Update app.py

app.py

@@ -9,7 +9,6 @@ import nltk
 # Download the required NLTK data
 nltk.download('punkt')
 nltk.download('punkt_tab')
-
 # Paths to your files
 faiss_path = "manual_chunked_faiss_index_500.bin"
 manual_path = "ubuntu_manual.txt"
@@ -52,7 +51,7 @@ except Exception as e:
 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):
@@ -79,6 +78,11 @@ def retrieve_chunks(query, k=5):
     relevant_chunks = [manual_chunks[i] for i in valid_indices]
     return relevant_chunks, distances, indices
 
+# Function to truncate long inputs
+def truncate_input(text, max_length=16385):
+    tokens = tokenizer.encode(text, truncation=True, max_length=max_length, return_tensors="pt")
+    return tokens
+
 # Function to perform RAG: Retrieve chunks and generate a response
 def rag_response(query, k=5, max_tokens=150):
     try:
@@ -87,8 +91,15 @@ def rag_response(query, k=5, max_tokens=150):
         if not relevant_chunks:
             return "Sorry, I couldn't find relevant information.", distances, indices
 
-        augmented_input = query + "\n" + "\n".join(relevant_chunks)
+        # Combine the query with retrieved chunks
+        augmented_input = query + "\n\n" + "\n\n".join(relevant_chunks)
         
+        # Truncate the input if it exceeds token limits
+        input_tokens = tokenizer.encode(augmented_input, return_tensors="pt")
+        if input_tokens.shape[1] > 16385:
+            # Truncate to fit within the model's maximum input length
+            augmented_input = tokenizer.decode(input_tokens[0, :16385])
+
         # Generate response using OpenAI API
         response = openai.ChatCompletion.create(
             model="gpt-3.5-turbo",