🤖 Conversational Q&A RAG Chatbot with Memory using LangChain, HuggingFace and Groq AI

A modular, memory-enhanced Retrieval-Augmented Generation (RAG) chatbot built using LangChain, Groq’s blazing-fast LLaMA3 API, Chroma vector store, and HuggingFace sentence embeddings.

This project allows multi-turn conversations with context-aware question reformulation and history tracking using session-based memory — making it ideal for educational assistants, AI agents, and document-based Q&A.

---

📋 Key LangChain Features Used

Feature	Description
ChatPromptTemplate	Modular prompt templates for structured messaging
StrOutputParser	Clean extraction of string responses from raw LLM outputs
RunnableWithMessageHistory	Maintains chat state across multiple LLM calls
Chroma.from_documents	Create vector store from document chunks with embeddings
as_retriever()	Convert vector store into retriever for use in RAG
LCEL (operator)	Expressive pipeline chaining of LangChain component
create_history_aware_retriever	Convert an updated retriever which remembers chat history

📚 What You’ll Learn

• How to build a RAG chatbot with:
  • Custom vector store retrieval
  • Prompt engineering for Q&A
  • Chain and agent-based memory integration
• How to maintain conversational context using RunnableWithMessageHistory
• How to scrape and chunk real web content for vector-based retrieval

---

🧠 Architecture Overview

🧠 App Preview

🧩 Tech Stack

• 🦜 LangChain: Core framework
• 🧠 Groq LLaMA3: High-performance LLM
• 🔍 ChromaDB: In-memory vector store
• 🧬 HuggingFace MiniLM Embeddings
• 🌐 WebBaseLoader: Web scraping from blog/article sources
• 🧱 RecursiveCharacterTextSplitter: Smart document chunking
• 💬 Session-based Chat History: Rehydration of past queries

---

📦 Installation

git clone https://github.com/your-username/Conversational-QnA-Chatbot.git
cd Conversational-QnA-Chatbot
pip install -r requirements.txt

---

🔐 Environment Variables

Create a .env file in the root directory with the following:

GROQ_API_KEY=your_groq_key
HF_TOKEN=your_huggingface_token

---

🧪 Usage

# Step 1: Run once to initialize & ingest docs
python rag_chatbot.py

# Step 2: Start chatting with conversational memory!
conversational_rag_chain.invoke(
    {"input": "What is Chain of Thought?"},
    config={"configurable": {"session_id": "your-session-id"}}
)

---

📘 Example Use Case

# First question
conversational_rag_chain.invoke(
    {"input": "What is Chain of Thought?"},
    config={"configurable": {"session_id": "abc456"}}
)

# Follow-up question with context memory
conversational_rag_chain.invoke(
    {"input": "What are common ways of doing it?"},
    config={"configurable": {"session_id": "abc456"}}
)

---

🛠️ Extending This

Want to plug in your own documents or data sources?

Just change the source URL here:

loader = WebBaseLoader(
    web_paths=("https://your-site.com/article",),
    ...
)

Or replace with PDF/Text loader for local files.

---

🧠 Memory Types You Can Explore

• RunnableWithMessageHistory (current)
• ConversationBufferMemory
• ZepMemory, ChromaMemory (advanced)

---

📄 Resources Used

• Lilian Weng’s Agent Blog
• LangChain Docs
• Groq LLM API

---

💬 License

MIT License. Feel free to fork, extend, or deploy in your projects.

---