Apple Product Q&A Assistant (RAG + Agent)

This Streamlit app is an intelligent assistant for answering questions about Apple products. It combines Retrieval-Augmented Generation (RAG) with rule-based agent routing for calculators, definitions, and product comparisons.

---

Architecture & Key Design Choices

• Hybrid RAG + Agent Architecture: Combines RAG-based generation with agent-based routing for different query types (e.g., calculator, definition, comparison).
• Rule-based Decision Routing: agent.py uses keyword and pattern matching to route queries to the appropriate tool or RAG engine.
• Modular Tooling: Implements distinct tools for calculation, definitions, and product comparisons, which can be extended easily.
• Local LLM (TinyLlama): Uses TinyLlama-1.1B-Chat for efficient local inference with no cloud dependency.
• FAISS-based Document Retrieval: Vector similarity search over pre-loaded Apple product documentation for relevant context.
• Transparent Query Handling: Displays decision path (RAG, calculator, etc.) and retrieved chunks used to build the answer.
• Streamlit Interface: Provides a clean and interactive UI for input/output, including visual feedback and sample use cases.

---

Features

• Ask general questions like Name all Apple products
• Ask product-specific specs like Specs of iPad Pro
• Use calculator with queries like Calculate 12 + 5
• Ask for definitions like Define RAM
• Ask for comparisons like Compare iPad Air vs iPad Pro
• Uses a local TinyLlama model for text generation
• FAISS-powered vector search from local documents

---

Sample Outputs

1. General Question

Query: Name all apple products : It gives a list of different porducts along with the different models available for the product like iphone and its models, mackbooks, apple watches etc.

---

1.Specific product

Query: Specs of all ipads : It gives a list of different models of the product with their specs.

---

3. Calculator

Query: Calculate 12+8

---

4. Dictionary Definition

Query: Define Macbooks

---

5. Product Comparison

Query: Compare iPad Pro vs iPad Air

---

How to Run Locally

1. Clone the repository

git clone https://github.com/Lalwaniamisha789/Apple-Knowledge-Assistant.git
cd Apple-Knowledge-Assistant

2. Install dependencies

pip install -r requirements.txt

3. Run the app

streamlit run app.py

Project structure

├── app.py               # Streamlit frontend
├── agent.py             # Main decision router
├── llm_chain.py         # Handles RAG prompts and LLM generation
├── retriever.py         # Loads FAISS retriever
├── tools.py             # Calculator, define, compare agents
├── doc_loader.py        # Loads and chunks documents
├── requirements.txt     # Python dependencies
├── vector_store/        # FAISS index (auto-created)
├── Docs/                # Text documents for RAG
└── images/              # Screenshots of outputs

Notes

• The app uses TinyLlama-1.1B-Chat-v1.0
• Works fully offline once vector store is created
• Ideal for building product-specific knowledge assistants
• You can see the retrieved document context and top 3 chunks used to generate each answer
• You can also see the decision route it followed whether it was rag, calculator, dictionary or comparator.

Future Improvements

• Add PDF/HTML loader support
• Extend toolset with price lookup, timeline, etc.
• Switch to a larger LLM for richer responses