title	RAG Knowledge Base
emoji	💬
colorFrom	blue
colorTo	indigo
sdk	docker
pinned	false
tags	rag knowledge-base customer-support groq langchain chromadb streamlit chatbot llm python

💬 RAG Knowledge Base — Customer Support Chatbot

Pre-indexed AI knowledge base for customer support — ask questions about your product without uploading files. Drop your PDFs in /docs, index once, answer forever. ⚡

---

✨ Features

• 📁 Pre-indexed knowledge base — no file upload needed from the user
• 🔍 Semantic search over your documents via ChromaDB
• ⚡ Ultra-fast responses powered by Groq (llama-3.3-70b)
• 💬 Chat interface with conversation history and source display
• 🛠️ Admin ingestion script — run once to index all your PDFs
• 🐳 Docker ready for easy deployment

---

🏗️ Architecture

Admin (you)                    End User
     │                             │
     ▼                             ▼
Drop PDFs in /docs          Asks questions
     │                             │
     ▼                             ▼
python src/ingest.py    →   streamlit run src/app.py
     │                             │
     ▼                             ▼
ChromaDB (indexed)      ←   RAG Pipeline (Groq + LangChain)

---

📦 Tech Stack

Component	Tool	Role
LLM	Groq (llama-3.3-70b-versatile)	Answer generation
Embeddings	HuggingFace (all-MiniLM-L6-v2)	Text vectorization
Vector Store	ChromaDB	Semantic search
Orchestration	LangChain	RAG pipeline
UI	Streamlit	Chat interface

---

🚀 Getting Started (Local)

Prerequisites

• Python 3.11+
• A Groq API key (free)
• Git

1. Clone the repository

git clone https://github.com/salmazenn/rag-knowledge-base.git
cd rag-knowledge-base

2. Create a virtual environment

python3 -m venv .venv-kb
source .venv-kb/bin/activate   # Mac/Linux
# .venv-kb\Scripts\activate    # Windows

3. Install dependencies

pip install -r requirements.txt

Or install manually:

pip install langchain-community \
            langchain-groq \
            langchain-huggingface \
            langchain-chroma \
            langchain-text-splitters \
            chromadb \
            pypdf \
            streamlit \
            sentence-transformers \
            python-dotenv \
            numpy

4. Set up environment variables

Create a .env file at the root of the project:

cp .env.example .env

Then edit .env with your values:

# Required
GROQ_API_KEY=your_groq_api_key_here

# Optional — defaults shown
GROQ_MODEL=llama-3.3-70b-versatile
CHROMA_PERSIST_DIR=./data/chroma
DOCS_DIR=./docs
CHUNK_SIZE=1000
CHUNK_OVERLAP=100
TOP_K=8

5. Add your documents

Drop your PDF files into the /docs folder:

cp your-faq.pdf docs/
cp your-product-guide.pdf docs/
# Add as many PDFs as needed

6. Run the ingestion script

python src/ingest.py

You should see:

🚀 Starting ingestion...

📁 Source folder: ./docs
💾 Vector store: ./data/chroma

📄 Loading: your-faq.pdf
   ✅ 12 pages loaded

✂️  163 chunks created

🔍 Creating embeddings (all-MiniLM-L6-v2)...
💾 163 chunks saved in: ./data/chroma

🎉 Ingestion complete! 163 chunks indexed and ready.
👉 Now run: streamlit run src/app.py

7. Launch the app

streamlit run src/app.py

Open http://localhost:8501 in your browser. 🎉

---

⚙️ Configuration Parameters

Variable	Default	Description
GROQ_API_KEY	—	Required. Your Groq API key
GROQ_MODEL	llama-3.3-70b-versatile	Groq model to use
CHROMA_PERSIST_DIR	./data/chroma	ChromaDB storage path
DOCS_DIR	./docs	Folder containing your PDFs
CHUNK_SIZE	1000	Size of each text chunk (tokens)
CHUNK_OVERLAP	100	Overlap between consecutive chunks
TOP_K	8	Number of chunks retrieved per query

Tuning tips

• Increase CHUNK_SIZE (e.g. 1500) for documents with long paragraphs
• Increase TOP_K (e.g. 10-12) for broad questions requiring more context
• Decrease TOP_K (e.g. 4) for precise factual questions — faster and more accurate
• Re-run ingest.py every time you add or modify documents in /docs

---

📁 Project Structure

rag-knowledge-base/
├── src/
│   ├── ingest.py       # Admin script — index PDFs into ChromaDB
│   ├── rag.py          # RAG engine — retrieval + generation
│   └── app.py          # Streamlit chat interface
├── docs/               # Drop your PDFs here
├── data/
│   └── chroma/         # ChromaDB vector store (auto-generated)
├── .env                # Your API keys (never commit this!)
├── .env.example        # Template for environment variables
├── .gitignore
├── requirements.txt
├── Dockerfile
└── README.md

---

🔄 Updating the Knowledge Base

When you add new documents or update existing ones:

# 1. Add/replace PDFs in /docs
cp new-document.pdf docs/

# 2. Re-run ingestion (old index is automatically replaced)
python src/ingest.py

# 3. Restart the app
streamlit run src/app.py

---

☁️ Deploy on HuggingFace Spaces

1. Create a Space

• Go to huggingface.co/new-space
• SDK: Docker → Blank
• Visibility: Public

2. Add secrets

In your Space settings → Variables and secrets:

• GROQ_API_KEY → your Groq key

3. Push to HuggingFace

git remote add hf https://huggingface.co/spaces/salmazen/rag-knowledge-base
git push hf main

⚠️ Important: the ChromaDB index must be pre-built and committed to the repo before deploying. Run python src/ingest.py locally and commit the data/chroma/ folder.

---

🔭 Roadmap

• Multi-language support
• Admin dashboard to manage documents
• Automatic RAGAS evaluation
• Agents layer on top of the knowledge base
• Authentication for the chat interface

---

📄 License

MIT — free to use, modify and distribute.