Update app.py

app.py

@@ -14,6 +14,7 @@ from huggingface_hub import InferenceClient
 from langchain.llms.base import LLM
 from typing import Optional, List, Any
 from transformers import AutoTokenizer, AutoModelForCausalLM
+from transformers import pipeline
 
 # Initialize Flask application
 app = Flask(__name__)
@@ -142,19 +143,12 @@ def init_db():
         conn.commit()
 
 def initialize_llm():
-    """Initialize the LLM using InferenceClient."""
+    """Initialize the LLM using transformers pipeline."""
     try:
-        # Get API token from environment variable
-        api_token = os.environ.get('HF_TOKEN')
-        if not api_token:
-            raise ValueError("No API token found")
-
-        # Initialize the InferenceClient
-        client = InferenceClient(api_key=api_token)
-
+        model_name = "mistralai/Mistral-7B-Instruct-v0.3"
+        llm = pipeline("text-generation", model=model_name)
         print("LLM initialized successfully!")
-        return client
-
+        return llm
     except Exception as e:
         print(f"LLM initialization error: {str(e)}")
         return None
@@ -270,12 +264,7 @@ try:
     print("Initializing database...")
     init_db()
     print("Database initialized successfully")
-
-    print("Initializing LLM...")
-    llm_client = initialize_llm()
-    if llm_client is None:
-        raise ValueError("LLM initialization failed")
-    print("LLM initialized successfully")
+    llm = initialize_llm()
 
 
 
@@ -313,11 +302,12 @@ Output Guidelines:
    - Mark critical information with [IMPORTANT] prefix and give small explanations with some bold headings if required and in white font always.
 """
 
-prompt = PromptTemplate(
-    input_variables=["user_request", "chat_history", "important_info"],
-    template=prompt_template
-)
-llm_chain = LLMChain(llm=llm, prompt=prompt)
+def generate_prompt(user_request, chat_history, important_info):
+    return prompt_template.format(
+        user_request=user_request,
+        chat_history=chat_history,
+        important_info=important_info
+    )
 
 
 def convert_to_html(raw_text):
@@ -469,45 +459,27 @@ def chat():
         user_input = data.get("message", "")
         print(f"Received message: {user_input}")
 
-        if not user_input:
-            return jsonify({
-                "success": False,
-                "response": "No message provided."
-            })
+        # Very simple example for the chat history and important information (can be more complex)
+        chat_history = ""  # Assuming you maintain chat history somehow
+        important_info = "Provide code examples for Python programming."
 
-        # Very simple prompt for testing
-        messages = [
-            {"role": "user", "content": user_input}
-        ]
-
-        try:
-            # Get response from the model using InferenceClient
-            if llm_client:
-                completion = llm_client.chat.completions.create(
-                    model="mistralai/Mistral-7B-Instruct-v0.3",  # Replace with the appropriate model
-                    messages=messages,
-                    max_tokens=150  # Adjust the max tokens as needed
-                )
-                response = completion['choices'][0]['message']['content']
-                print(f"Raw response received: {response}")
-
-                # Return the model's response
-                return jsonify({
-                    "success": True,
-                    "response": response
-                })
+        # Generate the prompt
+        prompt = generate_prompt(user_input, chat_history, important_info)
 
-            else:
-                raise ValueError("LLM client not initialized properly.")
+        # Get response from the model using transformers pipeline
+        if llm:
+            response = llm(prompt, max_length=150)[0]['generated_text']
+            print(f"Raw response received: {response}")
 
-        except Exception as e:
-            print(f"Model error: {str(e)}")
-            # Return a simple error message
+            # Return the model's response
             return jsonify({
-                "success": False,
-                "response": "Sorry, I couldn't generate a response right now."
+                "success": True,
+                "response": response
             })
 
+        else:
+            raise ValueError("LLM not initialized properly.")
+
     except Exception as e:
         print(f"Request error: {str(e)}")
         return jsonify({