app.py

# AI assistant with a RAG system to query information from
#  the gwIAS search pipline
# using Langchain and deployed with Gradio

# Thanks to Pablo Villanueva Domingo for sharing his CAMELS template
# https://huggingface.co/spaces/PabloVD/CAMELSDocBot

from rag import RAG, load_docs
from langchain_community.embeddings import HuggingFaceInstructEmbeddings
from langchain.chat_models import ChatOpenAI
import gradio as gr

# Load the documentation
docs = load_docs()
print("Pages loaded:", len(docs))

# LLM model
llm = ChatOpenAI(model="gpt-4o-mini")

# Embeddings
embed_model = "sentence-transformers/multi-qa-distilbert-cos-v1"
# embed_model = "nvidia/NV-Embed-v2"
embeddings = HuggingFaceInstructEmbeddings(model_name=embed_model)

# RAG chain
rag_chain = RAG(llm, docs, embeddings)

# Function to handle prompt and query the RAG chain
def handle_prompt(message, history):
    try:
        # Stream output
        out = ""
        for chunk in rag_chain.stream(message):
            out += chunk
            yield out
    except Exception as e:
        raise gr.Error(f"An error occurred: {str(e)}")


if __name__ == "__main__":

    # Predefined messages and examples
    description = "AI powered assistant to help with [gwIAS](https://github.com/JayWadekar/gwIAS-HM) gravitational wave search pipeline."
    greetingsmessage = "Hi, I'm the gwIAS Bot, I'm here to assist you with the search pipeline."
    example_questions = [
        "Can you give me the code for calculating coherent score?",
        "Which module in the code is used for collecting coincident triggers?",
        "How are template banks constructed?"
    ]

    # Define customized Gradio chatbot
    chatbot = gr.Chatbot([{"role": "assistant", "content": greetingsmessage}],
                         type="messages",
                         avatar_images=["ims/userpic.png", "ims/gwIASlogo.jpg"],
                         height="60vh")

    # Define Gradio interface
    demo = gr.ChatInterface(handle_prompt,
                            type="messages",
                            title="gwIAS DocBot",
                            fill_height=True,
                            examples=example_questions,
                            theme=gr.themes.Soft(),
                            description=description,
                            # cache_examples=False,
                            chatbot=chatbot)

    demo.launch()

# https://arxiv.org/html/2405.17400v2
# https://arxiv.org/html/2312.06631v1
# https://arxiv.org/html/2310.15233v2