watsonx.ai inference provider is not working for me #3165
Description
System Info
llama_stack==0.2.17
llama_stack_client==0.2.17
Information
- The official example scripts
- My own modified scripts
🐛 Describe the bug
When I try to call chat completions or responses from the LLS client referencing a watsonx.ai model and provider, I get a 500 error on the client and a 'WatsonXInferenceAdapter' object has no attribute '_openai_client' error on the server logs.
Error logs
INFO 2025-08-15 14:44:28,875 console_span_processor:65 telemetry: 18:44:28.848 [ERROR] Error executing endpoint route='/v1/openai/v1/responses'
method='post': 'WatsonXInferenceAdapter' object has no attribute '_openai_client'
and also:
ERROR 2025-08-15 14:47:16,968 __main__:244 server: Error executing endpoint route='/v1/openai/v1/chat/completions' method='post':
'WatsonXInferenceAdapter' object has no attribute '_openai_client'
Expected behavior
When I run
client.chat.completions.create(
model=LLAMA_STACK_MODEL_ID,
messages=[{"role": "user", "content": "What is the capital of France?"}]
)
or
client.responses.create(
model=LLAMA_STACK_MODEL_ID,
input="What is the capital of France?"
)
with LLAMA_STACK_MODEL_ID referring to a watsonx model, I should get a valid response (since I have a watsonx.ai API key and have configured my run.yaml to use it).