Update app.py

app.py

@@ -1,4 +1,3 @@
-
 import gradio as gr
 import faiss
 import numpy as np
@@ -6,12 +5,10 @@ import openai
 from sentence_transformers import SentenceTransformer
 from nltk.tokenize import sent_tokenize
 import nltk
-from transformers import AutoTokenizer, AutoModel
 import torch
 
 # Download the required NLTK data
 nltk.download('punkt')
-nltk.download('punkt_tab')
 
 # Paths to your files
 faiss_path = "manual_chunked_faiss_index_500.bin"
@@ -52,19 +49,14 @@ except Exception as e:
     raise RuntimeError(f"Failed to load FAISS index: {e}")
 
 # Load the tokenizer and model for embeddings
-embedding_tokenizer = AutoTokenizer.from_pretrained("microsoft/MiniLM-L12-H384-uncased")
-embedding_model = AutoModel.from_pretrained("microsoft/MiniLM-L12-H384-uncased")
+embedding_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')
 
 # OpenAI API key
-openai.api_key =  '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'
+openai.api_key = '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'
 
 # Function to create embeddings
 def embed_text(text_list):
-    inputs = embedding_tokenizer(text_list, padding=True, truncation=True, return_tensors="pt")
-    with torch.no_grad():
-        outputs = embedding_model(**inputs)
-    embeddings = outputs.last_hidden_state[:, 0, :].cpu().numpy()  # Use the CLS token representation
-    return embeddings
+    return np.array(embedding_model.encode(text_list), dtype=np.float32)
 
 # Function to retrieve relevant chunks for a user query
 def retrieve_chunks(query, k=5):
@@ -78,51 +70,50 @@ def retrieve_chunks(query, k=5):
         raise RuntimeError(f"FAISS search failed: {e}")
     
     if len(indices[0]) == 0:
-        return []
+        return [], distances, indices
 
     valid_indices = [i for i in indices[0] if i < len(manual_chunks)]
     if not valid_indices:
-        return []
+        return [], distances, indices
 
     relevant_chunks = [manual_chunks[i] for i in valid_indices]
-    return relevant_chunks
-
-# Load the tokenizer and model for generation
-generator_tokenizer = AutoTokenizer.from_pretrained("gpt-3.5-turbo")  # Replace with correct tokenizer if needed
-generator_model = AutoModel.from_pretrained("gpt-3.5-turbo")  # Replace with correct model if needed
-
-# Function to truncate long inputs
-def truncate_input(text, max_length=512):
-    inputs = generator_tokenizer(text, return_tensors="pt", truncation=True, max_length=max_length)
-    return inputs
+    return relevant_chunks, distances, indices
 
 # Function to perform RAG: Retrieve chunks and generate a response
-def rag_response(query, k=5, max_new_tokens=150):
+def rag_response(query, k=5, max_tokens=150):
     try:
-        relevant_chunks = retrieve_chunks(query, k=k)
+        relevant_chunks, distances, indices = retrieve_chunks(query, k=k)
         
         if not relevant_chunks:
-            return "Sorry, I couldn't find relevant information."
+            return "Sorry, I couldn't find relevant information.", distances, indices
 
         augmented_input = query + "\n" + "\n".join(relevant_chunks)
         
-        inputs = truncate_input(augmented_input)
-        
-        # Generate response
-        outputs = generator_model.generate(inputs['input_ids'], max_new_tokens=max_new_tokens)
-        generated_text = generator_tokenizer.decode(outputs[0], skip_special_tokens=True)
-        
-        return generated_text
+        # Generate response using OpenAI API
+        response = openai.Completion.create(
+            model="gpt-3.5-turbo",
+            prompt=augmented_input,
+            max_tokens=max_tokens,
+            temperature=0.7
+        )
+        generated_text = response.choices[0].text.strip()
+        return generated_text, distances, indices
     except Exception as e:
-        return f"An error occurred: {e}"
+        return f"An error occurred: {e}", [], []
 
 # Gradio Interface
+def format_output(response, distances, indices):
+    # Format output to include distances and indices
+    formatted_response = f"Response: {response}\n\nDistances: {distances}\n\nIndices: {indices}"
+    return formatted_response
+
 iface = gr.Interface(
     fn=rag_response,
     inputs="text",
     outputs="text",
     title="RAG Chatbot with FAISS and GPT-3.5",
-    description="Ask me anything!"
+    description="Ask me anything!",
+    live=True
 )
 
 if __name__ == "__main__":