Update app.py

app.py

@@ -76,10 +76,9 @@ languages = {
 components = {}
 dotenv.load_dotenv()
 seamless_client = Client("facebook/seamless_m4t")
+mistralmed_client = Client("https://tonic1-mistralmed-chat.hf.space/--replicas/crzkn/")
 HuggingFace_Token = os.getenv("HuggingFace_Token")
 hf_token = os.getenv("HuggingFace_Token")
-base_model_id = os.getenv('BASE_MODEL_ID', 'default_base_model_id')
-model_directory = os.getenv('MODEL_DIRECTORY', 'default_model_directory')
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
 image_description = "" 
@@ -364,75 +363,17 @@ def query_vectara(text):
         return f"Error: {response.status_code}"
 
 
-# Functions to Wrap the Prompt Correctly
-def wrap_text(text, width=90):
-    lines = text.split('\n')
-    wrapped_lines = [textwrap.fill(line, width=width) for line in lines]
-    wrapped_text = '\n'.join(wrapped_lines)
-    return wrapped_text
-
-
-def multimodal_prompt(user_input, system_prompt="You are an expert medical analyst:"):
-
-    # Combine user input and system prompt
-    formatted_input = f"{user_input}{system_prompt}"
-
-    # Encode the input text
-    encodeds = tokenizer(formatted_input, return_tensors="pt", add_special_tokens=False)
-    model_inputs = encodeds.to(device)
-
-    # Generate a response using the model //MODEL UNDEFINED, using peft_model instead.
-    output = peft_model.generate(
-        **model_inputs,
-        max_length=512,
-        use_cache=True,
-        early_stopping=True,
-        bos_token_id=peft_model.config.bos_token_id,
-        eos_token_id=peft_model.config.eos_token_id,
-        pad_token_id=peft_model.config.eos_token_id,
-        temperature=0.1,
-        do_sample=True
-    )
-
-    # Decode the response
-    response_text = tokenizer.decode(output[0], skip_special_tokens=True)
-
-    return response_text
-
-
-# Instantiate the Tokenizer
-tokenizer = AutoTokenizer.from_pretrained("stabilityai/stablelm-3b-4e1t", token=hf_token, trust_remote_code=True, padding_side="left")
-# tokenizer = AutoTokenizer.from_pretrained("Tonic/stablemed", trust_remote_code=True, padding_side="left")
-tokenizer.pad_token = tokenizer.eos_token
-tokenizer.padding_side = 'left'
-
-# Load the PEFT model
-peft_config = PeftConfig.from_pretrained("Tonic/stablemed", token=hf_token)
-peft_model = AutoModelForCausalLM.from_pretrained("stabilityai/stablelm-3b-4e1t", token=hf_token, trust_remote_code=True)
-peft_model = PeftModel.from_pretrained(peft_model, "Tonic/stablemed", token=hf_token)
-
-
-class ChatBot:
-    def __init__(self):
-        self.history = []
-        
-    @staticmethod
-    def doctor(user_input, system_prompt="You are an expert medical analyst:"):
-        formatted_input = f"{system_prompt}{user_input}"
-        user_input_ids = tokenizer.encode(formatted_input, return_tensors="pt")
-        response = peft_model.generate(input_ids=user_input_ids, max_length=512, pad_token_id=tokenizer.eos_token_id)
-        response_text = tokenizer.decode(response[0], skip_special_tokens=True)
-        return response_text
-
-
-bot = ChatBot()
-
-
 def process_summary_with_stablemed(summary):
-    system_prompt = "You are a medical instructor . Assess and describe the proper options to your students in minute detail. Propose a course of action for them to base their recommendations on based on your description."
-    response_text = bot.doctor(summary, system_prompt)
+    system_prompt = "You are a medical instructor. Assess and describe the proper options to your students in minute detail. Propose a course of action for them to base their recommendations on based on your description."
+    # Use the Mistral Med Gradio client API call
+    result = mistralmed_client.predict(
+        summary,  # Summary text
+        system_prompt,  # System prompt
+        api_name="/predict"
+    )
+    # Assuming the result is the response text
+    response_text = result if isinstance(result, str) else "Error in processing"
     return response_text
-
     
 # Main function to handle the Gradio interface logic