RAG Customer Support Bot

Built with LangChain · ChromaDB · LangGraph · Groq (LLaMA 3.1)

---

What This Does

A Retrieval-Augmented Generation (RAG) system that:

• Reads your PDF knowledge base
• Answers user questions using only that document
• Uses LangGraph for workflow control
• Escalates to a human agent when it's not confident

---

Setup (3 Steps)

Step 1 — Get a free Groq API key

1. Go to https://console.groq.com
2. Sign up (free)
3. Create an API key
4. Paste it in your .env file:

   GROQ_API_KEY=gsk_xxxxxxxxxxxxxxxxxxxx
   

Step 2 — Install dependencies

python -m venv venv
source venv/bin/activate        # Mac/Linux
venv\Scripts\activate           # Windows

pip install -r requirements.txt

Step 3 — Add your PDF

• Place your PDF inside the data/ folder
• Rename it to your_document.pdf (or edit PDF_PATH in ingest.py)

---

Running the Bot

First time only — ingest your PDF:

python ingest.py

This creates the chroma_db/ folder with your embeddings.

Every time — start the chatbot:

python main.py

---

File Structure

rag-support-bot/
│
├── data/
│   └── your_document.pdf       ← your knowledge base
│
├── chroma_db/                  ← auto-created after ingestion
│
├── 1_ingest.py                 ← PDF → Chunks → ChromaDB
├── 2_retriever.py              ← ChromaDB → Top-K chunks
├── 3_graph.py                  ← LangGraph workflow + HITL
├── main.py                     ← Chatbot entry point
│
├── .env                        ← Your Groq API key
├── requirements.txt            ← All dependencies
└── README.md

---

How It Works

PDF
 └─[ingest.py]─→ Chunks ─→ Embeddings ─→ ChromaDB
                                                │
User Query ─→ Embed Query ─→ ChromaDB Search ──┘
                                  │
                            Top 3 Chunks
                                  │
                         [LangGraph Workflow]
                                  │
                    ┌─────────────▼─────────────┐
                    │      Groq LLM (LLaMA 3)   │
                    └─────────────┬─────────────┘
                                  │
                          [Router checks]
                         /                \
                   HIGH conf           LOW conf
                       │                   │
               Answer to User      Human Agent Input
                                           │
                                   Answer to User

---

Groq Model Options

Edit GROQ_MODEL in graph.py:

Model	Speed	Quality
llama3-8b-8192	⚡ Fastest	Good
llama3-70b-8192	Medium	Best
mixtral-8x7b-32768	Fast	Great
gemma2-9b-it	Fast	Good

---

HITL (Human-in-the-Loop)

When the bot escalates, you'll see:

⚠️  ESCALATING TO HUMAN AGENT
User asked: <question>
Bot attempted: <bot's uncertain response>
👤 Human Agent — type your response: 

Type your response and press Enter. It gets sent to the user.