✦ Stella — LLM-Powered FAQ Assistant

An AI-powered FAQ assistant that answers questions from your own documents using Retrieval-Augmented Generation (RAG). Ask anything in plain English — Stella searches your knowledge base, retrieves the most relevant context, and generates a grounded answer with cited sources.

---

🧠 How It Works

User asks a question
        ↓
Stella converts it to a semantic embedding
        ↓
ChromaDB searches for the most relevant document chunks
        ↓
Retrieved context is passed to the LLM (via Ollama)
        ↓
LLM generates a grounded answer with sources
        ↓
Answer displayed in Streamlit UI with typewriter effect

---

Project Workflow

---

⚙️ Tech Stack

Layer	Technology
Frontend	Streamlit
Backend	FastAPI + Python
LLM	Ollama (llama3.2:1b or any model)
Embeddings	LangChain FastEmbedEmbeddings
Vector DB	ChromaDB
Orchestration	Docker Compose

CPU-only machines work but responses will be slow. Remove the deploy.resources GPU block from compose.yaml if you don't have a GPU.

---

🚀 Option A — Run From Source Code (GitHub)

1. Clone the repo

git clone https://github.com/yrarjun59/FAQ-Assistant.git
cd FAQ-Assistant

2. Configure environment (optional)

cp .env.example .env
# edit .env to change models or API URLs if needed

3. Start everything

docker compose up --build

installed ollama model or takes more time for that

This will automatically:

• Build backend and streamlit images
• Start Ollama server
• Pull llama3.2:1b models
• Start the API and UI

4. Open the app

http://localhost:8501

🔧 Configuration

Copy .env.example to .env and edit as needed:

# Model selection
LLM_MODEL=llama3.2:1b         

# API URLs (change only if running outside Docker)
OLLAMA_BASE_URL=http://ollama:11434
STELLA_API_URL=http://backend:8000

📥 Adding Your Own Documents

Place your FAQ or documentation files inside backend/knowledge/:

backend/
└── knowledge/
    ├── doc1
    ├── doc2

delete the vector_db folder and only run app and ensure docs are in json format as in mine otherwise not ingesting.....

🔄 Getting Updates

git pull
docker compose up --build