[ONE BIG DOCUMENT - 50 pages]
Return policy... shipping info... warranty... products...
contact info... payment methods... etc etc...

Chunk 1: Return Policy (just this topic)
Chunk 2: Shipping Info (just this topic)
Chunk 3: Warranty Info (just this topic)
Chunk 4: Product Details (just this topic)

JarMart Return Policy: All electronics can be returned
within 30 days of purchase. Items must be in original
packaging with all accessories.

Refunds are processed within 5-7 business days.

Shipping Information: Standard shipping takes 3-5 business
days within Japan. Express shipping takes 1-2 business days.

Free shipping on orders over 5,000 yen.

✂️ Chunk 1:
"JarMart Return Policy: All electronics can be returned
within 30 days of purchase. Items must be in original
packaging with all accessories."

✂️ Chunk 2:
"Refunds are processed within 5-7 business days."

✂️ Chunk 3:
"Shipping Information: Standard shipping takes 3-5 business
days within Japan. Express shipping takes 1-2 business days."

✂️ Chunk 4:
"Free shipping on orders over 5,000 yen."

Most Similar Chunks:
✓ Chunk 1: "All electronics can be returned within 30 days..." (95% match)
✓ Chunk 2: "Refunds are processed within 5-7 business days." (70% match)
✗ Chunk 3: "Shipping Information..." (20% match - not relevant)
✗ Chunk 4: "Free shipping on orders..." (15% match - not relevant)

You are a customer support assistant for JarMart.

Here is relevant information from our documentation:
---
JarMart Return Policy: All electronics can be returned
within 30 days of purchase. Items must be in original
packaging with all accessories.

Refunds are processed within 5-7 business days.
---

Customer Question: How many days do I have to return a product?

Answer the question based ONLY on the information provided above.
If you don't find the answer in the information, say "I don't know."

"You have 30 days from the date of purchase to return electronics.
Please ensure the item is in its original packaging with all
accessories included."

openai==1.12.0
chromadb==0.4.22
sentence-transformers==2.5.1

pip install -r requirements.txt

JarMart Return Policy:
All electronics can be returned within 30 days of purchase if you're not satisfied.
Items must be in original packaging with all accessories included.
The product must be in unused or like-new condition.
Refunds are processed within 5-7 business days after we receive the returned item.
You can initiate a return by contacting our support team or through your account dashboard.

JarMart Shipping Information:
We offer two shipping options for customers in Japan.
Standard shipping takes 3-5 business days and costs 500 yen.
Express shipping takes 1-2 business days and costs 1,200 yen.
Free standard shipping is available on all orders over 5,000 yen.
We currently only ship within Japan.

JarMart Warranty Information:
All electronics come with a 1-year manufacturer warranty.
Extended warranty plans are available for purchase at checkout.
Warranty covers manufacturing defects but not accidental damage.
To claim warranty, keep your purchase receipt and contact our support team.

Product: Fajar XX-123XYZ Headphones:
Price: 45,000 yen
These premium noise-cancelling headphones feature industry-leading noise cancellation.
Battery life: Up to 30 hours with ANC on, 40 hours with ANC off.
Includes: Carrying case, USB-C charging cable, audio cable, and airplane adapter.
Available in two colors: Black and Silver.

Product: Fajar Phone A12:
Price: 159,800 yen
The latest Fajar Phone with Premium chip and titanium design.
Storage options: 128GB, 256GB, 512GB, 1TB.
Camera: 48MP main camera with 5x optical zoom.
Includes: USB-C charging cable and documentation (no power adapter).
Available in four colors: Natural Titanium, Blue Titanium, White Titanium, Black Titanium.

import chromadb
from sentence_transformers import SentenceTransformer
from openai import OpenAI

# ==========================================
# STEP 1: LOAD OUR STORE DOCUMENTATION
# ==========================================
# Read the store policies from the file
with open("store_policies.txt", "r", encoding="utf-8") as f:
    full_document = f.read()

# Print so you can see what we loaded
print("📄 Loaded store documentation:")
print(full_document[:200] + "...\n")

# ==========================================
# STEP 2: BREAK INTO CHUNKS
# ==========================================
# Why? Because we want to find specific, relevant pieces of information
# We'll split by double newlines (paragraph breaks)
chunks = full_document.split("\n\n")

# Remove empty chunks and clean up
chunks = [chunk.strip() for chunk in chunks if chunk.strip()]

print(f"✂️  Split document into {len(chunks)} chunks")
print(f"Example chunk: '{chunks[0][:100]}...'\n")

# ==========================================
# STEP 3: CONVERT CHUNKS TO EMBEDDINGS
# ==========================================
# This is where we convert text to numbers!
# We use a free model that runs on your computer
print("🔢 Converting text to embeddings (numbers)...")

# Load the embedding model
# "all-MiniLM-L6-v2" is a good, fast, free model
embedding_model = SentenceTransformer('sentence-transformers/all-MiniLM-L6-v2')

# Convert each chunk to embeddings
chunk_embeddings = embedding_model.encode(chunks)

print(f"✅ Created embeddings with {len(chunk_embeddings[0])} dimensions")
print(f"Example: '{chunks[0][:50]}...' → [{chunk_embeddings[0][:3]}...]")
print()

# ==========================================
# STEP 4: STORE IN VECTOR DATABASE
# ==========================================
# This is our smart filing cabinet!
print("💾 Storing embeddings in vector database...")

# Create a ChromaDB client that persists to disk
# This will create a folder called "chroma_db" in your project directory
client = chromadb.PersistentClient(path="./chroma_db")

# Try to get existing collection, or create new one
try:
    collection = client.get_collection(name="store_docs")
    print("✅ Loaded existing database from disk\n")
except:
    collection = client.create_collection(name="store_docs")
    
    # Store each chunk with its embedding (only if creating new)
    for i, (chunk, embedding) in enumerate(zip(chunks, chunk_embeddings)):
        collection.add(
            ids=[f"chunk_{i}"],           # Unique ID for each chunk
            embeddings=[embedding.tolist()],  # The numbers
            documents=[chunk],             # The actual text
        )
    print(f"✅ Stored {len(chunks)} chunks in database\n")

# ==========================================
# STEP 5: SETUP OPENAI FOR ANSWERING
# ==========================================
# You need an OpenAI API key for this part
# Get one at: https://platform.openai.com/api-keys
openai_client = OpenAI(api_key="YOUR_API_KEY")

# ==========================================
# STEP 6: THE QUERY FUNCTION - THIS IS RAG!
# ==========================================
def ask_question(question):
    """
    This function takes a question and returns an answer using RAG
    """
    print(f"\n{'='*60}")
    print(f"❓ Question: {question}")
    print(f"{'='*60}\n")

    # STEP 6.1: Convert question to embeddings
    print("1️⃣ Converting question to embeddings...")
    question_embedding = embedding_model.encode([question])[0]
    print(f"   → Question converted to {len(question_embedding)} numbers\n")

    # STEP 6.2: Find similar chunks (THE RETRIEVAL PART!)
    print("2️⃣ Searching for relevant information...")
    results = collection.query(
        query_embeddings=[question_embedding.tolist()],
        n_results=3  # Get top 3 most similar chunks
    )

    retrieved_chunks = results['documents'][0]

    print(f"   → Found {len(retrieved_chunks)} relevant chunks:")
    for i, chunk in enumerate(retrieved_chunks):
        print(f"   {i+1}. {chunk[:80]}...")
    print()

    # STEP 6.3: Create context from retrieved chunks
    context = "\n\n".join(retrieved_chunks)

    # STEP 6.4: Build the prompt for GenAI Tools (THE AUGMENTED PART!)
    print("3️⃣ Creating prompt with retrieved information...\n")

    prompt = f"""You are a helpful customer support assistant for JarMart, an electronics store.

Here is relevant information from our store documentation:
---
{context}
---

Customer Question: {question}

Instructions:
- Answer based STRICTLY and ONLY on the information provided above
- Do NOT use any external knowledge or make assumptions
- If the specific information is not in the context above, you MUST say "I don't have that information in our documentation"
- Do NOT suggest that we might have products unless they are explicitly mentioned in the context
- Be helpful and friendly, but stay within the bounds of the provided information

Answer:"""

    # STEP 6.5: Get answer from GenAI Tools (THE GENERATION PART!)
    print("4️⃣ Asking GenAI Tools to generate an answer...\n")

    response = openai_client.chat.completions.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "user", "content": prompt}
        ],
        temperature=0  # 0 = more focused, less creative
    )

    answer = response.choices[0].message.content

    # Display the answer
    print(f"{'='*60}")
    print(f"💬 ANSWER:")
    print(f"{'='*60}")
    print(answer)
    print(f"{'='*60}\n")

    return answer

# ==========================================
# STEP 7: TEST IT!
# ==========================================
if __name__ == "__main__":
    print("\n🤖 JarMart Customer Support Bot Ready!\n")

    # Test with different questions
    test_questions = [
        "How many days do I have to return a product?",
        "What comes with the Fajar headphones?",
        "How much does express shipping cost?",
        "Do you sell laptops?"  # This should say "I don't have that info"
    ]

    for question in test_questions:
        ask_question(question)
        input("Press Enter to continue to next question...")

python support_bot_simple.py

📄 Loaded store documentation:
JarMart Return Policy:
All electronics can be returned within 30 days...

✂️  Split document into 9 chunks
Example chunk: 'JarMart Return Policy:...'

🔢 Converting text to embeddings (numbers)...
✅ Created embeddings with 384 dimensions
Example: 'JarMart Return Policy:...' → [0.043, -0.028, 0.015...]

💾 Storing embeddings in vector database...
✅ Stored 9 chunks in database

🤖 JarMart Customer Support Bot Ready!

============================================================
❓ Question: How many days do I have to return a product?
============================================================

1️⃣ Converting question to embeddings...
   → Question converted to 384 numbers

2️⃣ Searching for relevant information...
   → Found 3 relevant chunks:
   1. JarMart Return Policy: All electronics can be returned within 30 days...
   2. Refunds are processed within 5-7 business days after we receive...
   3. You can initiate a return by contacting our support team...

3️⃣ Creating prompt with retrieved information...

4️⃣ Asking GenAI Tools to generate an answer...

============================================================
💬 ANSWER:
============================================================
You have 30 days from the date of purchase to return electronics.
Items must be in original packaging with all accessories and in
unused or like-new condition. You can start a return by contacting
our support team or through your account dashboard.
============================================================

Q: "What comes with the Fajar XX-123XYZ headphones from JarMart?"
A: "I don't have specific information about JarMart's offerings..."

Q: "What comes with the Fajar XX-123XYZ headphones?"
A: "The Fajar XX-123XYZ headphones include a carrying case,
    USB-C charging cable, audio cable, and airplane adapter."

"All electronics can be returned within 30 days.
Items must be in original packaging.
Contact our support team to start the return process."

Chunk 1: "All electronics can be returned within 30 days."
Chunk 2: "Items must be in original packaging."
Chunk 3: "Contact our support team to start the return process."

[Photo 1: LEFT side]  [Photo 2: MIDDLE]  [Photo 3: RIGHT side]
                  ❌ Gap            ❌ Gap

[Photo 1: LEFT to MIDDLE]
            [Photo 2: MIDDLE to RIGHT]
                 ✓ Overlap!

Original Text:
"All electronics can be returned within 30 days.
Items must be in original packaging.
Contact our support team to start the return process."

✂️ Chunk 1 (with extra text at the end):
"All electronics can be returned within 30 days.
Items must be in original packaging."
                    ↓ OVERLAP ↓
✂️ Chunk 2 (starts earlier, includes some repeated text):
"Items must be in original packaging.
Contact our support team to start the return process."

No Overlap (0%):
[Chunk 1]  [Chunk 2]  [Chunk 3]
Risk: Missing context between chunks ❌

Small Overlap (10-20%):
[Chunk 1____]
        [Chunk 2____]
               [Chunk 3____]
Good balance! ✓

Too Much Overlap (50%+):
[Chunk 1_________]
     [Chunk 2_________]
          [Chunk 3_________]
Wastes space, costs more ❌

[0---------250--------300]  ← Chunk 1
                [250--------550]  ← Chunk 2 (overlaps from 250-300)
                          [500--------800]  ← Chunk 3 (overlaps from 500-550)

```python
def smart_chunk_text(text, chunk_size=300, overlap=50):
    """
    Break text into overlapping chunks

    Args:
        text: Your full document as a string
        chunk_size: How many characters per chunk (default: 300)
        overlap: How many characters to overlap (default: 50)

    Returns:
        A list of text chunks
    """
    chunks = []
    start = 0

    # Keep creating chunks until we've covered the whole text
    while start < len(text):
        # Get the end position for this chunk
        end = start + chunk_size

        # Extract the chunk
        chunk = text[start:end]

        # Add this chunk to our list
        chunks.append(chunk)

        # Move to the next position (with overlap!)
        # If chunk_size=300 and overlap=50, we move forward only 250 characters
        start = start + chunk_size - overlap

    return chunks

# Example usage:
document = """JarMart Return Policy: All electronics can be returned
within 30 days of purchase. Items must be in original packaging.
Contact our support team to start the return process."""

# Create chunks with 100 character size and 20 character overlap
chunks = smart_chunk_text(document, chunk_size=100, overlap=20)

print(f"Created {len(chunks)} chunks:")
for i, chunk in enumerate(chunks):
    print(f"\nChunk {i+1}:")
    print(chunk)

# Next step: Save these chunks to your vector database
# (Use the same method shown in the main code earlier)

Created 3 chunks:

Chunk 1:
JarMart Return Policy: All electronics can be returned
within 30 days of purchase. Items must

Chunk 2:
 Items must be in original packaging.
Contact our support team to start the

Chunk 3:
start the return process.

# Use this instead
embedding_model = SentenceTransformer('sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2')

ask_question("Do you sell cars?")
# Answer: "I don't have information about selling cars in our documentation."

# Instead of OpenAI, use a local model
from transformers import pipeline

generator = pipeline('text-generation', model='mistral-7b')

```python
def ask_question_with_monitoring(question):
    # ... retrieve chunks ...

    # Log what was retrieved
    print(f"📊 Retrieval Quality Report:")
    for i, chunk in enumerate(retrieved_chunks):
        print(f"  Chunk {i+1} (first 100 chars): {chunk[:100]}...")

    # This helps you debug if answers are wrong!

# Add new product
new_product = """
Product: Fajar Phone A16:
Price: 140,000 yen
Features: AI-powered camera
"""

# Re-run the embedding and storage steps
# No need to retrain any models!

collection.add(
    ids=[f"chunk_{i}"],
    embeddings=[embedding.tolist()],
    documents=[chunk],
    metadatas=[{"category": "return_policy", "last_updated": "2025-11-12"}]
)

# Later, search only in specific categories
results = collection.query(
    query_embeddings=[question_embedding],
    where={"category": "return_policy"}
)