Customer AI Provider - Agent Issue "Hit error when calling LLM"

Hi,

I am using groq as my Custom AI Resource connecting to openai/gpt-oss-20b.

This connection works fine for basic workflows etc, but when I test using it in a Agent environment I get the following error "Hit error when calling LLM: Invalid text action response data".

Anything I am missing here? The groq website says its supported. OpenAI GPT-OSS 20B - GroqDocs

Picture below of how the connection is setup -


Thanks

It looks like the issue is that Groq’s openai/gpt-oss-20b endpoint doesn’t return the exact JSON structure Retool’s “OpenAI” provider schema expects in Agent mode. The schema is strict — it expects a choices[0].message.content or choices[0].text field.

For Groq models, you may need to:

  • Change the Provider schema from OpenAI to Custom so you can map the exact response format.
  • Or, if you stick with OpenAI schema, wrap the Groq API call in a small serverless function / proxy that reformats the Groq response to match OpenAI’s.

Groq’s API for gpt-oss-20b often returns fields slightly differently than OpenAI’s, which is why basic calls might work but the Agent environment (which parses the response automatically) fails with Invalid text action response data.

If you want to stay fully no-code, the quickest path is switching this resource to Custom and updating the schema to match Groq’s exact JSON.

1 Like

Hi David,

Sorry for the late reply on this. Let me try testing out this week and let you know how I go.

Thanks!

2 Likes

Hi @RA_SYD,

Any updates on if you were able to implement @DavidTech's solution?