ChatClient returns 401 Invalid API Key when using Groq as OpenAI-compatible provider

Bug description
I'm trying to use [Groq](https://groq.com) as an OpenAI-compatible provider with Spring AI (spring.ai.openai) in version 1.0.0. I have correctly configured the api-key and base-url. The same API key works as expected using curl, but Spring AI consistently returns a 401 - Invalid API Key when invoking ChatClient.

Environment

• Spring AI version: 1.0.0
• Spring Boot version: 3.2.5
• Java version: 17
• Model: llama3-70b-8192
• Groq base URL: https://api.groq.com/openai
• API key format: gsk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx (valid and confirmed via curl)
• No vector store is used

Steps to reproduce

1. Configure application.yml as below:

   spring:
     ai:
       openai:
         api-key: gsk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
         base-url: https://api.groq.com/openai
         chat:
           options:
             model: llama3-70b-8192
             temperature: 0.7
   

2. Inject and use ChatClient:

   @Autowired
   private ChatClient chatClient;
   
   public String callGroq(String prompt) {
       return chatClient.prompt()
               .system("You are a helpful assistant.")
               .user(prompt)
               .call()
               .content();
   }
   

3. Call the method.

Expected behavior
A valid chat completion response from Groq, the same as received when making the request via curl.

Minimal Complete Reproducible example

Here’s the curl that works:

curl https://api.groq.com/openai/v1/chat/completions -s \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer gsk_xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx" \
  -d '{
    "model": "llama3-70b-8192",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

This returns a valid response.

But with Spring AI, I receive this:

401 - {
  "error": {
    "message": "Invalid API Key",
    "type": "invalid_request_error",
    "code": "invalid_api_key"
  }
}

Please advise if any custom headers or additional configuration is required to support Groq in Spring AI.

Let me know if you'd like to include a GitHub repo or sample project link as well to improve reproducibility.

Start with the ChatClient invocation and the application.yml configuration, then compare the request sent by Spring AI with the working curl request to Groq. Done means the same configuration produces a valid Groq chat completion without the 401 Invalid API Key response.