Update app.py

app.py

@@ -13,51 +13,8 @@ from openai import OpenAI
 
 st.set_page_config(layout="wide")
 
-
-my_initial_rag_text = f"""This is a RAG (Retrieval-Augmented Generation) chatbot application built with Streamlit that combines document context with LLM responses. Here's a breakdown of its main components:
-
-1. Initial Setup:
-- Uses Streamlit for the web interface
-- Employs SentenceTransformer for generating embeddings
-- Uses HuggingFace's InferenceClient for LLM interaction
-
-2. State Management:
-- Maintains several session state variables for:
-  - RAG text (the context document)
-  - LLM model selection
-  - Chat message history
-  - Embeddings and sentences
-  - HuggingFace client
-
-3. Interface Layout:
-- Split into two columns (1:2 ratio):
-  - Left column: Model selection dropdown and RAG text input
-  - Right column: Chat interface and message history
-
-4. Core Functionality:
-- RAG Implementation:
-  - Splits the context document into sentences
-  - When a user asks a question, it:
-    - Converts the question into embeddings
-    - Finds the 3 most relevant sentences from the context
-    - Adds these relevant pieces as context to the prompt
-
-- Chat Interface:
-  - Streams responses from the LLM
-  - Maintains a chat history
-  - Shows message history and augmented prompts
-  - Uses different avatars for user and AI messages
-
-5. Model Options:
-- Offers three LLM choices:
-  - Mistral-7B-Instruct-v0.3 (default)
-  - Qwen2.5-72B-Instruct
-  - Zephyr-7b-beta
-
-The application essentially creates an intelligent chatbot that can answer questions while taking into account the context provided in the RAG text, making it particularly useful for domain-specific Q&A scenarios.
-
-A disclaimer at the bottom reminds users about potential LLM inaccuracies and the need for verification of responses.
-"""
+with open("/app/labeled_text_small.txt", "r", encoding="utf-8") as f:
+    my_initial_rag_text = f.read()
 
 # Check if the LLM model is not already in the session state
 if "my_llm_model" not in st.session_state: