Step 1: Set up RAG in Seconds Using Gradient’s Accelerator Block for RAG

Gradient’s made it extremely easy to setup RAG. Simply drag and drop your documents into Gradient’s UI and Gradient will generate a “RAG Collection ID” for you, that can be used as a reference. With Gradient’s Accelerator Block for RAG, you’ll be receiving:

• The Fastest Way to Production RAG: Gradient is improving your development process by 10x, by removing the need for infrastructure, setup or in-depth knowledge around AI.

• Best-of-Breed Technologies: When you use Gradient’s Accelerator Block for RAG, you won’t be compromising on quality. Our RAG service is powered by best-of-breed technologies including: Gradient for state-of-the-art LLMs and embeddings, MongoDB Atlas for indexing high-dimensional vector data, and LlamaIndex for the use of their advanced RAG framework.

• Optimized RAG Performance: We’ve already done the work for you to enhance your RAG performance and accuracy. Our Accelerator Block for RAG incorporates best practices and methods for optimization including the use of an optimal chunking strategy, rerankers, and advanced retrieval strategies.

Step 2: Connect Your RAG Collection to Your Accelerator Block for Q&A

Now that you’ve set up your RAG collection via the Accelerator Block for RAG, you can connect it to Gradient’s Accelerator Block for Q&A. Our Q&A block, accurately responds to questions on documents stored in RAG or provided via text input. If you want to check out whether or not everything is up to par before connecting it to your final application, you can actually easily test this out using Gradient’s playground as shown below. You’ll also be able to see the RAG context that is retrieved from the query, so you can quickly debug the query live.

To do this via the Python SDK, reference the following example. You can find your RAG collection ID using the instructions here.

from gradientai import Gradient
gradient = Gradient()
question = "What are Meta's areas of investments?"
response = gradient.answer(
    question=question,
    source={
        "type": "rag",
        "collection_id": "YOUR_RAG_COLLECTION_ID",
    },
)

Step 3: Chat Over the Documents Using Streamlit

You can easily use the Gradient SDK in your Streamlit app and couple it with Streamlit’s built-in chat components to create your financial chatbot. After deploying the Streamlit app, you can start using the chat interface to ask questions about the financial reports from Meta, Alphabet, and Netflix.

Here is an example of using the Gradient Q&A Accelerator Block in your Streamlit code. Note that more Streamlit integration is necessary for the full chatbot interface to function.

import streamlit as st
from gradientai import Gradient
gradient = Gradient()
# Additional Streamlit logic necessary to handle full chat interface.
with st.chat_message("assistant"):
    with st.spinner("Thinking..."):
        # Using Gradient Accelerator Block
        response = gradient.answer(
            question=prompt,
            source={
                "type": "rag",
                "collection_id": "YOUR_RAG_COLLECTION_ID",
            },
        )
        
        # Write response to Streamlit UI and chat message history
        st.write(response)
        message = {"role": "assistant", "content": response}
        st.session_state.messages.append(message)