app.py

# app_pure_rag.py
import numpy as np
import faiss
import gradio as gr
from langchain.text_splitter import CharacterTextSplitter
from sentence_transformers import SentenceTransformer

# --- Load and Prepare Data ---
with open("gen_agents.txt", "r", encoding="utf-8") as f:
    full_text = f.read()

# Split text into passages
text_splitter = CharacterTextSplitter(separator="\n\n", chunk_size=512, chunk_overlap=20)
docs = text_splitter.create_documents([full_text])
passages = [doc.page_content for doc in docs]

# Initialize embedder and build FAISS index
embedder = SentenceTransformer('all-MiniLM-L6-v2')
passage_embeddings = embedder.encode(passages, convert_to_tensor=False, show_progress_bar=True)
passage_embeddings = np.array(passage_embeddings).astype("float32")
d = passage_embeddings.shape[1]
index = faiss.IndexFlatL2(d)
index.add(passage_embeddings)

# --- Provided Functions ---
def retrieve_passages(query, embedder, index, passages, top_k=3):
    """
    Retrieve the top-k most relevant passages based on the query.
    """
    query_embedding = embedder.encode([query], convert_to_tensor=False)
    query_embedding = np.array(query_embedding).astype('float32')
    distances, indices = index.search(query_embedding, top_k)
    retrieved = [passages[i] for i in indices[0]]
    return retrieved

# --- Gradio App Function ---
def get_pure_rag_output(query):
    retrieved = retrieve_passages(query, embedder, index, passages, top_k=3)
    rag_text = "\n".join([f"Passage {i+1}: {p}" for i, p in enumerate(retrieved)])
    # Wrap text in a styled div
    return f"<div style='white-space: pre-wrap;'>{rag_text}</div>"

def clear_output():
    return ""

# --- Custom CSS for a ChatGPT-like Dark Theme ---
custom_css = """
body {
    background-color: #343541 !important;
    color: #ECECEC !important;
    margin: 0;
    padding: 0;
    font-family: 'Inter', sans-serif;
}
#container {
    max-width: 900px;
    margin: 0 auto;
    padding: 20px;
}
label {
    color: #ECECEC;
    font-weight: 600;
}
textarea, input {
    background-color: #40414F;
    color: #ECECEC;
    border: 1px solid #565869;
}
button {
    background-color: #565869;
    color: #ECECEC;
    border: none;
    font-weight: 600;
    transition: background-color 0.2s ease;
}
button:hover {
    background-color: #6e7283;
}
.output-box {
    border: 1px solid #565869;
    border-radius: 4px;
    padding: 10px;
    margin-top: 8px;
    background-color: #40414F;
}
"""

# --- Build Gradio Interface ---
with gr.Blocks(css=custom_css) as demo:
    with gr.Column(elem_id="container"):
        gr.Markdown("## Pure RAG Output\nDisplays the retrieved passages from the corpus.")
        query_input = gr.Textbox(label="Query", placeholder="Enter your query here...", lines=1)
        with gr.Column():
            submit_button = gr.Button("Submit")
            clear_button = gr.Button("Clear")
        output_box = gr.HTML(label="Retrieved Passages", elem_classes="output-box")
        
        submit_button.click(fn=get_pure_rag_output, inputs=query_input, outputs=output_box)
        clear_button.click(fn=clear_output, inputs=[], outputs=output_box)

demo.launch()