Update app.py

app.py

@@ -5,7 +5,10 @@ sys.modules['sqlite3'] = sys.modules.pop('pysqlite3')
 from sentence_transformers import SentenceTransformer
 import chromadb
 from datasets import load_dataset
-from gpt4all import GPT4All
+from transformers import AutoModelForCausalLM, AutoTokenizer
+import gradio as gr
+import faiss
+
 
 # Embedding vector
 class VectorStore:
@@ -45,12 +48,31 @@ vector_store = VectorStore("embedding_vector")
 vector_store.populate_vectors(dataset)
 
 
-# loading gpt4all language model
-# load model Chat based model mistral-7b-openorca.gguf2.Q4_0.gguf
-# detail about gpt4all and model information - https://gpt4all.io/index.html
-model_name = 'Meta-Llama-3-8B-Instruct.Q4_0.gguf' # .gguf represents quantized model
-model_path = "gpt4all"
-# add path to download load the model locally, download once and load for subsequent inference
-model = GPT4All(model_name=model_name, model_path=model_path,device="cuda")
+# Load the model and tokenizer
+# text generation model
+model_name = "meta-llama/Meta-Llama-3-8B"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+
+# Define the chatbot response function
+def chatbot_response(user_input):
+    global conversation_history
+    results = vector_store.search_context(user_input, n_results=1)
+    context = results['documents'][0] if results['documents'] else ""
+    conversation_history.append(f"User: {user_input}\nContext: {context[:150]}\nBot:")
+    inputs = tokenizer("\n".join(conversation_history), return_tensors="pt")
+    outputs = model.generate(**inputs, max_length=150, do_sample=True, temperature=0.7)
+    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
+    conversation_history.append(response)
+    return response
+
+
+# Gradio interface
+def chat(user_input):
+    response = chatbot_response(user_input)
+    return response
+
+iface = gr.Interface(fn=chat, inputs="text", outputs="text")
+iface.launch()