Catalog Optimize

Transform marketing copy into agent-readable product data.

Overview

The hyperfold catalog optimize command rewrites product descriptions for AI agent comprehension. Marketing-heavy copy full of superlatives and promotional language is converted to fact-dense, structured content that agents can parse efficiently.

# Optimize entire catalog for agent readability
$ hyperfold catalog optimize

> [Scan] Found 1,489 products
> [Job] Starting batch optimization job: job_opt_445
> [Progress] Optimizing descriptions...
> [Progress] Analyzing images...
> [Progress] Generating embeddings...

✓ Optimization complete

SUMMARY
  Products processed: 1,489
  Descriptions improved: 892
  Images analyzed: 2,847
  Embeddings updated: 1,489

READABILITY SCORE
  Before: 0.62
  After: 0.89 (+43%)

Optimization creates a parallel "agent-readable" version of your descriptions. Original marketing copy remains unchanged for human-facing channels.

Agent Optimization

Different AI models process text differently. Optimize for specific buyer agents:

# Optimize for a specific buyer agent model
$ hyperfold catalog optimize --target="gpt-4o"

> [Config] Target model: gpt-4o
> [Style] Optimizing for: concise, fact-dense, structured

✓ Optimized for GPT-4o comprehension

# Available targets
$ hyperfold catalog optimize --list-targets

TARGET          STYLE
gpt-4o          Concise, structured, fact-dense
gpt-4o-mini     Ultra-concise, key facts only
claude-3        Detailed, contextual, nuanced
gemini-pro      Structured, attribute-focused
generic         Balanced for all agents (default)

Optimization Styles

Target	Description
`gpt-4o`	Concise bullet points, explicit specifications, structured format
`gpt-4o-mini`	Ultra-compact, key facts only, minimal prose
`claude-3`	Detailed context, nuanced descriptions, use-case focused
`gemini-pro`	Structured attributes, comparative context, feature lists
`generic`	Balanced approach suitable for any agent (default)

Description Rewriting

See how optimization transforms product descriptions:

# Example transformation during optimization

BEFORE (Marketing-focused):
"Our AMAZING AeroRun X2 will TRANSFORM your running experience!
 These incredible shoes feature cutting-edge technology that
 professional athletes LOVE! ⭐⭐⭐⭐⭐ Don't miss out - BUY NOW
 and get FREE shipping! Limited time offer!"

AFTER (Agent-optimized):
"AeroRun X2 Marathon Running Shoe

 Specifications:
 - Weight: 280g
 - Heel drop: 8mm
 - Upper: Gore-Tex waterproof mesh
 - Sole: EVA foam with carbon plate
 - Waterproof rating: IPX4

 Best for: Marathon, trail running, wet conditions
 Target user: Competitive runners, athletes

 Care: Machine washable, air dry"

Selective Optimization

Optimize specific products or filter by criteria:

# Optimize specific products
$ hyperfold catalog optimize --product="prod_aero_x2"

# Optimize products matching criteria
$ hyperfold catalog optimize --filter="category:footwear"
$ hyperfold catalog optimize --filter="updated_after:2025-12-01"

# Optimize products with low readability scores
$ hyperfold catalog optimize --filter="readability:<0.7"

# Preview changes without applying
$ hyperfold catalog optimize --dry-run

> [Preview] prod_001: Description rewrite (score: 0.45 → 0.91)
> [Preview] prod_002: Image tags added (5 new tags)
> [Preview] prod_003: No changes needed (score: 0.88)

Products with readability scores above 0.85 are skipped by default. Use --force to re-optimize all products.

Image Analysis

Extract visual attributes from product images using multimodal AI:

# Analyze product images for visual tags
$ hyperfold catalog optimize --images-only

> [Vision] Analyzing 2,847 product images...
> [Extract] Extracting colors, styles, materials...

EXAMPLE OUTPUT for prod_aero_x2:
  Primary color: Blue (#2563EB)
  Secondary: White, Black
  Style tags: athletic, modern, streamlined
  Material: mesh, synthetic, rubber
  Features: reflective elements, thick sole

# Force re-analysis of all images
$ hyperfold catalog optimize --images-only --force

# Skip image analysis (faster)
$ hyperfold catalog optimize --no-images

Extracted Attributes

Attribute	Description
Colors	Primary, secondary, and accent colors with hex codes
Style Tags	Aesthetic descriptors (modern, vintage, minimalist)
Materials	Visible material types (leather, metal, fabric)
Features	Visual features (buttons, zippers, patterns)
Vibe	Lifestyle/mood associations (professional, casual)

Embedding Refresh

Update vector embeddings for improved semantic search:

# Refresh vector embeddings only
$ hyperfold catalog optimize --embeddings-only

> [Embed] Regenerating embeddings for 1,489 products...
> [Model] Using: text-embedding-3-large
> [Index] Updating Vertex AI Vector Search...

✓ Embeddings refreshed

# Use specific embedding model
$ hyperfold catalog optimize --embeddings-only --model="text-embedding-3-small"

# Check embedding health
$ hyperfold catalog embeddings status

EMBEDDING STATUS
  Total products: 1,489
  With embeddings: 1,489 (100%)
  Stale (>30 days): 23
  Missing: 0

INDEX STATUS
  Vector dimension: 1536
  Index type: HNSW
  Last rebuild: 2025-12-18T14:30:00Z

Test your optimized catalog with semantic search.

Search Documentation