PythonPlaza.com

Working RAG Application.

We will build a RAG application using the following:

1. Flask
2. Groq
3. ChromaDB
4. LangChain
5. PDF documents in pdf_documents

Install the packages

pip install flask python-dotenv
pip install langchain
pip install langchain-core
pip install langchain-community
pip install langchain-groq
pip install langchain-chroma
pip install langchain-huggingface
pip install langchain-text-splitters
pip install chromadb
pip install pypdf
pip install sentence-transformers

This project was developed in PyCharm IDE.

Here's the code for Ingest.py

import os import shutil from langchain_community.document_loaders import PyPDFLoader from langchain_text_splitters import RecursiveCharacterTextSplitter from langchain_huggingface import HuggingFaceEmbeddings from langchain_chroma import Chroma PDF_FOLDER = "pdf_documents" DB_FOLDER = "chroma_db" # --- ADD THIS BLOCK TO CLEAR OLD DATA --- if os.path.exists(DB_FOLDER): print("Clearing old database...") shutil.rmtree(DB_FOLDER) # ---------------------------------------- print("Loading PDFs...") documents = [] for file in os.listdir(PDF_FOLDER): if file.endswith(".pdf"): pdf_path = os.path.join(PDF_FOLDER, file) loader = PyPDFLoader(pdf_path) docs = loader.load() for doc in docs: doc.metadata["source_file"] = file documents.extend(docs) print(f"Loaded {len(documents)} pages") # Chunking splitter = RecursiveCharacterTextSplitter( chunk_size=1000, chunk_overlap=200 ) chunks = splitter.split_documents(documents) print(f"Created {len(chunks)} chunks") # Embeddings embeddings = HuggingFaceEmbeddings( model_name="sentence-transformers/all-MiniLM-L6-v2" ) # Create ChromaDB vectorstore = Chroma.from_documents( documents=chunks, embedding=embeddings, persist_directory=DB_FOLDER ) print("Chroma database created successfully.")

Here's the code for app.py

import os from flask import Flask from flask import render_template from flask import request from dotenv import load_dotenv from langchain_groq import ChatGroq from langchain_core.prompts import ChatPromptTemplate from langchain_huggingface import HuggingFaceEmbeddings from langchain_chroma import Chroma # ------------------------------------ # Load Environment Variables # ------------------------------------ load_dotenv() GROQ_API_KEY = os.getenv("GROQ_API_KEY") GROQ_MODEL = os.getenv("GROQ_MODEL") # ------------------------------------ # Flask # ------------------------------------ app = Flask(__name__) # ------------------------------------ # LLM # ------------------------------------ llm = ChatGroq( api_key=GROQ_API_KEY, model=GROQ_MODEL, temperature=0 ) # ------------------------------------ # Embeddings # ------------------------------------ embeddings = HuggingFaceEmbeddings( model_name="sentence-transformers/all-MiniLM-L6-v2" ) # ------------------------------------ # Load Chroma # ------------------------------------ vectorstore = Chroma( persist_directory="chroma_db", embedding_function=embeddings ) retriever = vectorstore.as_retriever( search_type="similarity", search_kwargs={ "k": 7 # Pulls 7 chunks instead of 5 to provide more context } ) # ------------------------------------ # Prompt # ------------------------------------ prompt = ChatPromptTemplate.from_template(""" You are a helpful AI assistant. Answer the question based on the supplied context. Use factual details from the context to form your answer. If the context completely lacks any information to answer the question, reply exactly: "I could not find the answer in the uploaded documents." Context: {context} Question: {question} """) # ------------------------------------ # Home # ------------------------------------ @app.route("/", methods=["GET", "POST"]) def home(): answer = "" sources = [] if request.method == "POST": question = request.form["question"] # Retrieve documents docs = retriever.invoke(question) # Build context context = "\n\n".join( doc.page_content for doc in docs ) # Get sources for doc in docs: filename = doc.metadata.get( "source_file", "Unknown" ) if filename not in sources: sources.append(filename) # Create chain chain = prompt | llm response = chain.invoke({ "context": context, "question": question }) answer = response.content return render_template( "index.html", answer=answer, sources=sources ) # ------------------------------------ # Run # ------------------------------------ if __name__ == "__main__": app.run(debug=True)

Her's the code for .env file

Here's the code for index.html file

<!DOCTYPE html>
<html>

<head>

    <title>PDF RAG Chatbot</title>

    <style>

        body{
            width:80%;
            margin:auto;
            margin-top:40px;
            font-family:Arial;
        }

        textarea{
            width:100%;
            height:120px;
        }

        button{
            margin-top:10px;
            padding:10px;
        }

        .answer{
            margin-top:25px;
            padding:15px;
            background:#f4f4f4;
            border:1px solid #ddd;
        }

    </style>

</head>

<body>

<h2>Groq + Chroma PDF RAG</h2>

<form method="POST">

<textarea
name="question"
placeholder="Ask a question..."
required></textarea>

<br>

<button type="submit">
Ask
</button>

</form>

{% if answer %}

<div class="answer">

<h3>Answer</h3>

<p>{{ answer }}</p>

<h4>Sources</h4>

<ul>

{% for source in sources %}

<li>{{ source }}</li>

{% endfor %}

</ul>

</div>

{% endif %}

</body>
</html>

How To get a Groq API key

1. Log in or Create an Account: Visit the Groq Cloud Console.
2. Sign in with your existing email or quickly authenticate using a Google account.
3. Navigate to Keys: Click on the API Keys tab located in the left sidebar menu.
4. Generate the Key: Click the Create API Key button.

Running the application in PyCharm

1. Run the Ingest.py
2. Run the application.

Screenshots are attached below..