request

Mirrored from GitHub Pull Request

question

stale

No Label bug

Milestone

No items

No Milestone

Projects

Clear projects

No project

Assignees

Clear assignees

yindo

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: langgenius/dify-official-plugins#45