import gradio as gr
from transformers import AutoTokenizer, AutoModelForCausalLM, AutoModel

# Load Med42
med42_model_name = "m42-health/med42-70b"
med42_tokenizer = AutoTokenizer.from_pretrained(med42_model_name)
med42_model = AutoModelForCausalLM.from_pretrained(med42_model_name)

# Load ClinicalBERT
clinicalbert_model_name = "medicalai/ClinicalBERT"
clinicalbert_tokenizer = AutoTokenizer.from_pretrained(clinicalbert_model_name)
clinicalbert_model = AutoModel.from_pretrained(clinicalbert_model_name)

# Define functions
def med42_qa(question):
    inputs = med42_tokenizer(question, return_tensors="pt")
    outputs = med42_model.generate(**inputs, max_length=200)
    return med42_tokenizer.decode(outputs[0], skip_special_tokens=True)

def analyze_ehr(text):
    inputs = clinicalbert_tokenizer(text, return_tensors="pt")
    embeddings = clinicalbert_model(**inputs).last_hidden_state
    return f"ClinicalBERT generated embeddings with shape: {embeddings.shape}"

# Combine Gradio Interface
def combined_tool(input_text):
    qa_result = med42_qa(input_text)
    ehr_result = analyze_ehr(input_text)
    return f"Med42 Answer:\n{qa_result}\n\nClinicalBERT Analysis:\n{ehr_result}"

# Build Gradio UI
interface = gr.Interface(
    fn=combined_tool,
    inputs="text",
    outputs="text",
    title="Healthcare AI Tool",
    description="Use Med42 for medical Q&A and ClinicalBERT for EHR analysis."
)

if __name__ == "__main__":
    interface.launch()
import os
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load authentication token from environment variables
hf_token = os.getenv("HF_AUTH_TOKEN")

# Load the Med42 model with the token
med42_model_name = "m42-health/med42-70b"
med42_tokenizer = AutoTokenizer.from_pretrained(med42_model_name, use_auth_token=hf_token)
med42_model = AutoModelForCausalLM.from_pretrained(med42_model_name, use_auth_token=hf_token)
import os
from transformers import AutoTokenizer, AutoModelForCausalLM

# Load the token from the environment
hf_token = os.getenv("HF_AUTH_TOKEN")

# Load Med42 model and tokenizer
med42_model_name = "m42-health/med42-70b"
med42_tokenizer = AutoTokenizer.from_pretrained(med42_model_name, use_auth_token=hf_token)
med42_model = AutoModelForCausalLM.from_pretrained(med42_model_name, use_auth_token=hf_token)