app.py

transformers
torch
sentence-transformers
gradio

import gradio as gr
from transformers import pipeline
from sentence_transformers import SentenceTransformer, util

# Load models
retriever_model = SentenceTransformer('all-MiniLM-L6-v2')  # A sentence transformer for retrieval
generator_model = pipeline('text-generation', model='gpt2')  # Use a suitable model

# Sample document collection
documents = [
    "The quick fox jumps over the lazy dog.",
    "A speedy brown fox jumps over the lazy dog.",
    "A cat in the hat is wearing a red bow.",
    "The sun sets behind the mountains, casting a warm glow."
]

# Function to perform retrieval and generation
def rag_evaluation(query):
    # Compute embeddings for the documents
    doc_embeddings = retriever_model.encode(documents, convert_to_tensor=True)

    # Compute embedding for the query
    query_embedding = retriever_model.encode(query, convert_to_tensor=True)

    # Calculate cosine similarity
    similarities = util.pytorch_cos_sim(query_embedding, doc_embeddings)[0]

    # Get the most similar document
    most_similar_idx = similarities.argmax()
    retrieved_doc = documents[most_similar_idx]

    # Generate a response based on the retrieved document
    generated_response = generator_model(f"Based on the document: {retrieved_doc}, answer: {query}", max_length=50)[0]['generated_text']

    return retrieved_doc, generated_response

# Gradio interface
iface = gr.Interface(
    fn=rag_evaluation,
    inputs=gr.Textbox(label="Enter your query"),
    outputs=[gr.Textbox(label="Retrieved Document"), gr.Textbox(label="Generated Response")],
    title="RAG Evaluation App",
    description="Evaluate retrieval and generation performance of a RAG system."
)

iface.launch()