Documentation

Semantic Search

Find products using natural language. Search for "wireless ceiling mic" and find results even if they're described as "RF pendant microphone."

How It's Different

Regular Search

"wireless mic" only finds products containing those exact words

Semantic Search

"wireless mic" finds all wireless microphones - even "RF handheld" or "cordless transmitter"

Example Queries

  • "ceiling mic for meetings" → Shure MXA920, Sennheiser TeamConnect
  • "4K video switcher" → HDMI matrices, AV-over-IP systems
  • "DSP for conferencing" → Biamp, QSC, Shure processors
  • "USB camera for Teams" → PTZ cameras with USB

Setup

Semantic search requires a one-time setup to "teach" the system your products:

What happens during setup?

Each product's description is converted into a numerical "embedding" that captures its meaning. This lets the system match products by concept, not just keywords. It costs 1 credit per product and only needs to be done once.

Step 1: Enable and Generate

  1. Go to Products
  2. Turn on the Semantic Search (AI) toggle
  3. You'll see how many products need embeddings (e.g., "0 / 5000")
  4. Set batch size (start with 25-50 to test, or use max for speed)
  5. Click Generate Embeddings

Step 2: Wait for Processing

The LED gauge shows progress. This runs on the server - you can close your browser and come back later. The job continues in the background.

  • Green LEDs = percentage complete
  • Current product = what's being processed now
  • Succeeded/Failed = track any errors

Step 3: Search

  1. With Semantic Search (AI) enabled, type your query
  2. Press Enter
  3. Results show a Match % column - higher = better match

Getting Better Results

Search quality depends on your product descriptions. Two ways to improve:

Option 1: Enrich Products

Use Enrich with AI to generate detailed, search-optimized descriptions that include:

  • Synonyms (e.g., "ceiling mic", "overhead microphone", "beamforming array")
  • Use cases ("conference room", "huddle space")
  • Technical specs and compatible equipment

Option 2: Add Tags

Add searchable keywords to the Tags field (e.g., "wireless, conference, USB").

After Enriching: Regenerate Embeddings

When you enrich a product, its embedding is cleared. Go back to Step 1 and click Generate Embeddings again to include the new content. You can enrich products at any time - just regenerate afterward.

What's Indexed

Semantic search combines these fields into a searchable embedding:

  • Make, Model, Description, Long Description
  • Enriched Description (if available - best for search)
  • Product Type, Room Type, Category
  • Specifications, Tags

Credits

  • Embedding Setup: ~1 credit per product (one-time)
  • Searching: 1 credit per semantic search

Product embeddings persist until you delete the product or re-enrich it. Each semantic search requires embedding your query, which costs 1 credit.