An intelligent customer support chatbot for a fictional e-commerce store.
- RAG Retrieval — Answers from a 10-document FAQ knowledge base using ChromaDB
- Multi-turn Memory — Remembers your name and conversation context across turns
- Price Calculator — Calculates discounts and totals on request
- Smart Routing — Automatically routes questions to the right pipeline
- Self-Evaluation — Scores answer faithfulness and retries if quality is low
- Grounded Answers — Never fabricates; admits when it doesn't know
memory_node → router_node → [retrieval_node / tool_node / skip_node] → answer_node → eval_node → save_node → END
| Component | Technology |
|---|---|
| LLM | Groq (LLaMA3-8b-8192) |
| Agent Framework | LangGraph StateGraph |
| Vector Database | ChromaDB (in-memory) |
| Embeddings | SentenceTransformer all-MiniLM-L6-v2 |
| Memory | LangGraph MemorySaver |
| Evaluation | RAGAS (faithfulness metric) |
| UI | Streamlit |
ShopEasy FAQ Bot — Agentic AI capstone project built with LangGraph, ChromaDB, Groq (LLaMA3), and Streamlit. Features RAG-based FAQ retrieval, multi-turn memory, discount calculator, and RAGAS evaluation.