The agent reports an error using the gpt-4o-mini model

[yunfeng13]

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

0.15.0

Cloud or Self Hosted

Self Hosted (Docker)

Steps to reproduce

error detail：[openai] Bad Request Error, Error code: 400 - {'error': {'message': "Invalid value for 'content': expected a string, got null. (request id: 202501202012127715196675TbPifJ3) (request id: 20250120201212658215175TCApZ8tE) (request id: 2025012020121243955128724364135) (request id: 2025012020121234417090004257917) (request id: 20250120201212335003074lycgz9cp)", 'type': 'invalid_request_error', 'param': 'messages.[4].content', 'code': None}}
I found that once the agent adds context and retrieves it, it will report an error if asked again.

✔️ Expected Behavior

I hope to know the reason. If the context is not hit, it will not report an error. Once it hits, it will report an error.

❌ Actual Behavior

error detail：[openai] Bad Request Error, Error code: 400 - {'error': {'message': "Invalid value for 'content': expected a string, got null. (request id: 202501202012127715196675TbPifJ3) (request id: 20250120201212658215175TCApZ8tE) (request id: 2025012020121243955128724364135) (request id: 2025012020121234417090004257917) (request id: 20250120201212335003074lycgz9cp)", 'type': 'invalid_request_error', 'param': 'messages.[4].content', 'code': None}}

[dosubot]

I found a similar unsolved discussion titled Bug of Use Claude, which might be related to the issue you're experiencing ^[1].

The error you're encountering, "Invalid value for 'content': expected a string, got null," typically indicates that the 'content' field in the 'messages' object is not being populated correctly. This can happen if the context retrieval process fails to provide valid content, resulting in a null value being passed to the OpenAI API.

To address this issue, you might want to check the following:

1. Context Retrieval Logic: Ensure that the context retrieval logic is correctly implemented and that it returns valid string content. If the context is not found, make sure there is a fallback mechanism to handle such cases without resulting in a null value.

2. Error Handling: Implement error handling to catch and manage cases where the context retrieval returns null. This can prevent the null value from being included in the 'messages' object.

3. Debugging: Add logging or debugging statements to trace the values being assigned to the 'content' field in the 'messages' object. This can help identify where the null value is being introduced.

These steps should help you identify and resolve the issue with the 'content' field in the 'messages' object.

To continue talking to Dosu, mention @dosu.

---

^{Help Dosu learn! Give it feedback: Great Response | Irrelevant Answer | Incorrect Sources | Too Verbose | Hallucination | Other}

[laipz8200]

Hi @yunfeng13. Thank you for the report. Could you please provide your logs in the api Docker container? It will help us to locate this issue.

[yunfeng13]

This is a probability report of 400, but the content of 400 is the error report in the screenshot above. There is only 400 in the container.

[laipz8200]

That means some content fields your LLM API returned in the messages array have a None value, and we expect a string.

[yunfeng13]

Do you mean that the api returns a null value? Will returning null cause llm to report an error?

[laipz8200]

Yes, the path of the wrong result is messages[index].content, our data structure of this content expects a string, not None.

[yunfeng13]

Thank you very much. Then I will try another apikey.