Documentation

AI Product Descriptions

Automatically generate product descriptions, part numbers, and specifications using AI and real-time web search.

Two AI Features: Fill vs. Enrich

The system offers two complementary AI features for product data:

Fill with AI Enrich with AI
Purpose Quick product info lookup Detailed search optimization
Output Short description (<100 chars) Comprehensive (300-600 words)
Use when Adding new products Improving search results
Time 5-15 seconds 10-30 seconds per product
Cost ~1-2 credits ~1-3 credits

Typical workflow: Fill with AI (new products) → Enrich with AI → Generate Embeddings

Fill with AI: Quick Product Lookup

Enter whatever product information you have, and the AI searches manufacturer websites in real-time to find accurate details.

How to Use

  1. Go to a product's Edit page (or create a new product)
  2. Enter at least one of:
    • Make - e.g., "Shure", "QSC", "Crestron"
    • Model - e.g., "MXA920", "Core 110f"
    • Part# - e.g., "MXA920-W-60"
  3. Click "Fill with AI" (below the Part# field)
  4. Wait 5-15 seconds while the AI researches
  5. Review generated content:
    • Part Number - filled only if the field was empty
    • Description - concise summary (always updated)
    • Long Description - detailed overview (always updated)
  6. Click Save to keep the content
Input Flexibility
  • Minimum required: Just one field (Make, Model, OR Part#)
  • Best results: More info = more accurate descriptions
  • Corrections: "shure mxa 920" → "Shure" + "MXA920" + "MXA920-W"

AI Suggestions & Corrections

If the AI detects incorrect formatting or spelling, a blue "AI Suggestion" banner appears:

  • Shows recommended corrections for Make, Model, and/or Part#
  • Click Apply to accept suggestions
  • Review before accepting - changes aren't applied until you click Apply
Credits Required

Fill with AI uses ~1-2 credits per product based on actual token usage (minimum 1 credit). If you don't have sufficient credits, you'll see: "Insufficient credits. Required: X, Available: Y"

Enrich with AI: Better Search Results

Enrichment creates structured, keyword-dense descriptions (120-180 words) optimized for semantic search. When customers search "ceiling microphone for conference room," enriched products like the Shure MXA920 will surface - even without knowing the exact model name.

Structured Enrichment Format

Enriched descriptions use a structured format that maximizes search accuracy:

[PRODUCT] Shure MXA920
[TYPE] Ceiling array microphone, 24-inch, ceiling mount
[FUNCTION] Voice capture for conferencing
[DESCRIPTION] Ceiling-mounted beamforming microphone array that captures clear audio from meeting participants without table microphones.
[SPECS] 8 steerable lobes, 60-degree pickup, Dante/AES67, IntelliMix DSP, PoE+
[CONNECTIVITY] Dante, AES67, analog out, PoE+ 802.3at
[APPLICATIONS] Conference room, boardroom, training room, huddle space
[COMPATIBLE] Shure IntelliMix, Biamp Tesira, QSC Q-SYS, Crestron
[KEYWORDS] ceiling mic, beamforming array, overhead microphone, MXA series

What Each Field Provides

  • [DESCRIPTION] - Human-readable sentence explaining what it does
  • [SPECS] - Key technical specifications with values
  • [CONNECTIVITY] - Protocols and connection types (Dante, HDMI, PoE, etc.)
  • [APPLICATIONS] - Room types where the product is typically used
  • [COMPATIBLE] - Ecosystems and common equipment pairings
  • [KEYWORDS] - Synonyms and alternative search terms

Enriching a Single Product

  1. Go to a product's Edit page
  2. Scroll to the Enriched Description section
  3. Click "Enrich with AI"
  4. Wait 10-30 seconds while the AI generates content
  5. Review the generated text
  6. Click Save to persist the enrichment

Credits used and token count are displayed after generation completes.

Not happy with the result?

Click "Enrich with AI" again to regenerate. You can regenerate multiple times before saving - only the final version is kept.

Bulk Enrichment

Enrich multiple products at once from the Products page:

  1. Go to the Products page
  2. Select products using checkboxes (no limit - cancel anytime)
  3. Configure options:
    • Re-enrich - regenerate enrichment for products that already have it (default: skip existing)
    • Skip web search - uses existing Long Description instead of web lookup (saves credits)
  4. Click "Enrich"
  5. Progress modal shows status for each product (~1 second per product)
  6. Click "continue working in new tab" to keep working while enrichment runs
  7. Total credits and tokens displayed when complete
Skip Web Search Option

When enabled, products with Long Descriptions over 50 characters use that content as context instead of web search. This saves credits while still generating optimized enriched descriptions. Products with short or missing Long Descriptions will still use web search.

Bulk Enrichment Credits

The badge shows an estimate of ~1 credit per product. Actual costs vary based on product complexity and whether web search is used. Important: If you run out of credits mid-batch, processing stops immediately. Products enriched before the error are saved.

After Enrichment: Update Embeddings

Enriched descriptions are stored separately and won't overwrite your regular descriptions. To include them in semantic search:

  1. When you Save an enriched product, its embedding is automatically cleared
  2. Go to the Products page and run Generate Embeddings
  3. New embeddings will incorporate the enriched content for better search matching
Best Practices
  • Use official model numbers for the most accurate results
  • Always review AI-generated content before saving
  • Enrich products before generating embeddings
  • For large catalogs, process in batches of 50

Troubleshooting

  • "Insufficient credits" - Purchase more credits at Billing > Credits (coin icon in top bar).
  • No enrichment returned - AI couldn't generate content. Check that the product has valid Make/Model information and try again.
  • Product not found - If the AI can't find the product online, it generates a basic description from your input. You can edit and refine manually.