Docs / RAG & Search

RAG & Search Configuration

Configure Retrieval Augmented Generation (RAG) to enable your chatbot to search and provide accurate information about your WooCommerce products, orders, and store content.

What is RAG?

Retrieval Augmented Generation (RAG) is an AI technique that enhances your chatbot's responses by retrieving relevant information from your product database before generating answers. Instead of relying solely on the AI model's general knowledge, RAG allows the chatbot to access your specific product details, pricing, availability, and descriptions.

Customer Query
"Do you have blue shoes?"
→
Vector Search
Find relevant products
→
AI + Context
Generate informed response
→
Response
With real product data

Benefits of RAG

✓
Accurate Product Info Responses include real prices, stock status, and descriptions from your store
✓
Semantic Understanding Finds products by meaning, not just keywords ("running shoes" finds "athletic footwear")
✓
Always Up-to-Date Automatic sync keeps product data current with your WooCommerce catalog
✓
Reduces Hallucination AI doesn't make up product details - it uses your real data

Setting Up Supabase

WooAI Chatbot Pro uses Supabase as the vector database for RAG functionality. Supabase provides a powerful PostgreSQL database with pgvector extension for semantic search.

💡 Why Supabase?

Supabase offers a generous free tier (500MB database, 50K active users) that's perfect for most WooCommerce stores. It's also open-source, reliable, and easy to set up.

Creating a Supabase Account

  1. Visit Supabase

    Go to supabase.com and click Start your project.

  2. Sign Up

    Create an account using your GitHub account, Google account, or email address.

  3. Verify Email

    If you signed up with email, check your inbox and click the verification link.

Supabase Sign Up Page
Supabase sign-up page - create a free account

Creating a New Project

  1. Click "New Project"

    From your Supabase dashboard, click the New project button.

  2. Enter Project Details

    Fill in the project information:

    • Name: A descriptive name (e.g., "WooAI Store Products")
    • Database Password: Create a strong password (save this securely)
    • Region: Select the region closest to your WordPress hosting
  3. Create Project

    Click Create new project and wait for provisioning (takes 1-2 minutes).

Creating a New Supabase Project
Creating a new Supabase project with appropriate settings
⚠️ Save Your Password

The database password cannot be recovered after creation. Store it securely in a password manager. You'll need it if you ever need direct database access.

Getting API Credentials

After your project is created, you need to retrieve two pieces of information: the Project URL and API Key.

  1. Open Project Settings

    In your Supabase project dashboard, click the gear icon (⚙️) in the sidebar to access Settings, then select API from the submenu.

  2. Copy Project URL

    Find the Project URL field and copy the entire URL. It looks like: https://xxxxxxxx.supabase.co

  3. Copy API Key

    Under Project API keys, copy the anon / public key. This is the key you'll use in WooAI settings.

Supabase API Settings
Supabase API settings showing Project URL and API keys
Example Configuration
Project URL: https://xyzcompany.supabase.co
API Key (anon): eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9...
💡 Which Key to Use?

Always use the anon/public key for WooAI Chatbot. The service_role key has elevated privileges and should never be exposed in client-side code.

Configuring RAG in WooAI

Now that you have your Supabase credentials, configure them in your WordPress admin.

  1. Navigate to RAG Settings

    In WordPress admin, go to WooAI Chat → RAG & Search.

  2. Enable RAG

    Toggle the Enable RAG Search switch to ON.

  3. Enter Supabase URL

    Paste your Supabase Project URL into the Supabase URL field.

  4. Enter API Key

    Paste your Supabase anon key into the Supabase API Key field.

  5. Save Settings

    Click Save Settings to store your configuration.

RAG Settings in WooAI
RAG & Search settings panel with Supabase configuration

Product Indexing

After configuring Supabase, you need to index your WooCommerce products to make them searchable by the AI.

Initial Indexing

  1. Click "Index Products"

    In the RAG & Search settings, click the Index All Products button.

  2. Wait for Processing

    The plugin will process all your products in batches. A progress indicator shows the current status.

  3. Verify Completion

    Once complete, you'll see a success message with the number of products indexed.

💡 Indexing Time

Initial indexing typically takes 1-2 seconds per product. A store with 500 products will take approximately 10-15 minutes to fully index.

What Gets Indexed

The following product information is indexed for search:

Field Description Search Weight
Product Name The title of the product High
Description Full product description High
Short Description Product summary Medium
SKU Stock keeping unit Medium
Categories Product categories Medium
Tags Product tags Low
Attributes Color, size, material, etc. Medium
Price Current and sale price Metadata
Stock Status In stock, out of stock Metadata

Automatic Sync

WooAI Chatbot automatically keeps your index up-to-date:

  • New Products: Automatically indexed when published
  • Updated Products: Re-indexed when saved
  • Deleted Products: Removed from index on deletion
  • Price Changes: Updated in real-time
  • Stock Changes: Synced on status update
✅ Sync Verification

To verify sync is working, make a small change to a product (like adding a word to the description), save it, then ask the chatbot about that product. It should reflect the change immediately.

Search Settings

Fine-tune how RAG search works with these configuration options:

Search Parameters

Setting Description Default
Max Results Maximum number of products to retrieve per query 5
Similarity Threshold Minimum relevance score (0-1) for results 0.7
Include Out of Stock Whether to show out-of-stock products in results No
Search Categories Limit search to specific product categories All

Similarity Threshold Explained

The similarity threshold determines how closely a product must match the query to be included in results:

  • 0.9 - 1.0: Very strict - only near-exact matches
  • 0.7 - 0.9: Balanced - good relevance (recommended)
  • 0.5 - 0.7: Lenient - includes loosely related products
  • Below 0.5: Not recommended - may return irrelevant results
💡 Tuning Tips

Start with the default threshold of 0.7. If customers aren't finding products they're looking for, try lowering to 0.6. If results include too many irrelevant products, increase to 0.8.

Testing RAG Search

After setup, test that RAG is working correctly:

Using the Test Panel

  1. Open Test Interface

    In RAG & Search settings, find the Test Search section.

  2. Enter a Test Query

    Type a product-related query like "blue t-shirt size medium" or "laptop under $500".

  3. Review Results

    The test panel shows matched products with their similarity scores.

Frontend Testing

Test the actual customer experience:

  1. Visit your store's frontend
  2. Open the chatbot widget
  3. Ask product-related questions:
    • "What running shoes do you have?"
    • "Show me products under $50"
    • "Do you have anything in red?"
  4. Verify responses include accurate product information

Advanced Configuration

Custom Embedding Model

By default, WooAI uses Google's text-embedding model. You can configure alternative embedding models:

Model Dimensions Best For
text-embedding-004 768 General purpose (default)
text-embedding-3-small 1536 Higher accuracy, OpenAI
text-embedding-3-large 3072 Maximum accuracy, higher cost

Database Optimization

For stores with over 10,000 products, consider these optimizations:

  • Index partitioning: Split products by category for faster queries
  • Batch processing: Increase batch size for initial indexing
  • Regional selection: Choose Supabase region closest to your server
⚠️ Large Catalogs

If you have more than 50,000 products, contact support for recommendations on database scaling and optimization strategies.

Troubleshooting

Connection Failed

If you see "Failed to connect to Supabase":

  • Verify the Project URL is correct and includes https://
  • Check that the API key is the anon key, not service_role
  • Ensure your Supabase project is active (not paused)
  • Check if your server can make outbound HTTPS connections

Products Not Found

If searches return no results:

  • Verify products have been indexed (check index count in settings)
  • Lower the similarity threshold temporarily to 0.5
  • Try searching for exact product names first
  • Re-index products using the "Reindex All" button

Slow Search Performance

If searches are taking too long:

  • Reduce "Max Results" setting to 3
  • Check Supabase dashboard for database performance
  • Consider upgrading Supabase plan for more resources
  • Ensure Supabase region matches your WordPress hosting region

Index Sync Issues

If products aren't syncing automatically:

  • Check WordPress cron is running (wp cron event list)
  • Verify webhook endpoints are accessible
  • Manually trigger re-index for specific products
  • Check WordPress debug log for errors

Best Practices

Product Data Quality

  • Write detailed descriptions: More text = better semantic matching
  • Use consistent naming: Standardize terms like colors and sizes
  • Add relevant attributes: Material, dimensions, use cases
  • Update regularly: Keep descriptions current and accurate

Search Optimization

  • Monitor queries: Check Analytics to see what customers search for
  • Add synonyms: Include common alternative terms in descriptions
  • Test regularly: Periodically test searches from customer perspective
  • Review thresholds: Adjust based on result quality feedback

Performance

  • Regional proximity: Host Supabase in the same region as WordPress
  • Regular maintenance: Re-index monthly for optimal performance
  • Monitor usage: Watch Supabase quotas on free tier
  • Cache wisely: Enable caching for frequently searched products
← AI Providers Playbooks →