The solution that worked for me is importing Inference client directly.The HuggingFaceEndpoint,HuggingFacePipeline doesnt seem to be working for the new version of huggingface_hub.
from huggingface_hub import InferenceClient
client = InferenceClient(
"mistralai/Mistral-7B-Instruct-v0.2", #"openai/gpt-oss-120b"
token=os.environ["HUGGINGFACEHUB_API_TOKEN"]
)
resp = client.chat.completions.create(
model="mistralai/Mistral-7B-Instruct-v0.2", #"openai/gpt-oss-120b"
messages=[
{"role": "user", "content": "What is the health insurance coverage?"}
],
max_tokens=1000,
temperature=0.1
)
print(resp.choices[0].message["content"])