Coding up the RAG capability of the site – Looking Ahead into the Future of IoT-1
By Patrice Duren / May 30, 2024 / No Comments / AWS Certification Exam, Managing threats and risks
The first component of the coding starts with setting up the HTML landing page for us to be able to submit queries:
We first need to create the HTML code that will be used for the landing page. Create a Templates folder in the working directory.
Change directory into Templates and create a file named query_form.html.
Insert the following code into the file:
<!DOCTYPE html>
<html lang=”en”>
<head>
<meta charset=”UTF-8″>
<title>Query Form</title>
</head>
<body>
<form action=”/query” method=”post”>
<input type=”text” name=”query” placeholder=”Enter your query here” required>
<input type=”submit” value=”Submit”>
</form>
</body>
</html>
The code defines a simple web page with a form intended for user interaction. The language is set to English. The form comprises a single text input field where users can type a query. The text field is labeled Enter your query here, which serves as a placeholder inside the box, and it is a required field, meaning the form cannot be submitted without filling it out. There is also a submit button labeled Submit. When the form is submitted, it sends a POST request to the /query URL, which is typically handled by a server-side script to process the query. The form’s functionality is often used for search operations or data retrieval requests on a website.
With that, we have been able to create the landing page for the site. We can now move on to creating the RAG component of the site.
We now want to create a RAG capability. As part of this, we will look at defining the appropriate routes for the Flask app while powering the site with the RAG capability:
Create a new file called app.py. This will be where we put our main code.
Import the dependencies needed as follows:
import boto3
import json
import os
import sys
import numpy as np
from urllib.request import urlretrieve
import ssl
from langchain.text_splitter import CharacterTextSplitter, RecursiveCharacterTextSplitter
from langchain.document_loaders import PyPDFLoader, PyPDFDirectoryLoader
from langchain.embeddings import BedrockEmbeddings
from langchain.llms.bedrock import Bedrock
from langchain.chains.question_answering import load_qa_chain
from langchain.vectorstores import FAISS
from langchain.indexes import VectorstoreIndexCreator
from langchain.indexes.vectorstore import VectorStoreIndexWrapper
The following are descriptions of what each dependency does:
- boto3: This is the AWS SDK for Python. It allows Python developers to write software that makes use of services such as Amazon Simple Storage Service (Amazon S3) and Amazon Elastic Compute Cloud (Amazon EC2). In the context of your application, it’s used to interact with the AWS services, possibly for storing data or utilizing AWS AI/ML services.
- os: The os module, part of the standard library, offers a user-friendly method to perform operating system-specific tasks such as file reading/writing, path management, and environment variable access across different systems.
- sys: The sys module, also included in the standard library, allows users to access certain variables that the Python interpreter uses or maintains, as well as functions that have a significant interaction with the interpreter. This module is commonly utilized for adjusting the Python runtime environment.
- numpy: numpy, which stands for Numerical Python, is a Python library designed to enhance the language by adding support for extensive, multi-dimensional arrays and matrices. Additionally, it provides a wide range of advanced mathematical functions to work with these arrays.
- urllib.request: This module is used to open and read URLs. urlretrieve is a function within this module that is used to download a file from a remote URL to a local file path.
- ssl: The ssl module offers tools for implementing Transport Layer Security (TLS) encryption and verifying the identities of peers in network connections, applicable to both client and server sides. It’s useful for safeguarding communication between a client and a server.
- langchain.text_splitter: CharacterTextSplitter and RecursiveCharacterTextSplitter are classes within the langchain package used for splitting text into manageable pieces for further processing or analysis.
- langchain.document_loaders: This module contains utilities for loading and reading documents. The PyPDFLoader and PyPDFDirectoryLoader are classes used to load PDF files or directories containing PDF files into Python objects for manipulation or data extraction.
- langchain.embeddings: This module deals with generating embeddings for text. BedrockEmbeddings is a class that interacts with the Bedrock API to create numerical representations of text that capture semantic meaning.
- langchain.llms.bedrock: This module is a part of the langchain library that likely interfaces with the Bedrock language learning models. The Bedrock class is used to interact with a specific language model provided by Bedrock.
- langchain.chains.question_answering: This is a module within langchain dedicated to question-answering (QA) functionality. load_qa_chain is a function or method for loading a QA process chain, possibly setting up a series of steps for processing and answering questions.
- langchain.vectorstores: This module handles the storage and retrieval of vectors, which are the result of converting text into numerical embeddings. Facebook AI Similarity Search (FAISS) is a class or interface for working with the FAISS library, which is a library for efficient similarity search and clustering of dense vectors.
- langchain.indexes: These classes (VectorstoreIndexCreator and VectorStoreIndexWrapper) are for the indexing of vectors for efficient retrieval during operations such as similarity search or document retrieval in the context of the RAG system. Indexing is essential for scaling up to large datasets where quick retrieval is necessary.